Controlling Underestimation Bias in Constrained Reinforcement Learning for Safe Exploration

• URL: http://arxiv.org/abs/2601.11953v1
• Date: Sat, 17 Jan 2026 08:02:51 GMT
• Title: Controlling Underestimation Bias in Constrained Reinforcement Learning for Safe Exploration
• Authors: Shiqing Gao, Jiaxin Ding, Luoyi Fu, Xinbing Wang,
• Abstract summary: Constrained Reinforcement Learning (CRL) aims to maximize cumulative rewards while satisfying constraints.<n>Existing CRL algorithms often encounter significant constraint violations during training, limiting their applicability in safety-critical scenarios.<n>We propose the Memory-driven Intrinsic Cost Estimation (MICE) method, which introduces intrinsic costs to mitigate underestimation.
• Score: 42.720459553340326
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Constrained Reinforcement Learning (CRL) aims to maximize cumulative rewards while satisfying constraints. However, existing CRL algorithms often encounter significant constraint violations during training, limiting their applicability in safety-critical scenarios. In this paper, we identify the underestimation of the cost value function as a key factor contributing to these violations. To address this issue, we propose the Memory-driven Intrinsic Cost Estimation (MICE) method, which introduces intrinsic costs to mitigate underestimation and control bias to promote safer exploration. Inspired by flashbulb memory, where humans vividly recall dangerous experiences to avoid risks, MICE constructs a memory module that stores previously explored unsafe states to identify high-cost regions. The intrinsic cost is formulated as the pseudo-count of the current state visiting these risk regions. Furthermore, we propose an extrinsic-intrinsic cost value function that incorporates intrinsic costs and adopts a bias correction strategy. Using this function, we formulate an optimization objective within the trust region, along with corresponding optimization methods. Theoretically, we provide convergence guarantees for the proposed cost value function and establish the worst-case constraint violation for the MICE update. Extensive experiments demonstrate that MICE significantly reduces constraint violations while preserving policy performance comparable to baselines.

Related papers

• Conditional Sequence Modeling for Safe Reinforcement Learning [8.858563919623082]
  offline safe reinforcement learning aims to learn policies from a fixed dataset while maximizing performance under cumulative cost constraints.<n>Most existing offline safe RL methods are trained under a pre-specified threshold.<n>We propose RCDT, a CSM-based method that supports zero-shot deployment across multiple cost thresholds within a single trained policy.
  arXiv  Detail & Related papers  (2026-02-09T12:22:57Z)
• Conformal Thinking: Risk Control for Reasoning on a Compute Budget [60.65072883773352]
  Reasoning Large Language Models (LLMs) enable test-time scaling, with dataset-level accuracy improving as the token budget increases.<n>We re-frame the budget setting problem as risk control, limiting the error rate while minimizing compute.<n>Our framework introduces an upper threshold that stops reasoning when the model is confident and a novel lower threshold that preemptively stops unsolvable instances.
  arXiv  Detail & Related papers  (2026-02-03T18:17:22Z)
• Information-Theoretic Reward Modeling for Stable RLHF: Detecting and Mitigating Reward Hacking [78.69179041551014]
  We propose an information-theoretic reward modeling framework based on the Information Bottleneck principle.<n>We show that InfoRM filters out preference-irrelevant information to alleviate reward misgeneralization.<n>We also introduce IBL, a distribution-level regularization that penalizes such deviations, effectively expanding the optimization landscape.
  arXiv  Detail & Related papers  (2025-10-15T15:51:59Z)
• Boundary-to-Region Supervision for Offline Safe Reinforcement Learning [56.150983204962735]
  Boundary-to-Region (B2R) is a framework that enables asymmetric conditioning through cost signal realignment.<n>B2R redefines CTG as a boundary constraint under a fixed safety budget, unifying the cost distribution of all feasible trajectories.<n> Experimental results show that B2R satisfies safety constraints in 35 out of 38 safety-critical tasks.
  arXiv  Detail & Related papers  (2025-09-30T03:38:20Z)
• Cost-aware Stopping for Bayesian Optimization [46.95172329282389]
  We propose a cost-aware stopping rule for Bayesian optimization that adapts to varying evaluation costs and is free of tuning.<n>We prove a theoretical guarantee bounding the expected cumulative evaluation cost incurred by our stopping rule when paired with state-of-the-art acquisition functions.
  arXiv  Detail & Related papers  (2025-07-16T17:54:14Z)
• Off-Policy Primal-Dual Safe Reinforcement Learning [16.918188277722503]
  We show that the error in cumulative cost estimation causes significant underestimation of cost when using off-policy methods. We propose conservative policy optimization, which learns a policy in a constraint-satisfying area by considering the uncertainty in estimation. We then introduce local policy convexification to help eliminate such suboptimality by gradually reducing the estimation uncertainty.
  arXiv  Detail & Related papers  (2024-01-26T10:33:38Z)
• Imitate the Good and Avoid the Bad: An Incremental Approach to Safe Reinforcement Learning [11.666700714916065]
  Constrained RL is a framework for enforcing safe actions in Reinforcement Learning. Most recent approaches for solving Constrained RL convert the trajectory based cost constraint into a surrogate problem. We present an approach that does not modify the trajectory based cost constraint and instead imitates good'' trajectories.
  arXiv  Detail & Related papers  (2023-12-16T08:48:46Z)
• Safe Deep Reinforcement Learning by Verifying Task-Level Properties [84.64203221849648]
  Cost functions are commonly employed in Safe Deep Reinforcement Learning (DRL) The cost is typically encoded as an indicator function due to the difficulty of quantifying the risk of policy decisions in the state space. In this paper, we investigate an alternative approach that uses domain knowledge to quantify the risk in the proximity of such states by defining a violation metric.
  arXiv  Detail & Related papers  (2023-02-20T15:24:06Z)
• Solving Richly Constrained Reinforcement Learning through State Augmentation and Reward Penalties [8.86470998648085]
  Key challenge is handling expected cost accumulated using the policy. Existing methods have developed innovative ways of converting this cost constraint over entire policy to constraints over local decisions. We provide an equivalent unconstrained formulation to constrained RL that has an augmented state space and reward penalties.
  arXiv  Detail & Related papers  (2023-01-27T08:33:08Z)
• AutoCost: Evolving Intrinsic Cost for Zero-violation Reinforcement Learning [3.4806267677524896]
  We propose AutoCost, a framework that automatically searches for cost functions that help constrained RL to achieve zero-violation performance. We compare the performance of augmented agents that use our cost function to provide additive intrinsic costs with baseline agents that use the same policy learners but with only extrinsic costs.
  arXiv  Detail & Related papers  (2023-01-24T22:51:29Z)
• COptiDICE: Offline Constrained Reinforcement Learning via Stationary Distribution Correction Estimation [73.17078343706909]
  offline constrained reinforcement learning (RL) problem, in which the agent aims to compute a policy that maximizes expected return while satisfying given cost constraints, learning only from a pre-collected dataset. We present an offline constrained RL algorithm that optimize the policy in the space of the stationary distribution. Our algorithm, COptiDICE, directly estimates the stationary distribution corrections of the optimal policy with respect to returns, while constraining the cost upper bound, with the goal of yielding a cost-conservative policy for actual constraint satisfaction.
  arXiv  Detail & Related papers  (2022-04-19T15:55:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.