{"id":"b758f5db-3465-473d-ac31-08832c049e91","arxiv_id":"2607.10649","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"A structured survey of 125 CPP works organizes recent methods into six problem-driven categories and connects them to classical foundations and open challenges.","lead":"This survey organizes 125 coverage path planning papers (mostly 2015–2026) into six categories and links them to classical methods. It gives robotics researchers a single map of single-robot, multi-robot, 3D, constrained, learning-based, and visual coverage work plus open problems.","discovery_kind":"review","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The strongest claim is that the survey comprehensively organizes recent CPP work into six problem-driven categories while connecting them to classical foundations. Full-text inspection confirms the taxonomy is consistently applied (single-robot offline/online, multi-robot offline/online, 3D layered/contour/decomposition, constrained energy/curvature/tethered/reconfigurable, learning-based, visual classical/hierarchical), tables summarize formulations and trade-offs, and Table I explicitly positions the work against prior surveys. The only soft spot is the non-reproducible inclusion protocol for the 125 works—exactly the reader’s weakest assumption. That is a limitation of transparency, not a load-bearing failure of the central organizational claim; surveys of this type routinely rely on expert selection. No stronger technical concern (inconsistency, missing core classical lineage, or unsupported experimental claim) appears. Therefore the reader’s ACCEPT / HIGH confidence verdict stands; no adjustment is warranted.","tokens_in":32254,"tokens_out":529,"duration_ms":5816,"concrete_test":"Independently sample 20–30 CPP papers from IEEE Xplore / Google Scholar (2015–2026) using queries such as “coverage path planning multi-robot”, “3D coverage path planning UAV”, “energy-constrained coverage path planning”, and “visual coverage path planning / view planning”; check whether each falls cleanly into one of the six categories and whether any major algorithmic family is systematically missing from the survey’s tables and timelines. If the taxonomy absorbs the sample without large gaps, the representativeness claim holds.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper is a problem-driven literature survey whose central claim is organizational and navigational: a six-category taxonomy of 125 representative CPP works (primarily 2015–2026) that links recent advances to classical foundations and fills gaps left by earlier narrower reviews (Abstract; Table I; §I.C–D; §VIII). That claim is supported by the full text—explicit category definitions, method summaries with strengths/limitations, comparison tables (II–VII), and the Table I contrast with prior surveys. The reader’s weakest assumption (non-reproducible hand selection of the 125 works) is real but standard for robotics surveys and does not undermine the internal structure, the classical-to-recent linkage, or the stated open challenges. No derivation, experimental result, or formal guarantee is at stake; completeness rests on author judgment without creating an internal inconsistency or a broken argument.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"This manuscript surveys coverage path planning (CPP), organizing 125 representative works (primarily 2015–2026) into a problem-driven six-category taxonomy: single-robot, multi-robot, 3D, constrained, learning-based, and visual CPP, with offline/online splits where appropriate. It links recent methods to classical foundations (cellular decomposition, STC, Morse, Boustrophedon, etc.), summarizes formulations, representative algorithms, strengths, and limitations for each family, and contrasts prior surveys in Table I. Comparison tables (II–VII), a timeline of online single-robot methods (Fig. 4), and a concluding discussion of open challenges (scalable online planning, multi-robot coordination, 3D/visual coverage, platform-constrained coverage, learning-enhanced hybrids) support the claim of a unified, up-to-date overview of the field.","tokens_in":32485,"tokens_out":1042,"duration_ms":14742,"significance":"If accepted as a structured map of the literature, the paper fills a clear gap: earlier surveys are either classical/2D-focused ([19], [20]), application-narrow ([21]–[23]), or incomplete on multi-robot, 3D, constrained, learning-based, and visual CPP (Table I). The problem-driven taxonomy, explicit strengths/limitations per family, and classical-to-recent linkage are useful for both newcomers and specialists. The comparison tables and open-challenge section are concrete contributions typical of high-value robotics surveys. No machine-checked proofs or new algorithms are claimed; the value is organizational and navigational, and that value is delivered.","major_comments":[{"comment":"Abstract and §I.C claim a survey of “125 representative works” with a six-category taxonomy that “fully” covers the field (Table I). The manuscript does not state a reproducible inclusion protocol (search queries, venues, years, exclusion rules, or how borderline works were assigned). For a survey whose central claim is completeness and balance, a short Methods-style paragraph (even if selection remains expert judgment) would make the corpus claim auditable and reduce the risk that underrepresented subareas (e.g., marine multi-robot, industrial spray painting beyond PaintNet) appear systematically omitted. This is standard survey hygiene and does not require redoing the taxonomy.","section":null},{"comment":"§VI (learning-based CPP) and §VII (visual CPP) are thinner and less systematically tabulated than §§II–V. Table VII lists learning-enhanced components, but the text underplays failure modes that matter for the survey’s own future-work claim (completeness, safety, sim-to-real, and when learned policies should be subordinated to classical completeness mechanisms). A short subsection or expanded “Strengths and Limitations” that ties learning/visual methods back to the completeness and online-hole issues developed in §II.B would better support the hybrid-framework recommendation in §VIII.","section":null}],"minor_comments":[{"comment":"Fig. 2 taxonomy is dense; some leaf labels (e.g., “2). Rank-based methods” under offline single-robot) are hard to parse at a glance. A cleaner hierarchical layout or color coding by offline/online would help.","section":null},{"comment":"Table I uses “Full / Partial / Limited / None” without a one-line operational definition in the caption beyond the footnote; moving that definition into the caption would improve standalone readability.","section":null},{"comment":"Notation for ε* / ε*+ / C* is consistent in the text but appears with slight typographic variation (epsilon vs. ε, asterisk placement). Standardize in the camera-ready version.","section":null},{"comment":"Several 2025–2026 citations (including author-affiliated CAP, Multi-CAP, C*) are appropriate as exemplars but should be clearly marked as recent/preprint where applicable so readers can judge maturity.","section":null},{"comment":"§VIII future-work bullets are strong; a brief prioritization (e.g., which open problem is most blocking for field deployment) would make the section more actionable without lengthening it much.","section":null},{"comment":"Minor copy-edits: “UA Vs” spacing, occasional missing spaces before citations, and “H ¨offmann” / accent consistency in the bibliography.","section":null}],"recommendation":"minor_revision","confidential_remarks":"Fit for a robotics journal survey track is good. Author self-citation of ε*, C*, CT-CPP, CAP, Multi-CAP is visible but normal for active contributors surveying their own line; it does not appear to distort the taxonomy. No integrity red flags. I would not block on the missing inclusion protocol if the authors add a short transparent paragraph; that is the main polish item."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is a clean, problem-driven survey of coverage path planning. The real value is the six-category taxonomy (single-robot, multi-robot, 3D, constrained, learning-based, visual) plus the explicit bridge from pre-2015 classical methods to ~125 works mostly from 2015–2026. Table I is honest about what earlier surveys covered and what they left out; that alone makes the paper useful for anyone who still starts from Choset or Galceran & Carreras.\n\nWhat it does well: each family gets formulations, representative algorithms, strengths/limitations, and comparison tables (II–VII). Online single-robot evolution is laid out with a timeline; multi-robot and constrained sections separate offline/online and constraint types without collapsing everything into one bag. Open challenges in §VIII are concrete (scalable online non-local methods, heterogeneous teams, online 3D/visual with incomplete maps, hybrid learning + classical guarantees). Self-citations of the authors’ own planners (ε*, C*, CT-CPP, CAP, Multi-CAP) sit inside the categories they survey; that is normal for active contributors and does not warp the structure.\n\nSoft spots are real but standard for this genre. The 125-paper set is hand-selected with no search protocol, venue list, or exclusion criteria, so completeness rests on author judgment. Learning-based and visual CPP get less depth than single- and multi-robot sections. There is no new algorithm, theorem, or empirical bake-off—pure synthesis. None of that breaks the argument; it just means you should treat the corpus as curated, not exhaustive.\n\nWho it is for: anyone entering or re-entering CPP, especially people working multi-robot, energy/curvature/tether constraints, or visual inspection who need a map of the last decade. Math and citation pattern look solid for a survey; no load-bearing derivation to fail. I would send it to peer review and would cite the taxonomy and open-problem list myself. Engage with it.","headline":"Solid, usable CPP survey that actually unifies classical foundations with multi-robot, 3D, constrained, learning, and visual work—worth keeping on the shelf.","tokens_in":33108,"tokens_out":517,"would_cite":true,"duration_ms":7856,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A problem-driven survey of 125 coverage-path-planning works maps classical foundations onto six modern categories and states the field’s open challenges.","keywords":["coverage path planning","motion and path planning","unknown environments","autonomous robots","multi-robot systems","3D coverage","learning-based planning","visual coverage"],"falsifier":"A documented, large body of peer-reviewed CPP methods from 2015–2026 that either cannot be placed in the six categories without severe distortion or is systematically omitted relative to included work of similar impact, which would show the taxonomy or the 125-work sample is not representative.","tokens_in":33139,"feed_emoji":"🤖","tokens_out":694,"duration_ms":12479,"temperature":0.7,"pith_summary":"Coverage path planning asks a robot to traverse every part of a workspace while cutting path length, overlap, turns, and energy. Classical work largely treated a single robot on a known 2D map; newer systems face multi-robot teams, 3D surfaces, battery and curvature limits, learned policies, and camera-based inspection. This paper surveys 125 representative works, mostly from 2015–2026, and organizes them into six categories so that formulations, algorithms, strengths, and limits can be compared under the same problem factors: map knowledge, geometry, robot constraints, sensing goals, and coordination. The authors argue that subarea decomposition, non-local online guidance, and hybrid learning-plus-planning are the main recent trends, and they list concrete open problems in scalable online planning, heterogeneous multi-robot teams, online 3D/visual coverage, and resource-aware platforms. A sympathetic reader gets a single map of where the field has moved and which gaps still block deployable systems.","feed_headline":"Six categories organize 125 robot coverage advances","feed_subtitle":"A survey links classical methods to multi-robot, 3D, constrained, learning, and visual tasks","key_machinery":"The six-category problem-driven taxonomy (single-robot, multi-robot, 3D, constrained, learning-based, visual CPP), with each method further typed offline vs online and scored by how map knowledge, workspace geometry, robot constraints, sensing objectives, and coordination shape the formulation.","core_discovery":"The paper claims that recent coverage path planning is best understood through a problem-driven six-category taxonomy—single-robot, multi-robot, 3D, constrained, learning-based, and visual CPP—built from 125 representative works and explicitly linked to classical pre-2015 methods, and that this organization both summarizes current practice and exposes open challenges in online scalability, multi-robot coordination, 3D and visual coverage, platform constraints, and learning-enhanced planning.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Six categories organize 125 robot coverage path planning advances","Survey maps 125 CPP works from classical to multi-robot and visual","Taxonomy of coverage planning: six categories span 125 papers","Problem-driven survey links pre-2015 methods to modern CPP","Six-category review of robot coverage from single to learning-based"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"That the authors’ hand-picked set of 125 works and the six-category split fairly represent the whole field, without a published search protocol that would let someone check what was left out.","fun_headline_variants_meta":{"raw":{"variants":["Six categories organize 125 robot coverage path planning advances","Survey maps 125 CPP works from classical to multi-robot and visual","Taxonomy of coverage planning: six categories span 125 papers","Problem-driven survey links pre-2015 methods to modern CPP","Six-category review of robot coverage from single to learning-based"]},"model":"grok-4.5","effort":"low","cost_usd":0.004206,"raw_usage":{"total_tokens":1295,"prompt_tokens":843,"num_sources_used":0,"completion_tokens":71,"cost_in_usd_ticks":42060000,"prompt_tokens_details":{"text_tokens":843,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":381,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":843,"tokens_out":71,"duration_ms":4419,"temperature":1.0,"reasoning_tokens":381,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-14T10:13:04.196263+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"A documented, large body of peer-reviewed CPP methods from 2015–2026 that either cannot be placed in the six categories without severe distortion or is systematically omitted relative to included work of similar impact, which would show the taxonomy or the 125-work sample is not representative.","supporting_citations":[],"review_version":1}