Locomotion Policy
Locomotion policy research focuses on developing control algorithms that enable robots, particularly legged robots, to move effectively and robustly in diverse environments. Current research emphasizes learning-based approaches, often employing reinforcement learning (RL) with various architectures like Central Pattern Generators (CPGs) and model predictive control (MPC), sometimes integrated with deep learning for improved stability and safety guarantees. This field is crucial for advancing robotics, enabling more agile and adaptable robots for applications ranging from search and rescue to industrial automation and even assisting humans in challenging terrains.
Papers
An Open-Loop Baseline for Reinforcement Learning Locomotion Tasks
Antonin Raffin, Olivier Sigaud, Jens Kober, Alin Albu-Schäffer, João Silvério, Freek Stulp
DecAP: Decaying Action Priors for Accelerated Imitation Learning of Torque-Based Legged Locomotion Policies
Shivam Sood, Ge Sun, Peizhuo Li, Guillaume Sartoretti