Position control of hybrid pneumatic–electric actuators using discrete-valued model-predictive control

Abstract

The design, modeling and position control of a novel hybrid pneumatic–electric actuator for applications in robotics and automation is presented. The design incorporates a pneumatic cylinder and DC motor connected in parallel. By avoiding the need for a high ratio transmission, the design greatly reduces the mechanical impedance that can make collisions with conventionally actuated robot arms dangerous. A novel discrete-valued model-predictive control (DVMPC) algorithm is proposed for controlling the position of the pneumatic cylinder with inexpensive on/off solenoid valves. A variant of inverse dynamics control is proposed for the DC motor. A prototype was built for validating the actuator design and control algorithms. It is used to rotate a single-link robot arm. The actuator inertia and static friction torque values for the prototype were only 0.6% and 4%, respectively, of the values found for a comparable actuator from an industrial robot. Simulation results for position control of pneumatic actuators with different valve speeds and friction coefficients show that the DVMPC algorithm outperforms a sliding mode control algorithm in terms of position error and expected valve life. Experimental results are presented for vertical rotary cycloidal trajectories. Even with the poor quantization caused by the on/off valves, the pneumatic cylinder controlled by the proposed DVMPC algorithm achieved a 0.7% root mean square error (RMSE) and a 0.25% steady-state error (SSE). With the addition of the DC motor to form the hybrid actuator, the RMSE and SSE were reduced to 0.12% and 0.04%, respectively. By incorporating a novel estimation algorithm, the system was made robust to an unknown payload.