Back to AI News
arxiv

Constrained World-Model Planning for Budget-Aware Visual Inspection

The paper formulates visual inspection with frozen detectors as neuro-symbolic observation planning under partial observability, separating a compact recurrent world model from symbolic feasibility and cost rules. A receding-horizon beam search with online replanning schedules class queries, crops, enhancements and stopping, supported by slack-aware shadow pricing and legal fallback at larger budgets. Across eight terrestrial, optical and sonar benchmarks, multi-step lookahead outperformed static schedules and search-free offline RL under tight budgets, gaining up to 28.4 AP percentage points on RUOD and 15.1 on ARIS with 11 ms planning cost. The approach aims to turn frozen detectors into adaptive, budget-aware inspection agents without retraining.

AI Brief

It presents a concrete planning method that improves budget-constrained visual inspection using frozen detectors, relevant to teams deploying vision systems under compute and sensing limits.

arXiv:2611.24187v1 Announce Type: cross Abstract: Visual inspection in field robotics and environmental monitoring must localize multiple targets under strict compute and sensing caps. Exhaustive sweeps waste calls on absent classes, while myopic and search-free policies stumble over heterogeneous tariff cliffs or discard viewport edits whose value emerges only after re-querying. We formulate inspection over frozen detectors as neuro-symbolic observation planning under partial observability, decoupling a compact recurrent world model (WM) from exact symbolic feasibility and tariff rules. Receding-horizon beam search with online replanning jointly schedules class queries, crops, enhancements, and termination (stop), paired with slack-aware shadow pricing and legal fallback at larger budgets. Across eight terrestrial, optical, and sonar benchmarks, multi-step lookahead outperforms static schedules and search-free offline RL under tight budgets -- gaining up to 28.4 AP percentage points over static priority on RUOD and 15.1 on ARIS -- while dynamic pricing lifts large-budget performance on all eight domains at 11 ms planning cost. Ultimately, pairing lightweight learned dynamics with exact symbolic constraints turns frozen detectors into adaptive, budget-aware inspection agents without retraining.
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!