The field of robotics has made big strides becoming more and more adaptable to changing environments. The robots’ autonomy can be broken down into the following functions: perceiving, planning, manipulating, navigating and collaborating. Combining AI and robotics facilitates a more optimum degree of autonomy through enhanced learning. The existing robots can autonomously carry out a set specialized tasks such as driving a car, carrying boxes, picking up objects and so on.
In the last years, computers have improved the quality of sensing and vision leading to an improvement in another application of artificial intelligence in robotics: perception. This is crucial for planning as well as establishing a sense of self-awareness, which promotes smoother interactions between the robot and the other elements in its environment. This field is also known as social robotics, consisting of cognitive robotics and human-robot interactions.
The underlying idea of the human-robot interaction points to improving the robotic perception and non-verbal communications in order to enable easy navigation of an environment alongside humans. Cognitive robotics, on the other hand, is concerned with autonomous capacity of knowledge acquisition based on experience and imitation of the human cognitive system. This is where AI advancements can truly shine and overcome numerous challenges to redefine robotic intelligence. So, what are some of these AI-powered technologies bolstering the robotics landscape?
Here are some of the technologies that can supplement AI in empowering the future of robotics.
With automation becoming the norm in every industry today, there is an increasing need for robots and other technologies such as AI-ML to provide a range of services to customers. People choose to believe the notion that robots can replace human labor and make them obsolete. However, it is not feasible. A well-managed AI-enabled robot can supplement human effort to provide better services, but it can never make human intervention one-hundred percent obsolete. Let us now see the emerging applications of AI in Robotics.
Although related, the correct term must be machine vision or robot vision. Giving the robots the ability to see includes more than computer algorithms; engineers need to consider the camera hardware that helps robots in processing physical data. There is a slight difference in kinematics between robot and machine vision. Robot vision involves the robot's ability to affect its physical environment and reference frame calibration. The big data on the web has propelled research in computer vision and helped improve ML-based prediction learning techniques. Automatic vehicles and drones are utilizing lidar, ultrasound, and radar technologies for providing 360-degree vision-based applications. Here’s a list of some machine vision for robotics applications:
The semiconductor industry has been deploying robots for manufacturing operations for mechanical as well as intelligent tasks, which requires machine vision. These tasks revolve around defect detection and their classification. Robots use machine vision to process the images obtained from the high resolution camera implemented in the robots.
Automation manifests in a number of ways in the retail landscape. Some of the more established applications are associated with distribution and shipment activities i.e. sorting, packaging, etc. A relatively new application is the usage of robots in customer service to provide assistance in finding, locating, advising on the items the customers are looking for. Machine vision here helps with object (item) and face (customer) recognition.
Store inventory management is another application in the retail industry. This function has to do with scanning shelves, counting available inventory, etc. In order to carry out these tasks, the robot needs to be able to recognize product codes and prices, which is another application of machine vision. Two good examples of these retail robots are Tally (the world’s first robotic autonomous shelf auditing and analytics solution) and OSHbot (a human-sized robot used across select Orchard's stores to help them find the products they are looking for).
Many of the tasks performed by agricultural robots require machine vision algorithms. From seeding to handling weeds, machine vision ensures the successful execution of all these tasks. For instance, in driverless tractors responsible for plowing utilize machine vision to navigate their path, classify objects and avoid obstacles. The same principle holds true for seeding to enable the robot to closely follow the rows to implant the seeds where it should. Another set of tasks aided by this AI application is the picking of the fruits and vegetables. The process of navigating and detecting the vegetables and fruits is perfected through machine vision algorithms.
The imitation learning approach uses Bayesian or probabilistic models for allowing the robots to act in the world to maximize their rewards. Industry applications include legged locomotion, multi-terrain mobile navigators, and humanoid robotics, deploying inverse optimal control methods. This principle is also known as robot learning by demonstration or robot programming (LfD) by demonstration (PbD) and is rooted in the fact that the end user can teach robots to perform additional tasks without any programming. In this approach, the robot is usually provided with a series of already known behaviours and actions that were either learnt prior or preprogrammed. The graphics below illustrate examples of this.
It is a learning approach that helps make the robots' smart.' They utilize the AI to generate their training examples for improving performance. They use close-range data to detect and reject objects, for example, dust and snow and identify obstacles in rough terrain. Self-supervised learning is also used in modeling vehicle dynamics and 3D scene analysis. Another relevant application is the tool manipulation, which is made possible by a task-oriented grasp facilitating the orientation and thus manipulation of the tool. For example, the task of grabbing a hammer and using it to sweep or hammer.
An assistive robot captures data, processes the information, and performs actions that can benefit the older generation and people with disabilities. This AI technology is also used in driver assistance tools. Furthermore, there are movement therapy robots which offer therapeutic benefit. Here are some of medical robotics and technologies advances:
Perhaps one of the most commonly used medical robots, the Da Vinci Surgical System helps surgeons perform surgery whilst making a few very small and precise incisions. This means less blood loss as well as infection complications and thus a faster recovery process for the patient.
This is where medical robots can truly shine. Utilizing AI and ML, diagnostic tasks can be performed much better by the robot. An example is using facial recognition softwares to screen patients for a number of diseases and disorders. The AI can scan medical documents and records to identify at-risk patients for diseases such as stroke, diabetes, etc.
The MURAB project (MRI and Ultrasound Robotic Assisted Biopsy) is a technique aimed for early cancer diagnosis through biopsy. MURAB scans and collects data through a minimally invasive procedure, providing the surgeon with a 3D image of where they need to be exactly getting the biopsy from.
To sum up, AI and ML can massively transform the robotics industry. One has to concede that the technology is still in its infancy with a significant scope to grow. However, the developing technology will ensure that AI continues to empower the field of robotics. AI-ML, along with other emerging technologies, can push AI robots' boundaries to an entirely different level in the future.
Robotics Online Marketing Team, How Artificial Intelligence is Used in Today's Robots, September 09, 2018, https://www.robotics.org/blog-article.cfm/How-Artificial-Intelligence-is-Used-in-Today-s-Robots/117
Polly, How AI Affects the Robotics Industry and What the Future Holds, March 10, 2020, https://roboticsandautomationnews.com/2020/03/10/how-ai-affects-the-robotic-industry-and-what-the-future-holds/31197/
P.-Y. Oudeyer, "Socially guided intrinsic motivation for robot learning of motor skills," Autonomous Robots, vol. 36, pp. 273-294, 2014, https://hal.inria.fr/hal-00936938v2/document
RELATED TECH SPECS