Threat-guided Policy-aware Scene Perturbation for Safe Autonomous Driving with Online Reinforcement Learning

• URL: http://arxiv.org/abs/2608.10403v1
• Date: Tue, 11 Aug 2026 02:49:50 GMT
• Title: Threat-guided Policy-aware Scene Perturbation for Safe Autonomous Driving with Online Reinforcement Learning
• Abstract summary: We propose Threat-guided Policy-aware Scene Perturbation (TPSP) for safe autonomous driving with online RL.<n>TPSP captures the interaction between policy behaviors and surrounding environments, enabling scene perturbation aligned with the current policy.<n>Experiments demonstrate that TPSP improves safety learning efficiency, achieving strong safety performance on NAVSIM v2 with approximately 4 million kilometers of simulated driving data.
• Score: 21.268329695683203
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reinforcement learning (RL) has shown promising performance in autonomous driving, yet ensuring the safety of online RL policies remains challenging due to insufficient exposure to safety-critical driving scenes. The long-tailed nature of real-world traffic situations makes dangerous and rare interactions difficult to encounter through conventional sampling, limiting the ability of RL policies to learn robust safety behaviors. Existing methods improve training diversity by synthesizing challenging scenes or adversarial situations. However, these approaches typically optimize scene generation objectives separately from the evolving policy, without explicitly modeling how generated perturbations relate to the current policy's weaknesses and learning needs. In this paper, we propose Threat-guided Policy-aware Scene Perturbation (TPSP) for safe autonomous driving with online RL. TPSP introduces a policy-aware scene encoder to capture the interaction between policy behaviors and surrounding environments, enabling scene perturbation aligned with the current policy. Based on this representation, TPSP selectively perturbs critical objects rather than applying uniform modifications across the scene. Furthermore, we develop a threat-guided optimization strategy that evaluates perturbed scenes through threat-level differences between policy rollouts on original and perturbed scenes, guiding the generation of safety-critical scenes with higher training value. Comprehensive experiments demonstrate that TPSP improves safety learning efficiency, achieving strong safety performance on NAVSIM v2 with approximately 4 million kilometers of simulated driving data. Ablation studies verify that policy-aware targeted perturbations provide more informative safety-critical experiences than random or policy-unaware strategies, enabling safer driving under limited interaction budgets.

Related papers

• World Engine: Towards the Era of Post-Training for Autonomous Driving [63.61277741443647]
  We introduce World Engine, a generative framework that reconstructs high-fidelity interactive environments from real-world logs.<n>World Engine substantially reduces failures in rare safety-critical scenarios and yields significantly larger gains than scaling pre-training data alone.<n>When deployed on a production-scale autonomous driving system, the resulting policy reduces simulated collisions.
  arXiv  Detail & Related papers  (2026-06-18T06:28:33Z)
• From Attacks to Curricula: Learnability-Guided Adversarial Training for Safe Autonomous Driving [56.30087557121323]
  AlignADV is a learnability-guided closed-loop adversarial training framework.<n>It converts adversarial scenarios into resolvable and capability-aligned curricula.<n> Experiments show it reduces training steps by up to 40.6 percent.
  arXiv  Detail & Related papers  (2026-06-12T02:13:55Z)
• Transferable Reinforcement Learning via Probabilistic Latent Embeddings and Dynamic Policy Adaptation for Sim-to-Real Deployment [5.938653166789926]
  Deep reinforcement learning (RL) agents for many cyber-physical systems are first trained in simulators.<n>When deployed in real world environments, they often suffer from performance degradation or safety violations because of the inevitable Sim2Real gap.<n>We propose a novel reinforcement learning framework that enables safe and efficient policy transfer via probabilistic latent embeddings and dynamic policy adaptation.
  arXiv  Detail & Related papers  (2026-05-26T20:22:44Z)
• RACER: Epistemic Risk-Sensitive RL Enables Fast Driving with Fewer Crashes [57.319845580050924]
  We propose a reinforcement learning framework that combines risk-sensitive control with an adaptive action space curriculum. We show that our algorithm is capable of learning high-speed policies for a real-world off-road driving task.
  arXiv  Detail & Related papers  (2024-05-07T23:32:36Z)
• Safety-aware Causal Representation for Trustworthy Offline Reinforcement Learning in Autonomous Driving [33.672722472758636]
  offline Reinforcement Learning(RL) approaches exhibit notable efficacy in addressing sequential decision-making problems from offline datasets. We introduce the saFety-aware strUctured Scenario representatION ( Fusion) to facilitate the learning of a generalizable end-to-end driving policy. Empirical evidence in various driving scenarios attests that Fusion significantly enhances the safety and generalizability of autonomous driving agents.
  arXiv  Detail & Related papers  (2023-10-31T18:21:24Z)
• Guided Online Distillation: Promoting Safe Reinforcement Learning by Offline Demonstration [75.51109230296568]
  We argue that extracting expert policy from offline data to guide online exploration is a promising solution to mitigate the conserveness issue. We propose Guided Online Distillation (GOLD), an offline-to-online safe RL framework. GOLD distills an offline DT policy into a lightweight policy network through guided online safe RL training, which outperforms both the offline DT policy and online safe RL algorithms.
  arXiv  Detail & Related papers  (2023-09-18T00:22:59Z)
• Robust Driving Policy Learning with Guided Meta Reinforcement Learning [49.860391298275616]
  We introduce an efficient method to train diverse driving policies for social vehicles as a single meta-policy. By randomizing the interaction-based reward functions of social vehicles, we can generate diverse objectives and efficiently train the meta-policy. We propose a training strategy to enhance the robustness of the ego vehicle's driving policy using the environment where social vehicles are controlled by the learned meta-policy.
  arXiv  Detail & Related papers  (2023-07-19T17:42:36Z)
• Safety Correction from Baseline: Towards the Risk-aware Policy in Robotics via Dual-agent Reinforcement Learning [64.11013095004786]
  We propose a dual-agent safe reinforcement learning strategy consisting of a baseline and a safe agent. Such a decoupled framework enables high flexibility, data efficiency and risk-awareness for RL-based control. The proposed method outperforms the state-of-the-art safe RL algorithms on difficult robot locomotion and manipulation tasks.
  arXiv  Detail & Related papers  (2022-12-14T03:11:25Z)
• Minimizing Safety Interference for Safe and Comfortable Automated Driving with Distributional Reinforcement Learning [3.923354711049903]
  We propose a distributional reinforcement learning framework to learn adaptive policies that can tune their level of conservativity at run-time based on the desired comfort and utility. We show that our algorithm learns policies that can still drive reliable when the perception noise is two times higher than the training configuration for automated merging and crossing at occluded intersections.
  arXiv  Detail & Related papers  (2021-07-15T13:36:55Z)

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.