跳到正文
arXiv cs.AI· Xinyu Deng, Bohao Lan, Kunpeng Li, Mingkai Yuan, Weihang Pan, Binbin Lin·· 10 小时前AI 评分31

Risk-Triggered Search:面向冻结 JEPA 世界模型的行为保持安全控制

Risk-Triggered Search: Behavior-Preserving Safe Control

AI 导读

研究者提出 Risk-Triggered Search(RTS),在冻结的 JEPA 世界模型上展开动作条件未来潜在前缀,并用过程违规监督训练轻量风险头,同一前缀风险判据同时充当搜索触发门与硬可行性门。

正文 · 原文

View PDF HTML (experimental)

Abstract:Pre-trained visual world models generate multi-step action plans from current and goal observations, yet terminal-proximity objectives fail to prevent forbidden-zone traversal or rigid-body collision sliding. Purely visual interfaces may lack analytical dynamics, structured states, or explicit collision geometry, requiring internal prediction of action-conditioned process risk. We propose Risk-Triggered Search (RTS) for frozen Joint-Embedding Predictive Architecture (JEPA) world models. Rather than relying on static state features that cannot distinguish candidate plans sharing an observation, RTS unrolls action-conditioned future latent prefixes in the frozen world model and trains a lightweight risk head with process-violation supervision. To intervene selectively while preserving safe nominal behavior, RTS uses the same predicted prefix-risk criterion as both the search trigger gate and hard feasibility gate: safe nominal plans execute directly, whereas alarms trigger local Cross-Entropy Method (CEM) search to minimize nominal deviation within the feasible set. To prevent isotropic Euclidean minimal modification from shrinking obstacle-directed actions into zero-amplitude stalling, we introduce an amplitude-preserving metric $d_{\mathrm{AP}}$ that retains tangential progression while maintaining a unique zero at the nominal plan. Under unified gating and this unique-zero condition, bypassing safe-state search yields closed-loop trajectories strictly identical to every-step same-kernel search. Evaluated under a process-safety benchmark and Safe Arrival Rate (SAR) metric across five tasks, RTS improves SAR over Direct by 12.67--41.67 percentage points on navigation and articulated control and +2.08/+4.17 pp on Cube/PushT boundary cohorts, maintaining 100% action identity with every-step search while cutting controller time by 17.71%--70.41% across 1,968 paired trajectories.
Subjects: Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.24409 [cs.AI]
  (or arXiv:2611.24409v1 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2611.24409

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Xinyu Deng [view email]
[v1] Wed, 18 Nov 2026 13:37:59 UTC (1,213 KB)

来源:arXiv cs.AI · arxiv.org