Soft Growing Robot Explore Unknown Environments Through Obstacle Interaction

Abstract

In low-light, unstructured, and confined environments, performing Simultaneous Localization and Mapping (SLAM) with conventional methods presents significant challenges. Soft growing robots, characterized by their compliance and extensibility, interact safely with the environment, making them well-suited for navigation in such environments. Through collision-based guidance, the robot can gather environmental data via morphological adaptations. Based on this, we developed the sensing capabilities of the soft growing robot, retaining its flexibility while enabling effective environmental interaction and perception. The robot employs a gyroscope combined with an encoder to track the end-effector trajectory and uses flexible proximity sensing to detect obstacles. By fusing the information from these sensors, we propose a multi-sensor fusion strategy for environmental exploration of the soft growing robot. The robot navigates unknown environments by employing pre-bending based on prior environmental data and utilizing pneumatic artificial muscles. In multi-obstacle environmental exploration, the path prediction error is less than 3.5% of the robot's total length, enabling greater environmental coverage with fewer exploration attempts.