Reinforcement Learning–Based Control Policy Optimization for Transtibial Prosthesis via Physics-Informed Simulation

J. He, S. Krishnan, M. Lancastre, E. Zhao

Lab: Meta Mobility Lab Learn more

Professor: Inseung Kang Learn more

Pending arXiv preprint; planned submission to relevant conferences

We compare imitation learning and PPO-based RL controllers for ankle assistive devices in MuJoCo, demonstrating that IL yields more stable gait reproduction and longer walking durations under varied state-space and prosthetic configurations.

PDF Code

Multi-Agent Reinforcement Learning with Large Language Models for Safe Path Planning

R. Graves, A. Chase, S. Krishnan, E. Zhao, J. Ali

Lab: SAFE AI Lab Learn more

Professor: Ding Zhao Learn more

Pending arXiv preprint; planned submission to relevant conferences

We integrate GPT-4 into a hierarchical MARL framework to classify obstacle severity and dynamically adjust penalty-based rewards, enabling context-aware, safety-driven path planning in unstructured environments.

PDF Code