{"paper":{"title":"R2PS: Worst-Case Robust Real-Time Pursuit Strategies under Partial Observability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Belief preservation extends dynamic programming to partial observability for real-time robust pursuit policies.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dongbin Zhao, Runyu Lu, Ruochuan Shi, Yuanheng Zhu","submitted_at":"2025-11-21T16:34:00Z","abstract_excerpt":"Computing worst-case robust strategies in pursuit-evasion games (PEGs) is time-consuming, especially when real-world factors like partial observability are considered. While important for general security purposes, real-time applicable pursuit strategies for graph-based PEGs are currently missing when the pursuers only have imperfect information about the evader's position. Although state-of-the-art reinforcement learning (RL) methods like Equilibrium Policy Generalization (EPG) and Grasper provide guidelines for learning graph neural network (GNN) policies robust to different game dynamics, t"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"After reinforcement learning, our policy achieves robust zero-shot generalization to unseen real-world graph structures and consistently outperforms the policy directly trained on the test graphs by the existing game RL approach.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The belief preservation mechanism successfully extends the optimality properties of the dynamic programming strategies to the partially observable setting while preserving worst-case robustness against asynchronous evader moves.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"R2PS combines a proof that dynamic programming remains optimal under asynchronous evader moves, a belief preservation mechanism for partial observability, and integration into equilibrium policy generalization to produce real-time pursuer policies that zero-shot generalize to unseen graphs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Belief preservation extends dynamic programming to partial observability for real-time robust pursuit policies.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f14c27d8e7c62e7a97bae9327972bfc618e7ec7a4acf162cda7f10526adf01c9"},"source":{"id":"2511.17367","kind":"arxiv","version":2},"verdict":{"id":"842f93f8-75ea-40ad-929b-16bf2f5f81b4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T20:15:50.816748Z","strongest_claim":"After reinforcement learning, our policy achieves robust zero-shot generalization to unseen real-world graph structures and consistently outperforms the policy directly trained on the test graphs by the existing game RL approach.","one_line_summary":"R2PS combines a proof that dynamic programming remains optimal under asynchronous evader moves, a belief preservation mechanism for partial observability, and integration into equilibrium policy generalization to produce real-time pursuer policies that zero-shot generalize to unseen graphs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The belief preservation mechanism successfully extends the optimality properties of the dynamic programming strategies to the partially observable setting while preserving worst-case robustness against asynchronous evader moves.","pith_extraction_headline":"Belief preservation extends dynamic programming to partial observability for real-time robust pursuit policies."},"references":{"count":8,"sample":[{"doi":"","year":2019,"title":"Self-learning exploration and mapping for mobile robots via deep reinforcement learning","work_id":"ba50ce7d-0ee4-43ef-886d-f109c64c5d40","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1910,"title":"Soft actor-critic for discrete action settings.arXiv preprint arXiv:1910.07207,","work_id":"684554b0-ea6b-4199-b360-8e61b4bd9971","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Soft Actor-Critic Algorithms and Applications","work_id":"bb49c9fb-03b2-4226-9edb-50186b8193e4","ref_index":3,"cited_arxiv_id":"1812.05905","is_internal_anchor":true},{"doi":"","year":2001,"title":"Pursuit-evasion games with unmanned ground and aerial vehicles","work_id":"1d61f8c8-6e59-47ed-a30b-b2c5bebd63a2","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Solving urban network security games: Learning platform, benchmark, and challenge for AI research.arXiv preprint arXiv:2501.17559,","work_id":"bc133bbb-6108-4a06-bb6c-ac3662967a2b","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":8,"snapshot_sha256":"536094c12974b0cdb502a0af7eca5a8d25c404c9f71470c1ac7b89526790e82f","internal_anchors":1},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}