This is the fifth article of an&nbsp;<a href="https://www.wevolver.com/article/what-is-a-collaborative-robot" rel="noopener noreferrer" target="_blank">eight-part series</a>&nbsp;that explores how collaborative robots or cobots are transforming industrial workspaces. It will survey the technologies that have converged to make robotic influence possible, unpack the unique engineering challenges robotics pose, and describe the solutions that are leading the way to the influential robots of the future.The articles were originally published in an e-magazine, and have been substantially edited by Wevolver to update them and make them available on the Wevolver platform. This series is sponsored by Mouser, an online distributor of electronic components. Through their sponsorship,&nbsp;<a href="https://www.mouser.com/" rel="nofollow" target="_blank">Mouser Electronics</a>&nbsp;supports spreading knowledge about collaborative robots.<h2 data-element-id="headingsMap-4-0">Introduction</h2>A new generation of robots, termed “collaborative robots” or “cobots,” is now being integrated into manufacturing environments. Rather than being used to replace human labor, these robots are being used to work together with humans in a common workspace. These cobots are typically used on tasks that require improved process control and/or greater flexibility, scalability, and agility. To enable cobots to work together with the human workforce, numerous improvements have been incorporated to enhance worker safety. The result is increased productivity and better labor utilization at lower overall costs. Consequently, these cobots are affordable for small- and medium-sized companies to support lower-speed, lower-volume production runs. This article examines some of the technological advancements that enable collaborative robotics and provides some examples of how these cobots are being used in a<a href="https://www.wevolver.com/article/industrie.40.deep.dive.part.5.flexible.manufacturing" rel="noopener noreferrer" target="_blank">&nbsp;manufacturing environment.</a><h2 data-element-id="headingsMap-5-0">Enabling Technologies</h2>Collaborative robotic platforms incorporate numerous enabling technologies that differentiate them from traditional robots and enhance human/robotic interaction, improve worker safety, and reduce costs. Improvements in sensor, actuator, control, and computation technologies empower the<a href="https://www.wevolver.com/article/the-complexities-of-mimicking-humans-is-just-the-beginning" rel="noopener noreferrer" target="_blank">&nbsp;cobot’s perception</a>, data fusion and processing, artificial intelligence, and actuation. Other technologies promote ease of use, adaptability, and lower costs. A brief overview of some of these technologies follows:<h3 data-element-id="headingsMap-6-0">Perception</h3>Technologies that improve perception include sensors that can detect and measure any physical property, such as distance, proximity, force, torque, velocity, sound, vibration, volume, weight, surface finish, texture, chemical composition, temperature, electrical current, and magnetism. Other technologies, such as machine vision, allow the cobot to detect and identify objects, spaces, scenes, orientations, and locations of objects in the visible, infrared, and ultraviolet wavelengths. Together, these perception technologies allow improved accuracy, higher throughput, collision avoidance, and situational awareness. Voice recognition and motion tracking technologies may also allow operators to control the cobot through voice commands and hand gestures.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5MDM0MTY5Mjk4LXNpY2suanBnIiwgImVkaXRzIjogeyJyZXNpemUiOiB7IndpZHRoIjogOTUwLCAiZml0IjogImNvdmVyIn19fQ==">Cobots can be equipped with a variety of sensors. Image:&nbsp;<a href="https://www.sick.com/at/en/the-new-mobility-how-sensors-control-the-cobots-of-the-future/w/blog-sensors-cobots/" rel="noopener noreferrer" target="_blank">Sick</a><h3 data-element-id="headingsMap-7-0">Data Fusion and Processing</h3>Raw data from numerous sensors must be integrated, processed, and interpreted to enable the cobot to make informed and accurate deductions about the environment so that appropriate actions can be taken. This must be rapidly accomplished to avoid undue delays in the robot’s operation, especially if the robot needs to interrupt its operation to avoid a collision with an object or human. Sensor-fusion algorithms must effectively integrate data from different types of sensors, which may come from different locations within the workspace or may be sensed at different times to make informed and accurate deductions about the environment and situation.<h3 data-element-id="headingsMap-8-0">Artificial Intelligence</h3>A traditional robot operates according to preprogramed instructions that are repeated without regard to changes in its operational environment. A cobot, on the other hand, incorporates <a href="https://www.wevolver.com/article/industrie.40.deep.dive.part.1.introducing.smart.manufacturing" rel="noopener noreferrer" target="_blank">artificial intelligence (AI)</a>, which encompasses its ability to perceive its environment; reason and solve problems; recognize, learn, and adapt to new situations and environments; autonomously make decisions; and interact with humans. The “programming” of a cobot to operate in a new environment is not necessarily complicated and can simply involve a human operator manually walking through the desired motion by moving the cobot’s arms through the desired trajectory. AI is enabled through high-speed computer processors and software that may incorporate a variety of technologies such as fuzzy logic, probabilistic methods, neural networks, and expert systems. These technologies provide cobots with human-like qualities that promote human-robot interaction and collaboration.<h3 data-element-id="headingsMap-9-0">Actuators</h3>While traditional robots have very stiff joints with high force, high torque actuation devices, cobots have compliant joints that can provide smooth or soft motions and forces, so they may seamlessly interact with their human counterparts. Compliant actuators do not require expensive, high-precision drive mechanisms or motion sensors, as they rely on force or current feedback to compensate for potential errors in position accuracy. Furthermore, since cobots are lighter and designed to work with humans in a shared workspace, they do not need the expensive high-force actuation devices or motor drives that are commonly used in traditional robots.<h2 data-element-id="headingsMap-10-0">Applications</h2>New applications for collaborative robotics within manufacturing continue to develop and expand. Because cobots are affordable, highly adaptable, and almost plug-and-play capable, small- and medium-sized manufacturers are eager to take on this technology. Some of these notable applications include the following:<h3 data-element-id="headingsMap-11-0">Machine Tending</h3>A typical production machine, such as the kind used for computer numerical control (CNC) machining, must continuously be tended by an operator who will load raw material, start the machine, ensure its proper operation, and unload finished parts. Machine tending is one of the most common industrial jobs. Often monotonous and boring, it does not require a high skill level, and operators are prone to injury due to the repetitive lifting of heavy materials and parts. Other examples of machine tending operations include traditional machining operations (e.g., stamping, grinding, trimming, and welding), injection molding, printed circuit board (PCB) testing, part inspecting, and coordinate measuring.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5MDM0MzM5MTU5LTA0MDIgKDEpLnBuZyIsICJlZGl0cyI6IHsicmVzaXplIjogeyJ3aWR0aCI6IDk1MCwgImZpdCI6ICJjb3ZlciJ9fX0=">A cobot tending a CNC machine. Image:&nbsp;<a href="http://www.robotiq.com" target="_blank">Robotiq</a>Cobots may be an ideal fit for machine tending applications. They can be placed in front of a traditional manufacturing machine and easily programmed to perform the required loading and unloading tasks. Periodically, skilled technicians can examine the machine and tools and perform regular maintenance. This can lead to better workforce utilization, where the cobots are used for the lower-skill-level manual operations and technicians are used for the higher-skill-level tasks such as inspecting, repairing, and changing tools, performing quality control tests, and taking appropriate corrective actions. This, in turn, can result in significant labor savings, lower injury rates, less production downtime, a more motivated and skilled workforce, and higher productivity.<h3 data-element-id="headingsMap-12-0">Packaging and Palletizing</h3>Packaging and palletizing tasks typically involve the removal of products from an assembly line or conveyor belt, counting and wrapping them appropriately, applying labels, placing them into cartons, and sealing the cartons. High-speed, high-volume manufacturers can create specialized robotic machines and equipment for packaging. On the other hand, small to medium companies may still require manual packaging operations since automated equipment may not be cost-effective due to limited production runs or variations in product form factors.Cobots may be an ideal fit for such packaging and palletizing applications. They can be easily programmed to recognize and count items on a conveyor belt, apply packaging materials and labels, and place them into the appropriate cartons. Human workers can still work with the cobots while performing higher-skill-level tasks such as quality-control inspections. Programming for the cobot can be easily modified to adapt to changes in the product configuration or packaging requirements.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5MDM0NDAwNDM3LTA0MDMgKDEpLmpwZyIsICJlZGl0cyI6IHsicmVzaXplIjogeyJ3aWR0aCI6IDk1MCwgImZpdCI6ICJjb3ZlciJ9fX0=">Cobot in packaging application. Image:&nbsp;<a href="http://www.robotiq.com" target="_blank">Robotiq</a><h3 data-element-id="headingsMap-13-0">Collaborative or Adaptive Assembly</h3>Traditional (pick and place) robots require the precise location of the unfinished assemblies and parts. Cobots, on the other hand, are adaptable to non-precise placement, as they might incorporate vision systems that facilitate position feedback. A collaborative or adaptive assembly operation may involve use of a human operator to first inspect, sort, and select a particular part then roughly align the part onto an unfinished assembly (that is, a work-in-progress). In a subsequent assembly step, a cobot would complete the assembly by fastening the part to the unfinished assembly. An operation such as this might be advantageous in situations where parts are too small to align with unaided human vision; or where the final alignment and attachment of the part requires a great deal of dexterity or strength; or where final assembly must be performed at high temperatures, in a difficult environment, or with hazardous chemicals or adhesives. A cobot would, therefore, be useful if the location, orientation, or placement of parts onto an unfinished assembly is not accurate or repeatable.<h3 data-element-id="headingsMap-14-0">Human Augmentation</h3>Collaborative robots can be used to augment human capabilities during manufacturing and assembly operations by assisting in the performance of menial (repetitive) tasks or tasks that require strength or extended reach. For example, certain manufacturing operations might require lifting and positioning a heavy part, like an automobile door, onto an unfinished assembly, like an automobile body. Such a process might not lend itself to a completely automated operation if it requires fine adjustments to the part position to ensure the part properly fits onto the unfinished assembly. For example, in an automobile assembly operation, different vehicle styles might require varying amounts and styles of doors for installation, which may not be a suitable situation for a completely automated process. However, a cobot can be useful in this situation to lift the heavy part. It could then carry out “instructions” from the human operator by responding to the operator’s touch and move to position the part on the desired location of the unfinished assembly.<h2 data-element-id="headingsMap-15-0">Conclusion</h2>These are only a few examples that illustrate the role of collaborative robotics in manufacturing. As enabling robotic technologies in the areas of sensors, actuators, controls, and computation continue to evolve, we expect that more and more manufacturers will continue to adopt collaborative robotic systems into their workplaces. What might the eventual role of cobots be in manufacturing environments? The advancement in its technology and applications continue to unfold.This article was originally written by Phi Hipol for Mouser and substantially edited by the Wevolver team. It's the fifth article of a series exploring collaborative robots.&nbsp;Article One&nbsp;<a href="https://www.wevolver.com/article/what-is-a-collaborative-robot" rel="noopener noreferrer" target="_blank">introduces collaborative robots.</a>Article Two&nbsp;describes&nbsp;<a href="https://www.wevolver.com/article/a-history-of-industrial-robots" rel="noopener noreferrer" target="_blank">the history of Industrial Robots.</a>Article Three&nbsp;gives an&nbsp;<a href="https://www.wevolver.com/article/the-complexities-of-mimicking-humans-is-just-the-beginning" rel="noopener noreferrer" target="_blank">overview of collaborative robots sensor technologies.</a>Article Four&nbsp;<a href="https://www.wevolver.com/article/designing-collaborative-robots-maximizing-productivity-and-safety" rel="noopener noreferrer" target="_blank">examines the balance between cobot safety and productivity.</a>Article Six&nbsp;explores the&nbsp;<a href="https://www.wevolver.com/article/programmable-versus-fixed-function-controllers-alternatives-for-complex-robotic-motion" rel="noopener noreferrer" target="_blank">challenges of motion control of robotic arms.</a>Article Seven l<a href="https://www.wevolver.com/article/cobots-are-collaborators-ai-will-make-them-partners" rel="noopener noreferrer" target="_blank">ooks at the potential of AI in cobots.</a>Article Eight&nbsp;<a href="https://www.wevolver.com/article/social-impact-of-robotics-in-the-workplace-the-arrival-of-the-perfect-coworker" rel="noopener noreferrer" target="_blank">discusses the social impact of cobots.</a><hr><h2 data-element-id="headingsMap-16-0">About the sponsor: Mouser Electronics</h2><a href="https://www.mouser.com/" rel="nofollow" target="_blank">Mouser Electronics</a>&nbsp;is a worldwide leading authorized distributor of semiconductors and electronic components for over 800 industry-leading manufacturers. They specialize in the rapid introduction of new products and technologies for design engineers and buyers. Their extensive product offering includes semiconductors, interconnects, passives, and electromechanical components.<img class="fr-fic fr-dii" src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjg1NDQ1NDcyNzI5LTE2ODU0NDU0NzI3MjkucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==">

Article 5 of our Cobot Series: Traditionally, robotics have served limited roles in manufacturing environments; however, various technologies have converged to make robots work well as collaborators with humans and to succeed in various applications.

Technologies Converge to Help Drive Manufacturing Applications

This is the sixth article of an&nbsp;<a href="https://www.wevolver.com/article/what-is-a-collaborative-robot" rel="noopener noreferrer" target="_blank">eight-part series</a>&nbsp;that explores how collaborative robots or cobots are transforming industrial workspaces. It will survey the technologies that have converged to make robotic influence possible, unpack the unique engineering challenges robotics pose, and describe the solutions that are leading the way to the influential robots of the future.The articles were originally published in an e-magazine, and have been substantially edited by Wevolver to update them and make them available on the Wevolver platform. This series is sponsored by Mouser, an online distributor of electronic components. Through their sponsorship,&nbsp;<a href="https://www.mouser.com/" rel="nofollow" target="_blank">Mouser Electronics</a>&nbsp;supports spreading knowledge about collaborative robots.<h2 data-element-id="headingsMap-4-0">Introduction</h2>Motion control of today's sophisticated robotic arms, regardless of their size or power, often requires simultaneous management of multiple axes. Modern electronics—including motors, device drivers, control systems (formerly all analog, but now digital), feedback sensors, and power-switching devices such as metal-oxide semiconductor field-effect transistors (MOSFETs) and insulated-gate bipolar transistors (IGBTs)—now make achieving precise motion control easier than it was just a few years ago. At the same time, however, the demands on system performance have increased dramatically, so the overall project of implementing motion control is as difficult as ever.Nonetheless, there's one unavoidable fact: <a href="http://www.mouser.com/applications/robotics-technology/" target="_blank">Robotics</a> is largely a mechanical function, so in reality mechanical systems must be configured into the control loop. These systems include gear backlash, mechanical tolerances, vibration, motor performance, mass inertia rotation, momentum, mechanical structure flexibility, variable loads, and more. For these reasons, it’s important to choose the type of motor that best suits the robotic operation you’re trying to achieve—usually the best choice is a <a href="http://www.mouser.com/applications/motor-control/" target="_blank">brushless DC motor</a> or stepper motor in low to moderate power situations.Another vital decision relates to sensor-based feedback. Most robotic applications use some type of feedback sensor to accurately gauge the end-effector's position and thus its velocity and acceleration (recall that velocity is the time integral of position, and acceleration is the time integral of velocity). This feedback transducer can be a Hall-effect sensor, a synchro/resolver, or an optical encoder. While it’s easiest to put the encoder on the motor, because of the mechanical issues noted above, placing it there may not produce sufficiently accurate data about the end-effector's actual situation for the application. Therefore, the sensor may need to be mounted closer to the load endpoint.Some motion-control applications operate without a sensor, which reduces the cost and mechanical complexity. Rather than using a sensor for feedback, sensorless field-oriented control (FOC, also called vector control) uses precise, synchronized readings of the current and voltage at each phase of the motor windings; FOCs then perform complicated frame-of-reference transformations and matrix calculations in real time to determine a motor's position. Eliminating the sensor reduces the hardware cost, but it necessitates significant computational capability and more complex programming. Many robotic designs still prefer to use sensors because FOC doesn’t provide the same level of confidence, credibility, and robustness that using a direct-sensor readout offers.<h2 data-element-id="headingsMap-5-0">Understanding Basic Robotic Configurations</h2>While the general public may associate the term "robot" with a mobile, lifelike servant or assistant, most robotic systems in the industrial domain are stationary and use a variety of mechanical arms and configurations to perform tasks. Among the most common arrangements are:The Cartesian robot, which has three linear motion axes, one for each of the x-, y-, and z-planes. This robot is used for pick and place machine operations, sealant applications, and basic assemblies.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDcwNDc4NjIwLWNhcnRlc2lhbi5wbmciLCAiZWRpdHMiOiB7InJlc2l6ZSI6IHsid2lkdGgiOiA5NTAsICJmaXQiOiAiY292ZXIifX19">Figure 1: The Cartesian robot is the easiest to comprehend and control because it has the simplest &nbsp;equations and works in the x-, y-, and z-planes. Image: MachineDesignIn a cylindrical robot, all motion is confined to a cylinder-shaped zone. It combines linear motion in the y-plane, linear motion in the z-plane, and rotational motion around the z-axis. This robotic arrangement is used for assembly, tool handling, and spot welding.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDcxNDg4OTMwLUlMTFVTVFJBVElPTiAxNSAtMDEtMDEgY3JvcHBlZC5qcGciLCAiZWRpdHMiOiB7InJlc2l6ZSI6IHsid2lkdGgiOiA5NTAsICJmaXQiOiAiY292ZXIifX19">Figure 2: The cylindrical robot has motion along two linear axes and around one rotational axisThe spherical or polar robot combines two rotary joints and one linear joint, and the arm is connected to the base with a twisting joint (Figure 3). Its motion is defined by a polar coordinates system and confined to a spherical zone. This robot is found in welding, casting, and tool-handling applications.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDcxNjU5NTk3LUlMTFVTVFJBVElPTiAxNyBGaW5hbC0wMSBjcm9wcGVkLmpwZyIsICJlZGl0cyI6IHsicmVzaXplIjogeyJ3aWR0aCI6IDk1MCwgImZpdCI6ICJjb3ZlciJ9fX0=">Figure 3: The spherical or polar robot combines motion around two rotary axes and along one linear axis, and it requires numerous calculation-intensive transformations between coordinate frames of referenceThe robotic approaches cited here offer three degrees of freedom, using a combination of linear and rotary motion. However, some applications need only one or two degrees. More advanced robotic arms, or articulated robots, combine additional linear and rotary motion for almost human-like dexterity and flexibility (Figure 4). Some high-tech arms provide six, eight, or even more degrees of freedom.<img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDcyNDYzODExLUlMTFVTVFJBVElPTiAyMCBGaW5hbC0wMSBjcm9wcGVkLmpwZyIsICJlZGl0cyI6IHsicmVzaXplIjogeyJ3aWR0aCI6IDk1MCwgImZpdCI6ICJjb3ZlciJ9fX0=">Figure 4: The articulated robotic arm combines multiple rotation and linear motion modes for many&nbsp; degrees of freedom, but it also requires careful coordination among the actuators and arms.Other designs use special combinations of linear and rotary motion for application-specific situations, such as the parallelogram implementation, an implementation used for precise and rapid motion over short distances (for example, to pick and place tiny components). As the number of degrees of freedom increases, achieving rapid, smooth, accurate, and synchronized control along each of these degrees grows exponentially more challenging.<h2 data-element-id="headingsMap-6-0">Considering Trajectory Profiles</h2>The motion-control objective in robotics seems simple enough: Have the end-effector optimally reach its target position as quickly and accurately as possible with the supported load. Of course, there are trade-offs involved, as in all engineering decisions, depending on the priorities associated with the optimum result in the given application.For example, is it acceptable to accelerate and decelerate quickly to rapidly reach a higher velocity if the result is overshot and if there is even possible oscillation at the endpoint? Is it worth trading accuracy for speed, and to what extent? How are the choices of acceleration, velocity, and position related to the desired transition from position A to position B? What are the priorities and parameters that define "optimum" in a particular application?Specialists in motion control for robotics and other motion applications have developed standard trajectory profiles that provide various ways to implement the desired trade-off solution for a given application. All choices involve significant real-time calculations based on the present situation and feedback signal, but some impose a more substantial, high-resolution computational burden. These profiles include:<ul><li>The simple trapezoid, where the motor accelerates at a fixed rate from zero to a target velocity, stays at that velocity, and then ramps down at a fixed rate to zero velocity at the desired position (Figure 5). Higher rates might speed up the entire positioning cycle, but they might also induce sudden changes in acceleration motion (called the jerk) that, in turn, adds inaccuracy and overshoot.</li></ul><img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDczMTM5Mzc1LUlMTFVTVFJBVElPTiAxNiBGaW5hbC0wMSBjcm9wcGVkLmpwZyIsICJlZGl0cyI6IHsicmVzaXplIjogeyJ3aWR0aCI6IDk1MCwgImZpdCI6ICJjb3ZlciJ9fX0=">Image 5: The simplest motion-trajectory profile is the trapezoid, which has constant acceleration to the target velocity, constant path velocity, and constant deceleration between the start and endpoints.<ul><li>The S-curve, a frequently-used enhancement to the trapezoid, where the acceleration rate ramps up from zero then decreases as the target velocity is achieved (Figure 6). As the target position is reached, the deceleration rate is ramped up and then reduced as the endpoint is near. The S-curve has seven distinct phases, in contrast to the three phases of the trapezoid.</li></ul><img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDczNjMyOTc3LUlMTFVTVFJBVElPTiAxOCBGaW5hbC0wMS5qcGciLCAiZWRpdHMiOiB7InJlc2l6ZSI6IHsid2lkdGgiOiA5NTAsICJmaXQiOiAiY292ZXIifX19">Figures 6: S-curve path is more complicated than the basic trapezoid, but it eases the jerk (change in acceleration) at each transition point of the path.<ul><li>The&nbsp;contoured motion, where the user establishes a set of desired positions, and the motion controller directs a smooth, jerk-free transition profile through all these positions (Figure 7). This allows the ultimate flexibility and control, which is necessary for advanced-motion situations. The required control-direction calculations to achieve smooth curve-fitting are complex and must be accomplished without the loss of resolution due to rounding or truncating errors, despite the many calculations.</li></ul><img src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk5NDczNzk3MjU0LUlMTFVTVFJBVElPTiAxOSBGaW5hbC0wMSBjcm9wcGVkLmpwZyIsICJlZGl0cyI6IHsicmVzaXplIjogeyJ3aWR0aCI6IDk1MCwgImZpdCI6ICJjb3ZlciJ9fX0=">Figure 7: The contoured-motion path allows the user to define a series of position-marker points between the start and endpoints, and the controller must guide the end-effector through these points in a smooth curve.There are other profiles in use, some of which are associated with specific application groups or industries. Regardless of the desired profile, it's one thing to want it and another to make it happen. The well-known, highly effective proportional-integral-derivative (PID), closed-loop, control algorithm is the most common approach used to drive a motor and end-effector to do what one desires with a high-enough-level of accuracy and precision.<a href="https://www.wevolver.com/article/designing-collaborative-robots-maximizing-productivity-and-safety" rel="noopener noreferrer" target="_blank">Effective control</a> of a single axis is a manageable project, but robotic control becomes far more difficult when this control extends to two, three, or even more motors and degrees of freedom, which must be closely coordinated and synchronized, with the performance one element dependent on the status of the others.<h2 data-element-id="headingsMap-7-0">Determining Standard versus Custom Motion-Control Applications</h2>For standard motion-control applications, a dedicated, fixed-function, embedded controller integrated circuit (IC) offers ease of use and rapid time-to-market. In contrast, if a nonstandard, customized profile is necessary or if the correlation between the various axes is complicated and must accommodate unusual or unique events, then the Design team may consider a fully user-programmable processor. This solution is implemented using a processor with digital-signal processor (DSP) capabilities for the computation-intensive aspects or with a field-programmable gate array (FPGA). When considering programmable devices, the vendor, third-party tools, and available software modules are factors of importance in addition to the hardware functions of the IC itself in making a specific selection.Be aware that these controllers are generally not the same as motor drivers, which are the MOSFET/IGBT drivers and devices that control motor power. This is the case for two reasons: First, these power devices must be sized to the motor, independent of the controller. Second, the high-density, complementary, metal-oxide-semiconductor-based process technologies used for these digital controllers are very different than the processes for power devices. For smaller motors, however, it is possible to integrate a controller with a driver and power device. Despite the fundamental differences, the term "controller" often refers to the power-device functional blocks, which can lead to confusion in keyword searches.<h2 data-element-id="headingsMap-8-0">Conclusion</h2>Motion-control options for robotics range from basic, dedicated-function ICs to highly integrated, extremely flexible microcontroller units (MCUs) with a large array of auxiliary-processing and support functions. Although embedded devices may seem limiting, some of them allow a selection of a variety of motion profiles and the setting of critical parameters, and they are quite adequate, low in cost, and easy to use. For advanced designs with unique or extremely sophisticated requirements or for designs needing additional levels of connectivity along with control, MCUs offer effective solutions with evaluation and development kits supported by verified code packages, debug and code-development tools, and validation suites.This article was originally written by Bill Schweber for Mouser and substantially edited by the Wevolver team. It's the sixth article of a series exploring collaborative robots.&nbsp;Article One&nbsp;<a href="https://www.wevolver.com/article/what-is-a-collaborative-robot" rel="noopener noreferrer" target="_blank">introduces collaborative robots.</a>Article Two&nbsp;describes&nbsp;<a href="https://www.wevolver.com/article/a-history-of-industrial-robots" rel="noopener noreferrer" target="_blank">the history of Industrial Robots.</a>Article Three&nbsp;gives an&nbsp;<a href="https://www.wevolver.com/article/the-complexities-of-mimicking-humans-is-just-the-beginning" rel="noopener noreferrer" target="_blank">overview of collaborative robots sensor technologies.</a>Article Four&nbsp;<a href="https://www.wevolver.com/article/designing-collaborative-robots-maximizing-productivity-and-safety" rel="noopener noreferrer" target="_blank">examines the balance between cobot safety and productivity.</a>Article Five <a href="https://www.wevolver.com/article/technologies-converge-to-help-drive-manufacturing-applications" rel="noopener noreferrer" target="_blank">discusses the development of cobot applications in manufacturing.&nbsp;</a>Article Seven l<a href="https://www.wevolver.com/article/cobots-are-collaborators-ai-will-make-them-partners" rel="noopener noreferrer" target="_blank">ooks at the potential of AI in cobots.</a>Article Eight&nbsp;<a href="https://www.wevolver.com/article/social-impact-of-robotics-in-the-workplace-the-arrival-of-the-perfect-coworker" rel="noopener noreferrer" target="_blank">discusses the social impact of cobots.</a><hr><h2 data-element-id="headingsMap-9-0">About the sponsor: Mouser Electronics</h2><a href="https://www.mouser.com/" rel="nofollow" target="_blank">Mouser Electronics</a>&nbsp;is a worldwide leading authorized distributor of semiconductors and electronic components for over 800 industry-leading manufacturers. They specialize in the rapid introduction of new products and technologies for design engineers and buyers. Their extensive product offering includes semiconductors, interconnects, passives, and electromechanical components.<img class="fr-fic fr-dib" src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNjAwODYzNDkxOTQ3LW1vdXNlciBsb2dvIGZvb3Rlci5qcGciLCAiZWRpdHMiOiB7InJlc2l6ZSI6IHsid2lkdGgiOiA5NTAsICJmaXQiOiAiY292ZXIifX19" style="width: 788px;">

Article 6 of our Cobot Series: Motion control of today's sophisticated robotic arms is as difficult as ever.

Programmable Versus Fixed-Function Controllers: Alternatives for Complex Robotic Motion

Article 4 of our Cobot Series: Collaborative robots are becoming invaluable for industry, as they improve safety and increase productivity, but there are still improvements to be made in design and regulation.

Designing Collaborative Robots: Maximizing Productivity and Safety

The Kosmos IoT system by Temboo together with the Nordic Thingy:52 prototyping platform can help companies seeking to leverage the benefits of Industrial IoT, without the need for experienced coders.In industrial applications, peripheral sensors provide valuable real time data on asset tracking and monitoring, predictive maintenance and smart facilities, which can result in huge gains.Being able to predict whether a manufacturing device, generator or pump requires service reduces downtime. Monitoring where your assets are located increases effectiveness and diminishes wastage. Hourly logs of temperature, humidity or other important variables on a food or beverage production line help ensure quality.There is no doubt that this increase in effectiveness and quality is desirable. However, deploying so many sensors and enabling the data to be transferred to the cloud where it can be processed and analysed is a daunting task for those with no prior knowledge of embedded programming.The <a href="https://temboo.com/platform" target="_blank">Kosmos IoT System</a> by Temboo helps companies make their products, processes and facilities smart, without requiring extensive software or hardware expertise.<h2>Kosmos by Temboo makes IIoT accessible</h2>The Kosmos IoT System works by taking sensor data from a physical environment and relaying that information via a gateway device to the Kosmos Cloud. This is where things get truly accessible. The platform includes drivers for a variety of supported hardware. You can configure which devices you want to use in applications and Kosmos will machine generate the code to be used on the devices.<img class="fr-fic fr-dib" src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk2Nzk3NzU0MDYxLTEyNDRfMzU0MC5qcGciLCAiZWRpdHMiOiB7InJlc2l6ZSI6IHsid2lkdGgiOiA5NTAsICJmaXQiOiAiY292ZXIifX19" style="width: 787px;">This allows you to quickly setup and deploy applications with no software coding and just minimal amount of hardware configuration required. It’s so simple that even people with no prior experience in programming can bring their first IoT system to life within an hour.<h2>Works with Nordic</h2>The&nbsp;<a href="https://blog.nordicsemi.com/getconnected/inside-the-nordic-thingy?utm_campaign=2021%20wevolver" rel="noopener noreferrer" target="_blank">Nordic Thingy:52 prototyping platform</a> is one of the supported devices. The Thingy:52 contains a variety of sensors including temperature, air pressure, air quality measurement, and accelerometers for detecting motion. It uses the nRF52832 Bluetooth 5 qualified System-on-Chip to transfer sensor data to the gateway device of your choice.For a quick look at how simple it can be to tap into Industrial IoT’s potential by using the Kosmos platform together with the Thingy:52, check out this video:<iframe allowfullscreen="" class="fr-draggable" frameborder="0" src="https://www.youtube.com/embed/qL4OyHKhIFk?&amp;wmode=opaque" style="width: 789px; height: 443px;"></iframe><h2>Real-time monitoring and rule-based actions</h2>On the Temboo website you can log in to your application to monitor a visualization of the data from your sensors. Temperature, humidity, CO2 levels, water quality, wear and tear of critical parts in manufacturing devices can all be monitored. In short, all real-time information that is needed for qualified decision making can be observed online.The Kosmos platform also allows you to create action rules based on the data from the sensors. Say for instance you have identified a maximum safe operating temperature for a production line. You can create an SMS or email alert when the sensor reaches that temperature. This alert can then be used to trigger another action, such as shutting down production to allow service.Or better yet, you can receive a warning before the temperature is dangerously elevated, allowing maintenance measures to take place before a costly total shut-down.&nbsp;Innovative organizations in industries like healthcare, smart cities, facilities management, and wastewater treatment use Kosmos for environmental monitoring, alert systems, and sensor data collection.The Kosmos platform is compatible with a range of wireless protocols and types of sensors (<a href="https://temboo.com/kosmos/docs" target="_blank">full details</a>) and they are continuously expanding the portfolio of drivers. Temboo offers a free trial if you would like to test it out and judge just how much IIoT can do for your business.<hr>This article was first published on Nordic's&nbsp;<a href="https://blog.nordicsemi.com/getconnected" rel="nofollow" target="_blank"></a><a href="https://blog.nordicsemi.com/getconnected?utm_campaign=2021%20wevolver" rel="noopener noreferrer" target="_blank">Get Connected Blog</a>.

The Kosmos IoT system by Temboo together with the Nordic Thingy:52 prototyping platform can help companies seeking to leverage the benefits of Industrial IoT, without the need for experienced coders.

A Path to Industrial IoT Without Coding

The RESCURE consortium, consisting of the companies&nbsp;<a href="https://technikon.com/" target="_blank">Technikon</a>&nbsp;and&nbsp;<a href="https://www.intrinsic-id.com/" target="_blank">Intrinsic ID</a>, along with the&nbsp;<a href="https://www.tue.nl/en/" target="_blank">Eindhoven University of Technology (TU/e)</a>, was created to address the need for a security solution that spans the lifecycle of devices in the Internet of Things (IoT). The solution provided by the consortium resolves security vulnerabilities in this lifecycle using SRAM Physical Unclonable Function (PUF) technology in combination with state-of-the-art cryptography and security protocols. The IoT is expanding fast, and so are risks to its security. The fact that IoT devices are deployed frequently with failing security implementations becomes clear with even a quick look at <a href="https://nvd.nist.gov/vuln" target="_blank">NIST’s National Vulnerability Database</a>. This database includes instances of devices with <a href="https://nvd.nist.gov/vuln/detail/CVE-2019-17391" target="_blank">poorly hidden</a> or <a href="https://nvd.nist.gov/vuln/detail/CVE-2018-16546" target="_blank">repeated</a> secret keys, <a href="https://nvd.nist.gov/vuln/detail/CVE-2019-13523" target="_blank">improper</a> – or <a href="https://nvd.nist.gov/vuln/detail/CVE-2020-6769" target="_blank">absent</a> – authentication, and <a href="https://nvd.nist.gov/vuln/detail/CVE-2020-9544" target="_blank">unguarded software update conduits</a>, to name just a few security issues known to exist in devices already in the field.The consequences of any IoT breach can be severe as IoT devices are often used in safety-critical applications, such as transportation systems, smart electricity grids, and water-supply systems. Compromising a single device, even a seemingly insignificant one, can bring down complete infrastructures, violate the privacy of millions of people, and disclose confidential data. Because IoT devices typically run autonomously across a broad range of untrusted environments, it can be very difficult to detect an attack quickly, and very costly to physically access the devices for repairs.Protecting an IoT device is complex. It not only has to be protected in the field, but during its entire complex and far-flung lifecycle. Various phases of the lifecycle of an IoT device involve different parties with a variety of security needs. Often, there is no one party among these varied players that has any incentive, expertise, or even ability to completely take care of security.<h1>RESCURE</h1>RESCURE is a project funded by the EU and the EUREKA programme <a href="https://www.eurostars-eureka.eu/" target="_blank">Eurostars</a> (Grant: E11897, project duration: February 2018 - April 2020). The goal of the project, consisting of the companies <a href="https://technikon.com/" target="_blank">Technikon</a> and <a href="https://www.intrinsic-id.com/" target="_blank">Intrinsic ID</a>, along with the <a href="https://www.tue.nl/en/" target="_blank">Eindhoven University of Technology (TU/e)</a>, has been to provide a flexible framework that allows IoT device security to be updated throughout the entire lifecycle of the device.<iframe width="640" height="360" src="https://www.youtube.com/embed/MJ7lsTwl2iE?&amp;wmode=opaque" frameborder="0" allowfullscreen="" class="fr-draggable"></iframe> RESCURE achieves this by retrofitting security on existing IoT devices using a low-cost solution, based on SRAM PUF technology. Tiny uncontrollable variations in the manufacturing of SRAM transistors lead to SRAM start-up behavior that is unique for every individual chip. SRAM PUF technology uses SRAM start-up data as a “silicon fingerprint” for the microcontroller unit (MCU), turning this unique property of SRAM into an unclonable device identity by deriving a device root key from the silicon fingerprint. This root key is never stored on the device, but instead it is generated only when needed.The main advantages gained from this method of key generation and storage are: &nbsp;<ul><li>High security, as no root key is programmed externally and it is never stored</li><li>Low cost, as no costly protected memory or other security hardware is required</li><li>High flexibility, as it is even possible to retrofit this security on deployed devices</li></ul><h1>Security for the Complete Device Lifecycle</h1>Using this technology, RESCURE has created a working prototype of its security architecture that protects an IoT device throughout its lifecycle:<ul><li>Manufacturing – Manufacturers need to make sure each device has a unique identity. This identity is used to connect securely to a cloud service, typically using a Public Key Infrastructure (PKI). The devices also need to have secure storage for sensitive data, such as IoT measurement data and valuable software IP.</li></ul>In the RESCURE security architecture, an SRAM PUF is enrolled on the device in the manufacturing stage, allowing it to create device-unique keys. These include public/private keypairs for setting up secure cloud connections, as well as symmetric keys that are used for protecting all valuable data and IP on the device. These secret keys are only available at runtime and are never stored on the device, so they can never be stolen. The silicon fingerprint of each device is unique; and because the secret keys are derived from this unique fingerprint, there is no possibility of secret keys from one device being copied to another device.<img class="fr-fic fr-dib" src="https://images.wevolver.com/eyJidWNrZXQiOiAid2V2b2x2ZXItcHJvamVjdC1pbWFnZXMiLCAia2V5IjogImZyb2FsYS8xNTk0MTExMjUxMzI4LUxpZmVjeWNsZS5qcGciLCAiZWRpdHMiOiB7InJlc2l6ZSI6IHsid2lkdGgiOiA5NTAsICJmaXQiOiAiY292ZXIifX19" style="width: 786px;">After the manufactured devices are sold, typically to Service Providers (such as energy suppliers or railway operators), who place them in the field. These Service Providers handle the next phases of the IoT lifecycle:<ul><li>Setup &amp; Pairing – The generated public/private keypairs are used to obtain device unique certificates. With these certificates the devices are authenticated to the cloud and trusted PKI connections between device and cloud are established.</li><li>Operation – If the data is only encrypted while travelling from device to cloud, it can be decrypted in the cloud before it reaches the Service Provider. To guarantee data remains encrypted until it reaches its destination, another keypair is generated to apply end-to-end encryption from the IoT device to the cloud of the Service Provider. Additionally, all sensitive data on the device is encrypted during operation.</li><li>Update – Cryptographic keys are also used to provide encryption and authentication for over-the-air updates, so that attackers can no longer install malicious updates or intercept updates in transmissions.</li><li>End-of-Life – All sensitive key material of the RESCURE prototype can be zeroized, which means the entire identity of a device, including all its rights and authorizations, can be erased.</li><li>Refurbish – After zeroization of a device, it can be re-enrolled with a completely new identity. The old identity and corresponding keys, however, are gone forever. This way, IoT device hardware can safely be re-used for new purposes.</li></ul>For more information on RESCURE and its solution for securing IoT devices, have a look at our video above. Also, two papers about the project were recently accepted at scientific conferences and will be published at the <a href="https://www.ares-conference.eu/workshops/wisi-2020/" target="_blank">ARES 2020 workshop WISI</a> and at <a href="https://wisec2020.ins.jku.at/accepted-papers/" target="_blank">WISEC 2020</a>.

A consortium effort to address the need for a security solution that spans the lifecycle of devices in the Internet of Things.

RESCURE: A Security Solution for the IoT Lifecycle

There’s few people who will disagree that there is massive potential in machine learning (ML) and artificial intelligence (AI) technologies for the next evolution of <a href="https://response.nordicsemi.com/the-industrial-iot?utm_campaign=2021%20wevolver" rel="noopener noreferrer" target="_blank">Industrial IoT</a> solutions. However, there’s also significant barriers for developers wanting to get started, or move beyond the basics.Companies looking to harness these opportunities will have to invest significant resources in testing and experimentation. Algorithms must be tested against real data to truly measure business impact, something that may be difficult for small companies or even large companies with no experience of AI/ML to contemplate.<h2>Machine learning out-of-the-box</h2>Rather than hiring teams of data scientists, other solutions are coming to market. One of those, <a href="https://sensiml.com/" target="_blank">SensiML</a>, is capable of auto-generating machine learning inferencing code for complex sensors like motion and vibration pattern recognition. We’re interested in this because SensiML can run locally on the nRF52 and also supports out-of-the-box connectivity to the Nordic Thingy:52 prototyping platform.The business was born out of a project at Intel. When the company cancelled the project, the team felt there was plenty of value and potential and so the core group acquired the project to continue it independently in 2017.<h2>Extracting data value, without data scientists</h2>It aims to solve a problem for developers of niche consumer and industrial projects. That being, how can you use an algorithm to extract value from the raw data from a sensor-based embedded device, without hiring teams of data scientists?“For high volume consumer product companies such as FitBit, it’s justifiable. Not so for a small startup or even a larger company targeting a more specific application or variable use case, such as a motor with vibration sensors to detect unusual operation across many different environments,” explains SensiML’s Chris Rogers.<h2>Machine learning algorithms</h2>The SensiML platform enables someone with development skills and application experience to generate machine learning algorithms. There are two ways of working with SensiML, as a general user or a more technical user.At the advanced level you can programmatically interact with the SensiML AI code generation service using the full power of Python and command-line coding. The general user mode has a GUI interface that allows users to enter constraints (e.g. available RAM), parameters (for accuracy) and then hit a button to generate candidate models.“A lot of applications can use classic learning methods such as decision trees and distance based classifiers. Simple machine learning approaches often work well when combined with automated feature extraction as SensiML supports and they need much less computing power than a full deep learning neural network. In the end, the best algorithm is the one that delivers the desired performance most efficiently,” explains Rogers.<h2>Works with Nordic</h2>SensiML can run locally on the nRF52 and also supports out-of-the-box connectivity to the <a href="https://blog.nordicsemi.com/getconnected/get-started-with-the-nordic-thingy?utm_campaign=2021%20wevolver" rel="noopener noreferrer" target="_blank">Thingy:52</a>. The SensiML team is also looking at future support for the nRF91 series.“The challenge with complex sensing is that there is a lot of richness in the raw data. It’s impractical to stream all of that, so ideally you pre-process the data and only send what you’re interested in. With the M4 and BLE radio integrated, the Nordic nRF52 has enough power to do that,” says Rogers.<h2>SensiML out in the wild</h2>The number of potential application areas for SensiML are many and varied. Rogers says there’s been a lot of interest from industries that require predictive maintenance: “There’s a lot of customer interest in being able to understand the data that comes from measuring vibrations, for example.”Other interesting application areas include building intelligence into smart devices. This could be voice recognition technology or sound recognition, for example, a microwave being able to recognize popcorn being cooked and when to shut off, i.e. when the popping noises from popcorn have subsided.The passive infrared sensors in smart lighting solutions can also benefit, explains Rogers:“There’s effectively a two pixel camera in a common passive infrared sensor. Typically the two pixels compare signals differentially and if there is a change, that’s recorded as movement. But with SensiML you can tap into those elements as waveforms and begin to recognise patterns such as direction of movement or any exaggerated movement such as the waving of arms or running. But it could also be used for monitoring occupancy of meeting rooms or bathrooms. It solves the privacy issue of not wanting cameras.”<h2>The next generation of wearables</h2>Rogers believes there’s also plenty of potential in taking fitness-based wearables to the next level:“Such devices could move beyond measuring the amount of exercise into the quality of exercise. For example, by measuring the gait during a run. We have collected accelerometer and gyro data from people on treadmills. A fitness coach then rated technique and recorded faults such as heel strike and excess leg rotation by watching the video and labelled everything. From this, we created a classifier than enables the prospect of a virtual running coach providing real-time analysis and feedback locally on the device.”<h2>The long tail of intelligent IoT applications</h2>These are just a few of literally countless applications enabled by local AI/ML running on low power connected devices such as those powered by the nRF52. What has stood in the way of realizing such diverse usages has been the complexity, skillset, and time required to implement the underlying sensor processing algorithms driving such applications.Auto code-generating AI tools like SensiML are now available to IoT developers to enable this otherwise complicated effort quickly and efficiently. Learn more about SensiML at <a href="https://sensiml.com/" target="_blank">sensiml.com</a><hr>This article was first published on Nordic's&nbsp;<a href="https://blog.nordicsemi.com/getconnected" rel="nofollow" target="_blank"></a><a href="https://blog.nordicsemi.com/getconnected?utm_campaign=2021%20wevolver" rel="noopener noreferrer" target="_blank">Get Connected Blog</a>.

Making sensors more intelligent has required substantial investment in machine learning algorithms. The SensiML platform aims to bring that cost crashing down.

SensiML: Adding Real-Time Intelligence to Sensors

Tools that enhance human capabilities and flexible manufacturing systems are necessary to deliver on the core promises of the 4th industrial revolution.

Industry 4.0 Deep Dive. Part 5: Flexible manufacturing

For many years, silicon Physical Unclonable Functions (PUFs) have been seen as a promising and innovative security technology that was making steady progress. Today, Static Random-Access Memory (SRAM-)based PUFs offer a mature and viable security component that is achieving widespread adoption in commercial products. They are found in devices ranging from tiny sensors and microcontrollers to high-performance Field-Programmable Gate Arrays (FPGAs) and secure elements where they protect financial transactions, user privacy, and military secrets.<h2>Physical Unclonable Function (PUF)</h2><ul><li>A PUF provides a way to authenticate electronic devices or chips and generate cryptographic keys from their unclonable silicon fingerprint.</li><li>It uses uncontrollable deep-submicron variations from manufacturing to derive a stable, unclonable, and unique device identity.</li></ul><h2>Benefits of SRAM PUF</h2><ul><li>No need to program keys or obtain entropy externally.</li><li>No secrets stored in device when power is off.</li><li>Keys and identifiers based on unclonable silicon structure.</li><li>Uses existing, standard hardware (SRAM) as PUF.</li></ul><h2>Reliability and Certification</h2><ul><li>Advanced error correction.</li><li>Based on years of extensive reliability testing across technology nodes and operating environments.</li><li>Devices using SRAM PUFs have been government vetted and certified EMVCo-Visa and CC EAL6+.</li></ul><h2>Application Markets</h2><ul><li>Smartcard, FPGA</li><li>Sensors, microcontrollers</li><li>Mobile, automotive, IoT</li></ul><h1>The SRAM-based PUF</h1>Due to deep submicron manufacturing process variations, every transistor in an Integrated Circuit (IC) has slightly different physical properties. These lead to small but measurable differences in terms of electronic properties, such as transistor threshold voltages and gain factor. Since these process variations are not fully controllable during manufacturing, these physical device properties cannot be copied or cloned.Threshold voltages are susceptible to environmental conditions such as temperature, and voltage so their values cannot be used directly as unique secret keys or identifiers.&nbsp;The behavior of an SRAM cell, on the other hand, depends on the difference of the threshold voltages of its transistors. Even the smallest differences will be amplified and push the SRAM cell into one of two stable states. Its PUF behavior is therefore much more stable than the underlying threshold voltages, making it the most straightforward and most stable way to use the threshold voltages to build an identifier.<h2><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1591264573768-SRAM_behavior2.png">Figure 1: Extracting a strong secret key from SRAM behavior.&nbsp;</h2><h2>SRAM PUF Behavior</h2>An SRAM memory consists of a number of SRAM cells. Each SRAM cell consists of two cross-coupled inverters that each are built up by a p- and n-MOS transistor. When power is applied to an SRAM cell, its logical state is determined by the relation between the threshold voltages of the p-MOS transistors in the invertors. The transistor that starts conducting first determines the outcome, a logical &#39;0&#39; or &#39;1&#39;.It turns out that every SRAM cell has its own preferred state every time the SRAM is powered resulting from the random differences in the threshold voltages. This preference is independent from the preference of the neighboring cells and independent of the location of the cell on the chip or on the wafer.Hence an SRAM region yields a unique and random pattern of 0&#39;s and 1&#39;s. This pattern can be called an SRAM fingerprint since it is unique per SRAM and hence per chip. It can be used as a PUF.Keys that are derived from the SRAM PUF are not stored &#39;on the chip&#39; but they are extracted &#39;from the chip&#39;, only when they are needed. In that way they are only present in the chip during a very short time window. When the SRAM is not powered there is no key present on the chip making the solution very secure.<h2>PUF Reliability</h2>The deep submicron process variations that determine PUF behavior are frozen during manufacturing and do not change afterwards. Hence the preference of the SRAM cells is persistent and stable over time.However, there is still a degree of noise. A small number of the cells, close to equilibrium are unstable and display a seemingly random startup preference. So, each time the SRAM starts up, a slightly different pattern emerges. This noise component is dependent on temperature, voltage ramp, and operating conditions.The noise of SRAM-based PUF responses has been exhaustively characterized and tested under a wide variety of circumstances and foundry processes:<ul><li>Temperatures ranging from -55&deg;C to +150&deg;C (-67&deg;F to 300&deg;F)</li><li>Voltage variation +/-20%</li><li>Humidity up to 80%</li><li>EMC tests at 3V/m (EN55020 0.15&ndash;150 MHz and IEC 61000-4-3 80-1000MHz)</li></ul>In particular, the SRAM PUF has been qualified for automotive, industrial and military use in collaboration with customers and partners.Millions of measurements have been performed. Under all these circumstances the average noise level of the SRAM-based PUF response was found to be lower than about 15%. Despite this amount of noise, it is possible to reconstruct a high-entropy device unique and reliable key every time the SRAM is powered. This can be done by applying error correction techniques, such as &#39;helper data algorithms&#39;[1] or &#39;fuzzy extractors&#39;[2]. These algorithms perform two main functions that will be explained below: error correction and privacy amplification.<h2>Error Correction</h2>Error correction techniques for cryptographic key reconstruction require an enrollment phase and a reconstruction phase. In the enrollment phase (a one-time process) the PUF response is mapped onto a codeword of an error correcting code. Information about the mapping is stored in the Activation Code (AC) or helper data. The AC is constructed such that it does not leak any information about the key. It should be stored in memory that is accessible by the PUF algorithms, but it can be stored off-chip as it is not sensitive. Any change to the AC, malicious or not, will prevent key reconstruction. The AC is only valid for the chip on which it was created.Each time the device runs an authentication protocol and needs the secret PUF key, a new noisy PUF measurement is carried out and the PUF key (without noise) is extracted from the AC and this new PUF response. This is called the reconstruction phase. Both enrollment and reconstruction phases are illustrated in Figure 2.The error correction algorithms have been designed to reconstruct the key with an average error rate of less than 10-12 and an error rate of less than 10-9 under extreme circumstances of 25% average noise.<h2><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1591264692705-Enroll_reconstruct.png">Figure 2: Enrollment and reconstruction phase for the generation of PUF keys. Note that R is the initial PUF response during enrollment while R&#39; is a PUF response in the field with a noise component.</h2><h2>Privacy Amplification and Security</h2>Secret keys provide security based on the fact that they are completely random and hence unpredictable. Physical measurements, such as PUF responses, have a high degree of randomness, but are usually not completely uniformly random. Privacy amplification is used to generate uniformly random keys.By combining error correction and privacy amplification, a 1kByte SRAM PUF response can be turned into a 256-bit uniformly random key and only approximately 0.5 kByte is needed for a 128-bit key with full randomness. A typical SRAM PUF contains so much entropy that only a few dozen bytes are needed to provide a collision-free globally unique identifier that can be used as a unique (but noisy) electronic chip ID (ECID) or as a serial number.Dedicated security labs and security teams at customers have analyzed the security of Intrinsic ID&#39;s SRAM PUF against various invasive and non-invasive physical attacks without revealing any weaknesses. Attacks with Scanning Electron Microscopes, Lasers, FIBs and probes have not been successful. Side-channel attacks have not led to any leakage of sensitive information.<h2>Aging</h2>Accelerated aging tests have been performed on SRAM PUFs to investigate the noise level as a function of time. By using a patented anti-aging technique[3] a 25-year lifetime can be guaranteed for SRAM PUF technology.<h1>SRAM PUF Implementation &ndash; QuiddiKey &amp; BroadKey</h1>Intrinsic ID has bundled the above error correction, randomness extraction, security countermeasures and anti-aging techniques into its products. They extract cryptographic keys from the SRAM PUF in a very secure manner and are available as Hardware IP (netlist) called QuiddiKey&reg;, firmware (ANSI C Code) called BroadKey&trade;, or a combination of these by using BroadKey&#39;s hardware interface. These hybrid solutions combining hardware and software offer great efficiency when already integrated hardware accelerators, such as for (a-)symmetric cryptography, are used in combination with SRAM PUF technology.The hardware IP is small and fast &ndash; around 25K gates / 50K cycles &ndash; and connects to common interconnects, such as AMBA&reg; AHB, APB as well as proprietary interfaces. A Built-In Self-Test (BIST), diagnostics, and health checks are included in the logic. A driver is provided to ease integration with software. Since it is pure digital, single clock logic it synthesizes readily to any technology.Software reference implementations start from 4KB of code and are available for major platforms, such as ARM&reg;, ARC&reg;, Intel&reg;, MIPS and RISC-V. Software implementations can be used to add PUF technology to existing products by a firmware upgrade. Intrinsic ID also provides versions of BroadKey that are pre-integrated with ARM&reg; TrustZone&reg;.Both QuiddiKey hardware and BroadKey software solutions can be optimized for low footprint, low latency or low memory use depending on the application. Re-using or integrating with existing crypto cores and random number generators can further enhance performance and reduce footprint. Intrinsic ID solutions come with comprehensive product specifications and integration guidelines including reference code that illustrates usage of the API offered to the application programmer.<h2>Requirements</h2>These Intrinsic ID products use uninitialized SRAM. This can be a separate SRAM block or a part of a bigger existing SRAM. Standard SRAM suffices. To store the Activation Code (AC), access to a storage medium is needed. This can be embedded Non-Volatile-Memory (NVM), a separate memory on the board, e.g. flash, or cloud storage. For the firmware version BroadKey, a microcontroller is needed for which a C-compiler exists. The PUF algorithms can be stored in any NVM e.g., flash, ROM.Note that SRAM is embedded in almost any microprocessor and SoC, in every technology node and is part of the standard manufacturing process. There is no need for time-consuming qualification and chip testing, as extensive testing by Intrinsic ID and its partners has shown that technology reliably scales down to the smallest technology nodes currently available.<div id="ftn1"><h2>Use in the Field</h2>SRAM-based PUFs are commercially available from semiconductor companies. They are already deployed in a wide range of microcontrollers, FPGAs and smart card controllers. In other markets, software implementations[4] have enabled rapid deployment of this technology even as a retrofit solution. Intrinsic ID has partnered with leading semiconductor companies and developed solutions for protecting embedded systems, sensors, and controllers. More information about our SRAM PUF deployments can be found on our website: <a href="https://www.intrinsic-id.com" target="_blank">https://www.intrinsic-id.com</a>.<h1>Conclusion</h1>SRAM-based Physical Unclonable Functions have been successfully implemented in commercial products. SRAM PUFs combine high security and reliability with low cost, low footprint designs, and easy implementation. They have been deployed in many devices, from tiny microcontrollers and sensors to high performance FPGAs and secure elements.Numerous implementations have consistently demonstrated the reliability and security of the technology. SRAM PUFs are a mature and robust technology, designed for security and based on solid theoretical foundations. SRAM PUFs have gained their credibility in high security markets and are now gaining traction in markets ranging from low-cost IoT applications to high-end security solutions for government, defense, and the payment industry.<div id="ftn1"><h2>References</h2></div></div>[1] J.-P. Linnartz and P. Tuyls, &quot;New shielding functions to enhance privacy and prevent misuse of biometric templates,&quot; in International Conference on Audio and Video-based Biometric Person Authentication (AVBPA&#39;03), ser. LNCS, J. Kittler and M. S. Nixon, Eds., vol. 2688. Heidelberg: Springer-Verlag, 2003, pp. 393&ndash;402.<div id="ftn2">[2] X. Boyen, &quot;Reusable cryptographic fuzzy extractors,&quot; in ACM Conference on Computer and Communications Security (CCS&#39;04). New York, NY, USA: ACM, 2004, pp. 82&ndash;91. AND Y. Dodis, L. Reyzin, and A. Smith, &quot;Fuzzy extractors: How to generate strong keys from biometrics and other noisy data,&quot; in EUROCRYPT&#39;04, ser. LNCS, C. Cachin and J. Camenisch, Eds., vol. 3027. Heidelberg: Springer-Verlag, 2004, pp. 523&ndash;540.[3] R. Maes and V. van der Leest, &quot;Countering the effects of silicon aging on SRAM PUFs&quot;, Proc. IEEE Int. Symp. Hardw.-Oriented Secur. Trust (HOST), pp. 148-153 available at <a href="https://www.intrinsic-id.com/wp-content/uploads/2017/05/PUF_aging.pdf" rel="noopener noreferrer" target="_blank">https://www.intrinsic-id.com/wp-content/uploads/2017/05/PUF_aging.pdf</a>&nbsp;[4] H.Hodson &ndash; New Scientist, &quot;Silicon fingerprint on chips could make any gadget unhackable&quot; June 6, 2016 <a href="https://www.newscientist.com/article/mg23030771-400-physical-quirks-in-silicon-chips-are-key-to-unhackable-devices/" rel="noopener noreferrer" target="_blank">https://www.newscientist.com/article/mg23030771-400-physical-quirks-in-silicon-chips-are-key-to-unhackable-devices/</a> or &quot;Chip design quirks make our lives more secure&quot; in the June 11 printed issue page 24.</div>

Low-cost and strong secret key storage technology is critical today to enable affordable and effective security systems.

SRAM PUF: The Secure Silicon Fingerprint

The fourth industrial revolution and the technologies that create it explained: The whole is bigger than its parts.

Industry 4.0 Deep Dive. Part 1: Introducing smart manufacturing

Up to 30% of a compressor&#39;s normal output can be lost through leakage, and poorly maintained equipment will generally be leaking compressed air at around 20% of its total capacity. Leaks may also cause your whole system to lose pressure, leading to less efficient functioning of pneumatic devices, and loss of production output.Not just the compressor, but almost all your system equipment will have its lifespan reduced, as leakages force more frequent cycling and consequent wear. To estimate how much your pneumatic system may be leaking will require running the compressor for a specified time when it is not driving any devices, and measuring the average time it takes to load and unload the system. This loading and unloading occurs as leaks cause pressure drops in the system, and the compressor then cycles on and off to compensate.Not all leakages can easily be prevented, and system design should incorporate some standard allowance for a pressure drop rate per unit time. These parameters could be anticipated in the circuit design, so that appropriate pressure rates and reservoir sizes are determined.<h2>How To Find Leaks In Pneumatic Systems</h2>Sometimes, unnecessary additional compressor capacity is installed to compensate for pressure lost due to leakage. A better and more cost-effective alternative is to introduce a proactive programme of detection and repair, which can bring down your average leak rate to under 10%. The most common areas where leaks can occur include joints, couplings, seals and fittings, as well as hoses and areas which have missing welds or worn packing. Missing O-rings are a common problem, as are improperly installed FRL units and worn point-of-use tools.Air is invisible, of course, so once you&#39;ve calculated the amount of air you&#39;re losing through leakage, you&#39;ll have to find some way of detecting the specific locus of the leak(s). A modern solution to this is ultrasonic detection, by means of an acoustic device to identify the high frequency sounds made by leaking air.Since this type of equipment is commonly portable, it makes the detection of leaks very fast, accurate and versatile. Ultrasound is directional in its transmission, emitting a louder signal the closer it is to the source, so leak sites can be very easily pinpointed by scanning the general area where the equipment is located. Tests can be carried out while equipment is in use, and the detectors are capable of finding many different types of leak. This kind of device is very simple to operate, and it takes only a few minutes to learn how to use it.<h2>What Causes Leaks In Pneumatic Systems?</h2>Leaks are most often found at junction points between the system and end-use pneumatic applications, especially around connectors and joints. Fixing a leak can mean a straightforward tightening-up of a connection, or more complex replacements of faulty components like hoses, fittings, couplings, joints, pipe sections, traps or drains.<h2>Faulty Connectors</h2>Connectors are a very large family of components used to join hoses, tubes or peripheral devices. They need to be chosen carefully and precisely to ensure that a leak-proof joint is achieved, and sealed where necessary. Push-in connectors are very popular due to the simplicity of their installation, but they may be prone to suffer more leakage if a cycle involves frequent connection and disconnection.<h2>Pressure</h2>Leakage may also ensue if the pressure load on valves or connectors exceeds their specified rating. The flow rate of air depends on the pressure differential pertaining across a leakage point, so a lower overall system pressure will help reduce the rate of leakage. If you can stabilise the header pressure of the system at the minimum range that is practical, the leakage rate should also be minimised.<h2>Too Many Connection Nodes</h2>However carefully you seal your connections, if they are pushed to function outside their pressure parameters, some leakage is bound to occur. Furthermore, the more connections there are in the system, the greater the likelihood of multiple leakages leading to a cumulative drop in overall pressure. Increasing the number of connections is not therefore a cost-effective way to expand your pneumatic system, as it also makes the system more complicated, and more difficult to detect leaks.<h2>Wrong Connection Type</h2>Thread types come in many varying styles and sizes, and tend to differ according to their country of origin and use. Connections typically use male and female mating parts, so it&#39;s important to make sure that the thread types match. There are also recommended torque values for most threaded joints, and connections should not be over or under-tightened in relation to these values. Threaded joints should be secured with teflon tape.Seal DegradationExposure of some materials to chemicals may make them degrade. Items such as rubber washers may develop cracks that allow leakage, or thread sealant might be of the wrong type, or not properly applied. All fittings, tubing, hoses, disconnects, etc, should be of high quality, and properly installed with the right thread sealant.<h2>Cylinder Bore Wear</h2>It&#39;s possible for a pneumatic cylinder to suffer excessive wear on the bore if it doesn&#39;t have the proper lubricant. Misalignment or improper mounting can also cause lateral forces to be applied on the bore, which could cause air to leak along the piston&#39;s periphery.<h2>Predictive Leak Prevention</h2>Establishing a good leakage identification and repair programme is critical to maintaining a stable, reliable, efficient and cost effective pneumatic system. Such a programme should involve employee teams in the identification and tagging of leaks, tracking performance, carrying out repairs and verification. Whenever leaks are identified and repairs instituted, the whole system should be reassessed. Scheduled predictive maintenance can help you detect and deal with leaks, so that you save on reactive maintenance and unscheduled downtime. An aggressive programme of leakage prevention and reduction should therefore form part of the overall performance improvement of pneumatic systems.<hr>This post was originally posted on&nbsp;<a data-fr-linked="true" href="https://www.rowse.co.uk/blog/post/how-to-find-leaks-in-pneumatic-systems" rel="noopener noreferrer" target="_blank">www.rowse.co.uk</a>.

If a compressed air system develops a leak, it can waste a lot of energy, especially in a complex automated industrial setting.

How To Find Leaks In Pneumatic Systems

Automated systems governing smart manufacturing, and enabled by IoT devices, give us the IIoT (Industrial Internet of Things), which allows for the expansion and streamlining of all aspects of business. In addition, the deployment of AI and robots is particularly useful in industry as they increase product quality, production efficiency and overall speed.The introduction of collaborative robots (or cobots) is becoming increasingly common, in industry as well as in laboratory work and commerce. Cobots can take instructions from humans in order to work more productively with them, including instructions that were not originally anticipated in the initial programming. Robots don&#39;t get bored, hungry or tired. The <a href="https://www.rowse.co.uk/blog/post/the-rise-of-artificial-intelligence" target="_blank" title="The Rise of Artificial Intelligence">capabilities of Artificial Intelligence</a> are far beyond human capacity when it comes to such things as miniaturisation and precision measurements, and they deliver vastly superior quality assurance.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1584368176061-1584368176061.png" style="width: 1080px;" class="fr-fic fr-dib">So what are the benefits we can expect to see as industry moves forward into the 21st century world of cobots and AI?<h2>1. Direct Automation</h2>IIoT connects all IoT enabled devices to the factory floor, integrating manufacturing processes with big data and making them programmable via a logic controller. Increased use of precision sensory equipment means that information can be generated, recorded and analysed for all aspects of the production process, covering anything from temperature to item picking and packaging. Programmable logic controllers with AI capacity for deep learning can then respond automatically to the seamlessly generated information, and make alterations to the minutest function without recourse to human intervention. The big data analytics processed by AI can substantially improve performance across the entire production process, and can be operated remotely.<h2>2. 24/7 Production</h2>Human beings are biological organisms and need regular maintenance, e.g. food and sleep. For any production facility to continue working round the clock it is necessary to introduce shifts, using three human workers for every 24 hour period. Robots don&#39;t get tired or hungry, and are capable of working on the production line 24/7. This allows the expansion of production capabilities, which is increasingly necessary to meet the demands of worldwide customers. Furthermore, robots are more efficient in many areas such as the assembly line, and picking and packing departments. They can greatly reduce turn-round times in many areas of the business operation.<h2>3. Safety</h2>Human beings are fallible and prone to making mistakes, especially if they are tired or distracted. Errors and accidents do occur on the factory floor and in any construction or processing environment; a tendency which can be all but eradicated by AI and robotic assistance. Remote access control means a reduction in human resources, especially when the work is dangerous or requires superhuman effort. Even regular working environments will cut down on the incidence of industrial accidents and lead to an overall improvement in safety. In addition, more advanced sensory equipment integrated with IIoT devices make the installation of safety guards and barriers a simpler and more effective measure to protect human lives.<h2>4. Lower Operational Costs</h2>Many companies are viewing the introduction of AI into the manufacturing industry with trepidation, as it requires a huge capital investment. On the other hand, the ROI is significant and increases as time goes on. Once intelligent machines begin to take over the daily activities of a factory floor, businesses will benefit through considerably reduced operating costs, with predictive maintenance helping additionally to reduce machine downtime.These days, consumers are increasing their demand for unique, personalised or customised products, while continuing to expect the best value. With such advances as 3D printing and IIoT connected devices, it is becoming simpler and cheaper to meet these needs, and using virtual or augmented reality design techniques means that the whole production process will be more cost-effective. Integrating machine learning and CAD means that systems can be designed and tested in a virtual model before they are put into production, thus reducing the cost of trial-and-error machine testing.<h2>5. Greater Efficiency</h2>IIoT enables the collection of vast amounts of data and advanced analytics which can be used to gain insights into consumer behaviour. Trends can be predicted, patterns recognised and market developments forecast across time, socioeconomic sectors and geographic markets, as well as taking into account political developments, macroeconomic cycles and even weather patterns. AI has the capacity with machine learning to anticipate information, to refine processes, and to track incongruities, all the way down the supply chain from source to finished product. This is particularly aided by such technologies as&nbsp;<a href="https://www.rowse.co.uk/blog/post/rfid-in-manufacturing" target="_blank" title="RFID In Manufacturing">RFID tracking</a>, which enable materials to be tracked without having to go through a physical process such as a bar code reader.<h2>6. Quality Control</h2>AI is also extremely useful for carrying out predictive maintenance on machinery and equipment. Using sensors to track performance and operating conditions, machines can learn to predict malfunctions and failures, and take action to remedy them before they occur. This can result in faster feedback, helping companies to eradicate unplanned downtimes.Sensors can also detect the mot microscopic defects, scanning them at resolutions far beyond the capacity of human vision, thus improving productivity and increasing the percentage of items that will pass quality control. AI helps to speed up many routine processes and improves accuracy to an enormous extent. This obviates the requirement for quality control and in-process inspection by human beings, which is time-consuming and often fallible.<h2>7. Quick Decision Making</h2>When IIoT is coupled with cloud computing and virtual or augmented reality, companies can share simulations, confer on production activities and exchange critical or important information in real-time, irrespective of geographical location. The data gathered from sensors and beacons helps to determine consumer activity, allowing companies to anticipate future needs and make rapid decisions on production, as well as speeding up the exchange between manufacturers and suppliers.<hr>This post was originally posted on <a data-fr-linked="true" href="https://www.rowse.co.uk/blog/post/7-manufacturing-ai-benefits" rel="noopener noreferrer" target="_blank">www.rowse.co.uk</a>.

As the world becomes increasingly automated, the IoT (Internet of Things) is already transforming our domestic and business lives. Nowhere is this more apparent than in the use of AI and robotics in the manufacturing industry, with all the benefits offered by Industry 4.0.

7 Benefits of AI In Manufacturing

Most robots used in factories today are powerful but blind. They are programmed to do specific tasks repeatedly with high precision and speed, but until now have not been able to adapt to changes in the environment or handling variables. Robot arms often lack the ability to grasp an object if it is moved a few inches, not to mention picking or sorting millions of different products in a warehouse.As a result system integrators design shaker tables, feeders, conveyor belts, etc., so robots can be offered the same parts in the right orientation. Larger industrial robots need cages so they don’t accidentally hurt nearby people. Unlike human hands, which can perform tens of thousands of different tasks, robot arms can only be programmed for single tasks, and they need the proper end effectors.Additionally, most robots cannot sense, unless you equip them with force or vision sensors. They don’t know if tasks are performed properly, because they cannot see or feel.Traditional computer vision requires pre-registration of all objects, i.e. scanning and building computer-aided design (CAD) models for all items beforehand with expensive 3D cameras. The process is time-consuming and inflexible. Even if there was a box with exactly the same shape, but a different size, a robot couldn’t automatically recognize the differences and figure out how to grasp the new item.In addition, translucent packaging, reflective surfaces, and deformable objects (such as food and clothing) make robotic vision an even more difficult problem to solve. The additional non-robot hardware and integration end up costing four times more than a robot arm, which costs anywhere between $10,000 to $100,000.Adding intelligence to robotsTo solve these issues, deep reinforcement learning (DRL) is unleashing the full potential of robotics, giving them the ability to recognize and react to their surroundings and handle product variations. Robots can now learn from experiences with off-the-shelf cameras and deep neural networks.With enough data and practice, robots can now teach themselves new abilities, as was seen with robots’ ability to identify images and win video games. With every grasp and trial, a robot gets smarter and more adept at mastering the task. Furthermore, cloud-connected robots allow learnings to be shared with each other. This is a tremendous shift, making robotics solutions more dexterous, flexible, and scalable.Deep Learning (DL) is primarily used for image classification. Deep neural networks learn representations, rather than task-specific algorithms.Reinforcement learning (RL) is inspired by behavioral psychology. It learns control policies by trial-and-error through reward or punishment.Deep Reinforcement Learning (DRL) combines these two approaches. It is currently suited for:· Simple or simulated tasks· Fault-tolerant tasks· Just enough variation to make explicit modeling difficult· Easy-to-define goals (reward functions)· Semi-structured environmentsWarehouse as a testing ground for AIAutomation in warehouses is widely considered the “low-hanging fruit” for these technologies because tasks tend to be similar across warehouses, and the environment is more structured. For example, order picking represents more than 40% of operational costs across most warehouses. While tasks are repetitive, they involve some variation with products and packaging design, making it impossible to hard-program robots in advance.Furthermore, labor costs account for up to 70% of a warehouse’s total budget. Pressured by e-commerce companies such as Amazon, retailers are looking to automation to address labor shortages and reduce costs. Customer demand, price competition, and fast delivery requirements are driving a strong business case for warehouses to be the first adopter of AI-enabled robots.Chart: Why do we see AI-enabled robots in the warehouse first?<img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNTY4OTAzMzQ5MzI3LTFfTDhkY3A1c1lfcEU0MTdqdW9VaEMwZy5wbmciLCJlZGl0cyI6eyJyZXNpemUiOnsid2lkdGgiOjk1MCwiZml0IjoiY292ZXIifX19" style="width: 766px;" class="fr-fic fr-dib"> Most tasks in a warehouse are fault tolerant, unlike tasks on a production floor or in the automotive industry. If a robot accidentally drops an object in a warehouse, it can try again as long as it achieves the required speed and success rates without damaging the goods.DRL is particularly suited for fault-tolerant simple tasks — actions a human can perform within a few seconds — with enough variation in semi-structured environments. As technology improves in environments like warehouses, it can then be applied in other industries, such as manufacturing, where tasks are more complicated and more uncertainties exist. Thus, an automated warehouse acts as an incubator, predicting a knowledge spill-over to many other market segments.Piece-picking is the holy grailWarehouse automation has been around for decades, in the form of automated storage, retrieval systems, and mobile robots like those from Kiva (now Amazon Robotics), that move items around.But most of the pick-and-pack operations within a warehouse are still performed by humans, with labor costs accounting for 50 to 70% of the overall warehouse budget. Worker shortages have driven up salaries by a rate of 6 to 8% every year, while robot costs have fallen considerably since the 1990s.Piece-picking has long been the holy grail of robotics. Year after year, incumbents such as Amazon and KUKA host robotics challenges for startups and academic teams to create machines that can identify, pick up and stow goods. The emergence of deep learning allows robots to recognize, pick, and place hundreds or up to thousands of items.But the technology is not yet perfect. It’s still challenging for a machine to recognize tens of millions of objects, and manipulate deformable objects or items with transparent packages. However, our interviews with Locus Robotics and Osaro indicate that many industry experts expect the technology will mature. In addition, we will have a robot vision camera that can recognize almost everything in a warehouse within five years. If that happens, it will affect not just warehouses, but retail, delivery, and many other applications that we haven’t yet imagined.Pushing towards standard modelsTechnology improvements in warehouse automation and beyond will also push the mature, but fragmented, robotics industry to become more standardized and modular. Given the nature of narrow AI, most Silicon Valley startups in the AI-enabled robotics industry choose to focus on specific vertical applications and collect proprietary data as a defensive advantage.Piece-picking by robots is the best example. As the technology improves, we will see warehouse piece-picking startups enter manufacturing, food, and agriculture markets. Moving away from manually programming a robot and having it learn and act autonomously on specific simple tasks — and sharing the learning — will become more important in the future.However, it’s not easy to find scalable and technically feasible business cases like piece picking that customers are willing to pay for. For example, assembly tasks vary from factory to factory, making it difficult to get sufficient training data or scale a solution across different customers.Given current technology constraints, there’s still a trade-off between precision and flexibility. Deep learning allows robots to handle variation, but these AI-enabled robots still cannot achieve the same precision and accuracy as traditional robots, at least not today.Some startups are providing modules or technology stacks for other players in the ecosystem to mitigate the complexity and scalability issues mentioned above. New areas where startups are likely to enter and disrupt include teleoperation modules, controllers that allow for safer human-robot collaboration, human-centered AI protocols, and sensors that can quantify human behavior.Eventually, we may see unified operating systems for AI-enabled robots similar to Windows or Android in the consumer electronics space, or standard user interfaces that will fundamentally transform the fragmented robotics industry and further accelerate its disruption.Standard modules and interfaces will bring down the cost to enter the market and shorten the time to market. As DRL and robotics technology improves, we will see a boom across industries including agriculture, food processing, home, and even surgical robotics.How technology will transform the industry landscapeLed by industry leaders like Amazon and Ocado, a British online grocery store, retailers see efficient and cost-effective supply chains as a key to success and are incentivized to automate their warehouses for lower cost, lower management complexity, and higher scalability. While Amazon keeps its Kiva mobile robots and automated warehouse solutions for in-house use only, Ocado launched its fulfillment platform to help other companies store, pick, and transport groceries. But most retailers still rely on material-handling companies and system integrators to build and maintain an automated warehouse.Incumbent material-handling companies tend to exploit the current technology, focusing on high-revenue-potential retrofit projects. This gives startups an opportunity to enter from the lower-end markets, where incumbents are less interested, and thus would not respond immediately. Dozens of startups offer scalable and affordable solutions that address customer needs for robots that can handle high-variability, low-precision tasks, such as piece picking in the warehouses.On the other hand, large robotics companies have long focused on developing high-speed and more-precise solutions to serve their main customers in the automotive and manufacturing industries. The pursuit of absolute precision is contrary to the process of deep learning, where results improve over time through trial and error.Some robotics companies saw an opportunity to enter new segments that require robots to handle high-variability, but lower-precision tasks, and began to build in-house AI teams and explore partnerships with startups developing deep neural network solutions for robots.There are currently more than 20 warehouse piece-picking robot startups worldwide, including Osaro and Righthand Robotics. Most of these startups (between seed round and Series C funding) are working on pilot projects with incumbent material-handling, retail, and robotics companies.Software and services will also capture more values, while premiums in hardware will decrease. With this power shift, as well as potential mergers and acquisitions, the landscape will change significantly over the next few years.The future of AI-enabled roboticsThoughts vary on the effect that AI and DRL will have on the robotics industry. One scenario is the robotics industry will consolidate because of the AI network effect, which says results improve as the amount of training data they’re given increases. This idea is that whoever owns the most proprietary data will win the competition.However, new researches in the areas of Transfer Learning, Meta Learning, and Zero-Shot Learning make it possible to learn from fewer and fewer examples, generalizing learning results across different fields. Cloud-based AI and robotics will also help accelerate learning and diffusion.As these technologies evolve, the landscape may face yet another significant change. This all starts from the warehouse!

IIoT

Latest Posts