Back to AI News
arxiv

Risk-Triggered Search: Behavior-Preserving Safe Control

The paper proposes Risk-Triggered Search (RTS) for frozen Joint-Embedding Predictive Architecture (JEPA) world models, training a lightweight risk head on action-conditioned future latent prefixes to predict process violations. Safe nominal plans execute directly, while predicted risk triggers local Cross-Entropy Method search using an amplitude-preserving metric intended to avoid zero-amplitude stalling. Evaluated on five tasks with a Safe Arrival Rate metric, RTS improves SAR by 12.67–41.67 percentage points on navigation and articulated control and by 2.08/4.17 points on Cube/PushT boundary cohorts. It reports 100% action identity with every-step search while cutting controller time by 17.71%–70.41% across 1,968 paired trajectories.

AI Brief

It offers a concrete method for adding selective safety intervention to frozen world-model planners, with reported gains in safety and controller efficiency.

arXiv:2611.24409v1 Announce Type: new 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.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!