Robust RL

RL: Reinforcement Learning

Overview

Robust reinforcement learning studies how agents can make reliable decisions when dynamics, observations, rewards, or other agents differ from training conditions. Common approaches include robust Markov decision processes and distributionally robust optimization, adversarial training and domain randomization, and policies that use history or uncertainty estimates to adapt. Safety shields and formal verification add ways to constrain or certify behavior, with evaluation spanning control benchmarks, robotics, driving, and offline or multi-agent settings. A central tension is that worst-case protection can make policies overly conservative, while nominal performance may not predict transfer to unseen shifts; measuring that trade-off remains a key challenge.

Latest papers

CardsList

246 papers23 papers in the last four weeks, down 8% on the four weeks before. 0.2% of all new papers.

  1. PA-LOCO: Learning Perturbation-Adaptive Locomotion for Quadruped Robots

    Jul 5, 2024Zhiyuan Xiao, Xinyu Zhang, Xiang Zhou +1RL for Legged LocomotionQuadruped Locomotion

  2. Lipschitz-Regularized Critics Lead to Policy Robustness Against Transition Dynamics Uncertainty

    Apr 22, 2024Xulin Chen, Ruipeng Liu, Zhenyu Gan +1Robotic RLRobust RL

  3. Distributionally Robust Reinforcement Learning with Interactive Data Collection: Fundamental Hardness and Near-Optimal Algorithms

    Apr 4, 2024Miao Lu, Han Zhong, Tong Zhang +1Distributionally Robust RLRobust RL

  4. Corruption Robust Offline Reinforcement Learning with Human Feedback

    Feb 9, 2024Debmalya Mandal, Andi Nika, Parameswaran Kamalaruban +2Offline RLCorruption Robustness

  5. Robust Recurrent Reinforcement Learning under Evolving Hidden Disturbances with Application to Rover Wheel Slip

    Jul 29, 2023Saki Omi, Hyo-Sang Shin, Namhoon Cho +2RL for RoboticsReinforcement Learning

  6. Security of Deep Reinforcement Learning for Autonomous Driving: A Survey

    Dec 12, 2022Ambra Demontis, Srishti Gupta, Maura Pintor +6Robust RLRL for Autonomous Driving