FT-CPG: Learning Central Pattern Generators for Fault-Tolerant Quadruped Locomotion under Multi-Joint Failures
Pei Zhang, Zhaobo Hua, Qiyu Qiu, Jinliang Ding
AI summary
Problem
Quadruped robots in rescue and exploration face unpredictable concurrent multi-joint leg failures that compromise mobility and safety. Existing controllers lack robust fault-tolerant capabilities for these complex scenarios, often failing to maintain omnidirectional movement or requiring extensive manual tuning.
Approach
The authors propose FT-CPG, a model-free reinforcement learning framework that uses a biomimetic central pattern generator modulated by two neural policies—one for normal gaits and another for fault-specific adjustments—along with a discriminator to detect failures and generate adaptive, safe locomotion.
Key results
- Enables rapid locomotion recovery under concurrent multi-joint locking and power loss faults
- Maintains full omnidirectional velocity tracking despite leg failures
- Achieves successful zero-shot sim-to-real transfer on a physical Unitree Go1 robot
- Eliminates the need for complex mathematical modeling or manual joint constraint tuning
Why it matters
This framework significantly enhances the safety and operational reliability of quadruped robots in rescue and exploration missions by enabling autonomous recovery from complex hardware failures.
Abstract
Quadruped robots used for rescue and exploration are susceptible to various leg failures, where unpredictable joint locking or power loss can pose an immediate risk of falling. Traditional controllers lack fault-tolerant control capabilities in the case of multi-joint concurrent faults, and erroneous controller outputs may lead to robot damage. This paper proposes a model-free reinforcement learning framework based on central pattern generators (CPG) for fault-tolerant control (FT-CPG). The framework uses biomimetic gait generation and section- wise training to address various types of multi-joint concurrent faults. FT-CPG adopts a fault-tolerant CPG module to generate safe gaits, while utilizing neural network-based policies to infer failures and coordinate the rhythmic behaviors of the CPG, ensuring the ability to track velocity commands under fault con- ditions. Experiments show that FT-CPG is robust in unexpected situations, where a single leg experiences failures across any number of joints, with each joint randomly encountering locking or power loss faults. Furthermore, the proposed framework preserves the robot’s omnidirectional mobility. Finally, zero-shot sim-to-real transfer was successfully implemented on the real- world Unitree Go1 robot, effectively addressing various multi- joint leg failures.