Machine learning-driven innovation design of clustered tensegrity continuum robot

Abstract

Tensegrity structures have the advantage of superior deformation ability and high load-to-weight ratio, making them potential candidates for cable-driven continuum robot design. However, designing a clustered tensegrity continuum robot is still challenging due to the difficulty in modeling the tensegrity structure. In this study, we propose an innovation design method for a clustered tensegrity continuum robot based on machine learning (ML). Our ML-driven design method includes topology design of the clustered tensegrity continuum robot using genetic algorithm (GA) and deep reinforcement learning (DRL) approach, and driving law design (motion planning) of the continuum robot using deep reinforcement learning. We emphasize the obstacle avoidance and reaching point motion as one of the most important challenges for a cable-driven continuum robot and design the topology and driving law of the clustered tensegrity continuum robot through ML-based approach. This study demonstrates the applicability of tensegrity structures in the field of clustered tensegrity continuum robot design and illustrates the feasibility of using machine learning in the design of clustered tensegrity continuum robot.