|Ground to hip: (legged mode)||0.14||m|
|Ground clearance: (legged mode)||0.10||m|
|Leg–wheel (i.e., rim) diameter:||0.21||m|
|Maximum radius of leg–wheel:||0.14||m|
|Driving:||60||W dc Motor (8x)|
|Steering mechanism:||19.2||W dc motor (×1)|
|Hall-effect||4||@2 DOF mech.|
|Temperature sensor||1||@computing uni|
|Current measurement||1||@motor power output|
|Battery voltage measurement||1||@power output|
|Potentiometer meter||1||@steering mech.|
|- Battery||48||V-6.3 Ah Li-ion battery|
The robot has four leg-wheel modules. Each leg-wheel module is composed of two half-circle legs and can be driven to rotate and move radially with respect to the hip joint, where the motion is similar to the definition of the 2-degrees-of-freedom polar coordinates.
When the legs form a circle and rotate, the robot moves like an ordinary four-wheeled vehicle, and when the legs are extended, the robot moves like a quadruped. The leg-wheels of the robot can be replaced by another type of leg-wheel. By using the passive torsion spring, the mechanism can react rapidly to external forces. In addition, by actively using only the rotational DOF, this method also reduces the loading of the DC motors.
The robot has four leg-wheel modules. Each leg-wheel module is composed of two half-circle legs and can be driven to rotate and move radially with respect to the hip joint, where the motion is similar to the definition of the 2-degrees-of-freedom polar coordinates. When the legs form a circle and rotate, the robot moves like an ordinary four-wheeled vehicle, and when the legs are extended, the robot moves like a quadruped.
The leg-wheels of the robot can be replaced by another type of leg-wheel shown in the image below. By using the passive torsion spring, the mechanism can react rapidly to external forces. In addition, by actively using only the rotational DOF, this method also reduces the loading of the DC motors.
The CompactRIO cRIO-9024 and cRIO-9118 were chosen as the computation resource of the robot, owing to their robust, modular, rugged, and real-time characteristics. The improved performance of the new CompactRIO in comparison with their predecessor cRIO-9014 and cRIO-9104 (installed in the old version of the robot) allows to implement bio-inspired, high-level, and calculation-intensive control strategies for robot behavior.
On the cRIO-9118, the Field Programmable Gate Arrays (FPGAs) is used to link the real-time control system to the I/O hardware which is linked to custom motor driving boards and various feedback sensors. The FPGA makes it possible to accomplish a simple and flexible low-level control system in a short time. When high-speed computing is needed in motor position and current control, the FPGA provides a fast and powerful way to address the signal processing method and PID control. With the power the FPGA, the real-time program can be greatly simplified, reserving the computing resource for a complex high-level control system.
Six NI I/O modules are installed on the cRIO, including (1) two NI 9401 high-speed D I/O modules for receiving high-speed motor encoder signals; (2) one NI 9403 D I/O module for interfacing digital sensory inputs (i.e., hall-effect sensors, temperature sensors, etc.); (3) one NI 9205 AI module for interfacing analog sensor inputs (i.e., inclinometer, potentiometer, infrared sensor, an inertial measurement unit etc.); (4) one NI 9264 AO module for controlling DC motors and (5) one NI 9227 current input module to implement force control by current feedback. The structure of the mechatronic system of the robot is shown in the image below.
The LabVIEW program of the robot includes three portions: the main control program run in the microprocessor (RT), the interfacing program executed on the FPGA, and the GUI program installed on a laptop. The GUI mainly has four functions: sending high-level control commands to the robot, monitoring the robot’s health, adjusting control parameters, and data logging for further analysis.
In evaluating the performance of TurboQuad, the quantitative analysis is an important process. Owing to the changing computing resource demand in a complex real-time control system, some data points may be missed when using the traditional write to file function. In contrast, the FIFO function is a reliable way to record precise experiment data under heavy loading conditions.
In addition, LabVIEW is also used to simulate the trajectory of the robot. For different gaits/modes of the robot, coordination of the leg-wheel should be maintained. During the development stage, the planned trajectory is first examined by the simulator before being sent directly to the robot. Figure 7 shows the trajectory simulator for the walking gait.
Bio-inspired control strategy
The theory behind the Central Pattern Generator (CPG) states that the CPG located in the spinal cord of animals can generate rhythmic neural signals without sensory inputs or accurate calculations from the brain. This neural signal is then modified according to sensory feedback signals and high-level commands to generate suitable gaits and trajectories for different terrain adaptation. In such a way, the animals can exhibit smooth gait transitions. A similar strategy is programmed into the robot.
The CPG applied to TurboQuad consists of three layers: the command layer for high-level commands like speed and gait, the mode generating layer for trajectory generation and the motion layer for motor PD control. There are four dependent oscillators for each of the four legs. Each of the oscillators is linked to its adjacent ones, which makes coordinating the positions of all legs possible. Three modes/gaits and two model-based gaits are implemented, including a high-speed wheeled mode for use on flat terrain, a stable low-speed legged walking gait for use on rough terrain, a legged trotting gait for use on medium-rough terrain and two legged pronking gaits based on R-SLIP model and SLIP models. Based on the CPG strategy, the legs-wheels in all modes/gaits are coordinated in a rhythmic manner and the mode/gait can transit rapidly and smoothly. Figure 8 demonstrates the mode/gait transition for the three modes. Figure 9 shows a pronking gait based on the R-SLIP model with torsion spring leg-wheel and a pronking gait based on the SLIP model using force control with 2-DOF mechanisms.
Describes the design, control strategy, implementation, and performance evaluation of a leg–wheel transformable robot. Also discusses a bio-inspired control strategy that is applied based on a central pattern generator and coupled oscillator networks.