Morphology-Aware Graph Reinforcement Learning for Tensegrity Robot Locomotion

TL;DR

This paper presents a graph neural network-based reinforcement learning framework that effectively controls tensegrity robots by capturing their physical structure, resulting in improved learning speed, robustness, and real-world transferability.

Contribution

The paper introduces a morphology-aware GNN integrated with SAC for tensegrity robot control, enabling faster, more stable learning and direct simulation-to-hardware transfer.

Findings

GNN-based policies outperform MLP policies in learning speed and stability.

The method achieves robust locomotion across different primitives and noise conditions.

Policies transfer seamlessly from simulation to real hardware without fine-tuning.

Abstract

Tensegrity robots combine rigid rods and elastic cables, offering high resilience and deployability but at the same time posing major challenges for locomotion control due to their underactuated and highly coupled dynamics. This paper introduces a morphology-aware reinforcement learning framework that integrates a graph neural network (GNN) into the Soft Actor-Critic (SAC) algorithm. By representing the robot's physical topology as a graph, the proposed GNN-based policy captures coupling among components, enabling faster and more stable learning than conventional multilayer perceptron (MLP) policies. The method is validated on a physical 3-bar tensegrity robot across three locomotion primitives, including straight-line tracking and bidirectional turning. It shows superior sample efficiency, robustness to noise and stiffness variations, and improved trajectory accuracy. Additionally, the learned policies transfer directly from simulation to hardware without fine-tuning, achieving stable real-world locomotion. These results demonstrate the advantages of incorporating structural priors into reinforcement learning for tensegrity robot control.