← Back ICRA 2026

Goal-Oriented Control Strategies for Soft Growing Robots

Wentao Huang, Pengchun Li, Ziyi Zhang, Zuankai Wang, Dekai Zhou, Longqiu Li

PDF

AI summary

Key figure (auto-extracted from paper)

A dual-thread control strategy combining graph convolutional networks and deep reinforcement learning enables soft growing robots to accurately track dynamic targets over long distances.

Soft growing robots Reinforcement learning Path planning Motion control Graph convolutional networks Dynamic target tracking

Problem

Soft growing robots lack reliable control and precision in dynamic environments due to their soft structure, nonlinear mechanics, and low stiffness. Existing planning and feedback methods struggle to balance global path optimization with real-time motion correction.

Approach

The method integrates graph convolutional networks with deep reinforcement learning for global path planning, coupled with a multi-loop closed-loop motion controller that continuously corrects deformation and tracking errors using real-time sensor feedback.

Key results

Achieves 11.83 mm tracking accuracy over a 5-meter range
Successfully tracks and approaches a non-cooperative dynamic target
Improves computational efficiency and planning accuracy over inverse kinematics methods
Effectively mitigates deformation errors from low stiffness and external disturbances

Why it matters

Enables reliable, goal-oriented navigation and manipulation for soft robots in dynamic environments, advancing applications in search, rescue, and inspection.

Abstract

Soft growing robots, as highly mobile pneumatic membrane robots, are limited in control performance due to their soft structure and nonlinear mechanical properties, especially un- der dynamic conditions. Therefore, developing reliable control strategies for the robot is essential. This study proposes a dual- thread, goal-oriented control strategy for soft growing robot that combines planning and control. By integrating graph convolutional networks with deep reinforcement learning, the global path plan- ning method is better suited to the self-growing behaviors of soft robots, leading to improvements in both computational efficiency and accuracy compared to inverse kinematics planning methods. Motion control reduces the adverse effects of deformation errors caused by its own low stiffness or by disturbances in the exter- nal environment. This strategy effectively combines reinforcement learning-based global planning with a multiple closed-loop motion control system, addressing the issues of low precision and reliability under dynamic conditions. Experimental results demonstrate that therobotachievesatrackingaccuracyof11.83mmwithina5-meter range and successfully tracks and approaches a non-cooperative dynamic target. These results highlight the significant potential of the proposed approach in applications such as target capture and dynamic manipulation.

Index terms

Modeling Control and Learning for Soft Robots Motion and Path Planning Reinforcement Learning