Directional Compliance in Snake Robot Obstacle-Aided Locomotion

This work investigates the role of compliance in high-degree-of-freedom snake robots. Prior work used geometric methods to reduce the dimension of the robot’s state space to efficiently control the robot: instead of controlling all of the robot’s degrees-of-freedom, we need only control two for meaningful motions. Recently, we adapted this technique to handle unmodeled terrain irregularities. We defined a compliant controller on the reduced state space, which used torque information measured at the robot’s joints. This improved the robot’s performance, but still the robot arrived at stuck configurations. To address this problem, we introduce a biologically inspired strategy called directional compliance, which selectively allows some portions of the robot to comply to the environment, while others remain stiff to push off for forward progression. Unfortunately, with pure directional compliance method, the robot can still get stuck. Therefore, we develop an estimator to determine the state of the robot and a controller that switches between compliance and directional compliance modes to help the robot get unstuck. We experimentally find that our method enables the snake robot to locomote more consistently than the pure compliant controller in obstacle-rich environments.