arXiv cs.MA· Sajad Marvi, Julian Schmidt, Alessandro Canevaro, Julian Jordan, Abhinav Valada·· 3 天前AI 评分26
不确定性驱动的自博弈如何提升多智能体驾驶策略的鲁棒性
Uncertainty-Driven Self-Play for Robust Multi-Agent Driving Policies
AI 导读
研究提出一种不确定性驱动的自博弈方案：一半智能体因提升邻居的认知不确定性而获得奖励，另一半则学习安全驾驶，探索智能体借此自动挖掘困难场景，无需人工设计课程。所得策略零样本迁移至 nuplan Val14 达 99.12%（标准自博弈为 97.97%），Argoverse 2 达 82.39%（对比 78.03%），并在 Waymo 上优于标准自博弈。
正文 · 原文
Abstract:Learning robust driving policies that generalize beyond their training domain remains a central challenge in motion planning. We propose an uncertainty-driven self-play scheme, in which half of the agents are rewarded for increasing their neighbors' epistemic uncertainty, while the other half learn to drive safely as the exploring agents continuously seek out their weaknesses. Epistemic uncertainty indicates where the policy is least confident and serves as an automatic proxy for scenario difficulty. Exploring agents use this signal to seek out hard cases, generating a challenging curriculum without manual scenario design. The resulting policy outperforms standard self-play, achieving best-in-class zero-shot transfer to nuplan Val14 (99.12% vs. 97.97%) and Argoverse 2 (82.39% vs. 78.03%) without fine-tuning, while also improving performance over standard self-play on Waymo. Our diversity-driven approach achieves 85% higher LSTM hidden-state coverage than the baseline, indicating a much richer set of behaviors. Further analysis shows that epistemic uncertainty serves as a calibrated difficulty detector, with highest-uncertain agents exhibiting 14.8 times the combined failure rate across collision, off-road, and DNF, of lowest-uncertain agents, enabling automatic prioritization of challenging scenarios without ground-truth labels.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.19714 [cs.RO] |
| (or arXiv:2610.19714v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19714 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Mohammad Sajad Marvi [view email]
[v1]
Wed, 14 Oct 2026 07:40:27 UTC (3,005 KB)
来源：arXiv cs.MA · arxiv.org