Gait Synthesis and Resilient Task Planning for Ambulating Soft Robots

Abstract

Soft robots are capable of motions and environmental interactions that are not typically achievable by rigid robotic systems. In addition, the materials and fabrication methods used often result in actuators that can be designed and produced quickly. The combination of these attributes presents an opportunity for developing rapidly manufacturable and versatile soft robots capable of addressing a variety of environments and tasks. However, taking full advantage of these properties requires tools to automatically synthesize motions and controllers for these systems while also addressing the potential for actuator failure. This work addresses the challenges of motion and control synthesis for ambulating soft robots. First, we present an algorithm for synthesizing gaits for arbitrarily composed, modular soft robots. The algorithm, which makes no assumptions about dynamics or specific soft material models, is demonstrated on two different composed robots. Second, we present a framework for synthesizing controllers for multigait soft robots that are resilient to actuator failure. This abstraction, which utilizes Linear Temporal Logic (LTL) to encode multigait behavior and a sensor based abstraction of actuator performance, is demonstrated on a physical robot. Finally, we present methods for modeling ambulating soft robots as a precursor towards gait synthesis.