mlsanityData to 7 Oct 2026

Revisiting Safe Exploration in Safe Reinforcement learning

arXiv 2409.01245 ↗First posted 2 Sep 2024cs.LG cs.AI cs.ROcited by 0

Not on the map. Why?

Abstract (arXiv)

Safe reinforcement learning (SafeRL) extends standard reinforcement learning with the idea of safety, where safety is typically defined through the constraint of the expected cost return of a trajectory being below a set limit. However, this metric fails to distinguish how costs accrue, treating infrequent severe cost events as equal to frequent mild ones, which can lead to riskier behaviors and result in unsafe exploration. We introduce a new metric, expected maximum consecutive cost steps (EMCC), which addresses safety during training by assessing the severity of unsafe steps based on their consecutive occurrence. This metric is particularly effective for distinguishing between prolonged and occasional safety violations. We apply EMMC in both on- and off-policy algorithm for benchmarking their safe exploration capability. Finally, we validate our metric through a set of benchmarks and propose a new lightweight benchmark task, which allows fast evaluation for algorithm design.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 16 of its 24 references that are in the map. First posted on arXiv 2 Sep 2024.

Builds on

16 of 24 references.

References · 16

  1. 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,635
  2. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  3. 2016Concrete Problems in AI Safetycited by 1,582
  4. 2017Constrained Policy Optimizationcited by 928
  5. 2015Risk-Constrained Reinforcement Learning with Percentile Risk Criteriacited by 320
  6. 2020Responsive Safety in Reinforcement Learning by PID Lagrangian Methodscited by 302
  7. 2018A Lyapunov-based Approach to Safe Reinforcement Learningcited by 278
  8. 2015Risk-Sensitive and Robust Decision-Making: a CVaR Optimization Approachcited by 162
  9. 2019Lyapunov-based Safe Policy Optimization for Continuous Controlcited by 140
  10. 2020Learning to be Safe: Deep RL with a Safety Criticcited by 109
  11. 2020Learning to Walk in the Real World with Minimal Human Effortcited by 103
  12. 2023State-wise Safe Reinforcement Learning: A Surveycited by 48
4 more
  1. 2022Towards Safe Reinforcement Learning via Constraining Conditional Value-at-Riskcited by 40
  2. 2021Joint Synthesis of Safety Certificate and Safe Control Policy using Constrained Reinforcement Learningcited by 25
  3. 2015Safe Policy Search for Lifelong Reinforcement Learning with Sublinear Regretcited by 15
  4. 2024Constrained Reinforcement Learning with Smoothed Log Barrier Functioncited by 12

Built on by

Not cited yet.