mlsanityData to 7 Oct 2026

Uniformly Safe RL with Objective Suppression for Multi-Constraint Safety-Critical Applications

arXiv 2402.15650 ↗First posted 23 Feb 2024cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Safe reinforcement learning tasks are a challenging domain despite being very common in the real world. The widely adopted CMDP model constrains the risks in expectation, which makes room for dangerous behaviors in long-tail states. In safety-critical domains, such behaviors could lead to disastrous outcomes. To address this issue, we first describe the problem with a stronger Uniformly Constrained MDP (UCMDP) model where we impose constraints on all reachable states; we then propose Objective Suppression, a novel method that adaptively suppresses the task reward maximizing objectives according to a safety critic, as a solution to the Lagrangian dual of a UCMDP. We benchmark Objective Suppression in two multi-constraint safety domains, including an autonomous driving domain where any incorrect behavior can lead to disastrous consequences. On the driving domain, we evaluate on open source and proprietary data and evaluate transfer to a real autonomous fleet. Empirically, we demonstrate that our proposed method, when combined with existing safe RL algorithms, can match the task reward achieved by baselines with significantly fewer constraint violations.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 22 of its 30 references that are in the map. First posted on arXiv 23 Feb 2024.

Builds on

22 of 30 references.

References · 22

  1. 2013Playing Atari with Deep Reinforcement Learningcited by 3,211
  2. 2017CARLA: An Open Urban Driving Simulatorcited by 2,431
  3. 2016Concrete Problems in AI Safetycited by 1,582
  4. 2017Constrained Policy Optimizationcited by 928
  5. 2020Deep Reinforcement Learning for Autonomous Driving: A Surveycited by 698
  6. 2018Reward Constrained Policy Optimizationcited by 397
  7. 2016Deep Reinforcement Learning for Robotic Manipulation with Asynchronous Off-Policy Updatescited by 342
  8. 2015Risk-Constrained Reinforcement Learning with Percentile Risk Criteriacited by 320
  9. 2020Responsive Safety in Reinforcement Learning by PID Lagrangian Methodscited by 302
  10. 2016Safe, Multi-Agent, Reinforcement Learning for Autonomous Drivingcited by 301
  11. 2019End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Taskscited by 256
  12. 2018Safe Exploration in Continuous Action Spacescited by 237
10 more
  1. 2022A Review of Safe Reinforcement Learning: Methods, Theory and Applicationscited by 169
  2. 2023Waymax: An Accelerated, Data-Driven Simulator for Large-Scale Autonomous Driving Researchcited by 158
  3. 2020Recovery RL: Safe Reinforcement Learning with Learned Recovery Zonescited by 156
  4. 2022Imitation Is Not Enough: Robustifying Imitation with Reinforcement Learning for Challenging Driving Scenarioscited by 124
  5. 2020Conservative Safety Critics for Explorationcited by 107
  6. 2023ReasonNet: End-to-End Driving with Temporal and Global Reasoningcited by 92
  7. 2017OptLayer - Practical Constrained Optimization for Deep Reinforcement Learning in the Real Worldcited by 64
  8. 2023State-wise Safe Reinforcement Learning: A Surveycited by 48
  9. 2022Towards Safe Reinforcement Learning with a Safety Editor Policycited by 28
  10. 2022Continuously Discovering Novel Strategies via Reward-Switching Policy Optimizationcited by 23

Built on by

Not cited yet.