Optimizing actual PID control for walking quadruped soft robots using genetic algorithms.

The construction of soft robots's models and controllers remains a significant challenge. In this paper, we propose a new walking control method for the quadruped soft robot named genetic algorithm-optimized PID. First, we construct the control model correlating valve voltage with leg bending based on the geometrical analysis. This modeling approach leverages the characteristics of novel leg structure and bend sensor, thereby streamlining the control model for locomotion of quadruped soft robotic. Moreover, We apply the genetic algorithm to automatically tune parameters and optimize PID controllers, aiming to enhance control performance. The application of the proposed method to the walking control has been uniquely demonstrated on a real 3D-printed quadruped soft robot. Experimental results indicate that the genetic algorithm-optimized PID controller significantly improves trajectory tracking compared to the Ziegler-Nichols tuning method. This optimization increases the robot's walking speed from 5 mm/s to 8 mm/s, reduces the error rate by 2.4064%, decreases overshoot by 12.55%, and shortens response time by 0.5 s, substantially enhancing the controller's overall performance. Additionally, compared to particle swarm optimization, the proposed method further improves performance by reducing the error rate by 0.4079%, overshoot by 8.4%, and response time by 1.0 s.