An Automatic Navigation Framework for Magnetic Fish-Like Millirobot in Uncertain Dynamic Environments

Abstract

Magnetic soft robots have shown great potential in various applications, including targeted delivery and therapy. However, current magnetic soft robots lack sufficient intelligence for autonomous decision-making, limiting their ability to adapt to dynamic and unstructured environments. To address these limitations, this study designs a magnetic fish-like millirobot and proposes an innovative control framework inspired by digital twin technology. By integrating a virtual agent and a leader-follower mechanism, this approach enables magnetically controlled fish-like robot to autonomously navigate and make decisions in both dynamic and unknown environments. To further enhance this framework, we developed a fuzzy logic improved dynamic window algorithm for intelligent obstacle avoidance, based on an analysis of the robot's kinematic characteristics. Through rigorous experiments combining virtual and real-world environments, we validated the efficacy of our proposed control framework. The results unequivocally demonstrate substantial improvements in autonomous navigation and intelligent responsiveness to dynamic environments, indicating a promising pathway for the application of magnetically controlled soft robots in complex environments.