{"id":"62bea482-2c1f-4d8f-aac0-ed39af44ba43","arxiv_id":"2411.15953","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":2.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A survey of multi-robot exploration methods paired with an unevaluated system proposal for 3D mapping and fire detection.","lead":"This preprint reviews existing 2D and 3D multi-robot exploration algorithms and sketches a modular pipeline using Cartographer, OctoMap, and potential fields. It provides no experiments or data, and the promised fire detection capability is never described.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"GPS-derived elliptical target points in Section III-B cannot yield interior exploration: the fixed ellipse is independent of the online OctoMap and has no mechanism to enter the building, so the central 'effective exploration' claim is unsupported.","rationale":"I read the paper in good faith as a survey-plus-system-description. The literature review is reasonable, and the system diagram is coherent. However, the central claim in the abstract—that the modular approach 'facilitate[s] effective exploration'—depends on the trajectory generator producing useful exploration targets. The described generator uses GPS-derived ellipse waypoints around the building center and deliberately avoids replanning. This is the weakest link because, unlike frontier-based or next-best-view methods reviewed in the paper, the target points are not informed by the online map and do not address entering a structure. The reader correctly identified this as the weakest assumption; I agree. My concrete test would settle whether the concern lands: if the ellipse trajectory cannot observe a meaningful fraction of interior voxels, the system cannot perform 'building exploration and fire detection' as claimed. This is an internal correctness risk, not merely disagreement with consensus. No formal verification, code, or quantitative evaluation exists to offset it. Hence no change to the REJECT verdict.","tokens_in":6781,"tokens_out":4622,"duration_ms":43911,"concrete_test":"Analytical coverage check: model a rectangular building of footprint 20m x 10m with wall height 3m and a 1m door on one facade. Place a UAV on the Section III-B ellipse generated from the building-center GPS coordinates (e.g., semi-axes 25m x 15m at 5m altitude) with a 30m-range, 90-degree FOV depth sensor. Compute the set of observed interior voxels over one full ellipse traversal using ray casting. If fewer than 5% of interior free-space voxels are observed, the GPS-ellipse target assumption is refuted. A complementary Gazebo experiment using the described Cartographer/OctoMap pipeline would confirm under realistic noise.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section III-B states that 'target points for trajectory generation are currently obtained from the Global Positioning System (GPS) location of the building' and shows ellipse-shaped waypoints around the building center (Fig. 5). The text also states that 'trajectory planning does not involve replanning during execution,' so the global path is fixed before mapping and is not updated from the OctoMap. For a closed or partially enclosed building, an exterior elliptical trajectory observes only outer walls and roof; interior voxels remain unknown. The only adaptation is a local potential-field obstacle-avoidance correction of the current trajectory, which cannot insert new goal points inside the structure. Therefore the load-bearing premise that this pipeline 'facilitate[s] effective exploration' of 3D building environments fails unless the robot has a separate, undescribed mechanism for entering openings and selecting interior targets. No experimental results, coverage metrics, or code are provided to show otherwise; Figure 6 only shows a Gazebo/RViz snapshot. The conclusion further limits the contribution to a 'single robot,' whereas the title and abstract promise multi-robot exploration and fire detection. Consequently, the central claim rests on an untested and, as described, geometrically insufficient target-generation rule.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript surveys 2D and 3D exploration strategies for multi-robot systems and then describes a modular 3D exploration framework built around Google Cartographer, an OctoMap representation, GPS-derived elliptical target points, and a potential-field obstacle avoidance module. The stated goal is effective autonomous exploration of building environments with eventual application to fire detection. No experiments, quantitative metrics, or comparisons to existing methods are reported; the only supporting visual is a Gazebo/RViz snapshot in Fig. 6.","tokens_in":6955,"tokens_out":3098,"duration_ms":32625,"significance":"If the described framework were demonstrated to explore 3D building interiors effectively, it would be relevant to search-and-rescue and inspection applications. The survey portion is organized and cites a reasonable set of prior works, and the use of OctoMap plus potential-field avoidance is a plausible starting point. However, the paper does not establish that the proposed system works: there is no experimental evidence, no coverage or mapping quality metrics, and the target-point generation rule appears geometrically unable to produce interior exploration. The title and abstract promise multi-robot operation and fire detection, but the described system is single-robot and fire detection is not addressed beyond the title. As submitted, the contribution is a broad overview plus an underspecified, untested architecture, so its significance as a research paper is limited.","major_comments":[{"comment":"The target-point generation rule cannot support the central claim of effective interior exploration. The text states that target points are obtained from the GPS location of the building and form an ellipse around the building center, and that trajectory planning does not involve replanning during execution. An elliptical exterior trajectory observes only outer walls and roof; it has no mechanism for generating goal points inside the building or even at detected openings. The potential-field obstacle avoidance only locally corrects the current path and cannot insert new interior goals. Therefore the pipeline, as described, cannot explore a closed or partially enclosed building's interior, and the abstract's claim that the framework facilitates effective exploration is unsupported.","section":"Section III-B, Fig. 5"},{"comment":"The paper contains no experiments, no exploration-time measurements, no coverage ratio, no map-accuracy evaluation, and no comparison against baseline exploration algorithms. Figure 6 is a single Gazebo/RViz snapshot and does not demonstrate that the environment was explored or that the map is complete. Without quantitative evidence, the phrases 'effective exploration' and 'autonomous exploration of 3D environments using a single robot' in Section IV are not substantiated.","section":"Section III (overall), Fig. 6"},{"comment":"There is a scope mismatch between the stated contribution and the actual content. The title promises 'Multi-Robot Exploration Strategies' with 'Fire Detection Capabilities,' and the abstract repeats this framing, but Section IV explicitly limits the described contribution to exploration using a single robot and lists multi-robot coordination and map sharing as future work. Fire detection is not described in any part of the system. This mismatch misrepresents the paper's contribution and should be corrected if the paper is revised.","section":"Section IV and title/abstract"},{"comment":"The relationship between the online OctoMap and trajectory generation is inconsistent. The paper says the resulting OctoMap is used for trajectory planning and execution, but it also says the target points come from GPS and that replanning does not occur during execution. Since the trajectory is fixed before execution and only locally modified by potential fields, the online map appears not to influence global exploration decisions. The manuscript should clarify how the OctoMap actually guides the generation of exploration goals, or remove the implication that it does.","section":"Section III-B"}],"minor_comments":[{"comment":"The manuscript switches between 'we' and 'I' (e.g., 'We explore' in the abstract, then 'I propose' and 'our work' later). The voice should be made consistent.","section":"Abstract and throughout"},{"comment":"The formatting in the text is broken in places, such as 'Theparameter' and 'thecell.Theparameter'. The equation itself is a standard cost-utility formulation and is adequately attributed to prior work, but the surrounding text should be cleaned up.","section":"Section II-B, Eq. (1)"},{"comment":"Several bibliographic entries are incomplete, notably Refs. [6], [10], and [12], which lack full publication data. Please complete all references according to the journal style.","section":"References"},{"comment":"Figures 3, 4, and 6 are referenced in the text but not numbered consistently with their captions, and Figure 3 appears before the paragraph that introduces it. Please recheck figure placement and numbering.","section":"Figures"}],"recommendation":"reject","confidential_remarks":"This manuscript reads as an early-stage write-up rather than a complete archival paper. The survey component is reasonable but the original contribution is both underspecified and untested, and the described trajectory generation mechanism appears fundamentally incapable of interior exploration. The scope mismatch with the title further weakens the submission. I see no basis for acceptance or even major revision without a substantial redesign and a full experimental evaluation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Plain take: this is a decent literature overview of 2D/3D multi-robot exploration, but the paper's substantive claim—a working modular 3D exploration system with fire detection—is not supported by the text. I agree with the reader's REJECT verdict and with the stress-test note.\n\nWhat the paper does well: the survey sections are competent. It covers frontier-based, cost-utility, and next-best-view methods, cites standard sources (Yamauchi, Burgard, Bircher, etc.), and explains OctoMap and potential fields accurately. For someone entering the field, it works as a compact reading list.\n\nThe soft spots are not minor. The system description in Section III-B uses GPS-derived elliptical waypoints around the building center, with no replanning during execution. A fixed exterior ellipse cannot generate goals inside a building; it only observes outer walls and roof. Local potential-field avoidance can correct the current path but cannot insert new interior goals. So the claim that the pipeline 'facilitate[s] effective exploration' of 3D building environments is geometrically unsupported. The paper itself admits in the conclusion that the approach is limited to a single robot, undercutting the multi-robot title. The promised fire-detection capability never appears in the body. There are no experimental results, no coverage metrics, no baseline comparisons, and no code or data. The only evidence is a Gazebo/RViz screenshot.\n\nThe stress-test is correct: this is a load-bearing flaw. If the authors have a working system, they need to show it exploring an interior, with replanning or a target-selection mechanism that responds to the online OctoMap. Right now, the system diagram is a standard Cartographer + OctoMap + potential-fields stack with an inadequate target-generation rule.\n\nWho is this for? A reader wanting a quick survey of exploration strategies might skim it. As a research contribution, it does not deserve a full peer review; it would be a desk reject or, at best, a workshop paper if reframed as a survey. I would not cite it, and I would not bring it to a reading group except to illustrate how a promising system description can fail on its own stated assumptions.","headline":"A reasonable survey of multi-robot exploration, but the claimed modular system with fire detection is unsupported: the GPS-ellipse target rule cannot explore building interiors, and there are no experiments.","tokens_in":7476,"tokens_out":1969,"would_cite":false,"duration_ms":19912,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper argues that a modular pipeline combining online OctoMap mapping, GPS-derived ellipse targets, and potential-field obstacle avoidance can let a single UAV explore an unknown 3D building for fire detection.","keywords":["multi-robot exploration","3D mapping","OctoMap","frontier-based exploration","potential fields","obstacle avoidance","fire detection","UAV exploration"],"falsifier":"Fly the described system in a building with no GPS signal or with a non-convex layout, and check whether the ellipse-derived trajectories reach and map the interior rooms. If the robot repeatedly hits walls, flies outside the building, or leaves large interior spaces unmapped, then the central exploration claim fails.","tokens_in":6560,"feed_emoji":"🚁","tokens_out":8317,"duration_ms":70726,"temperature":0.7,"pith_summary":"This paper argues that 3D building exploration for fire detection can be assembled from modular pieces instead of a single monolithic planner: localize and map online, turn the point-cloud map into an OctoMap, compute a trajectory, and correct that trajectory locally with potential-field obstacle avoidance. It reviews 2D frontier-based and cost-utility exploration methods and explains why these do not simply transfer to 3D, where memory and computation become the limiting resources. The proposed system is meant to work without prior knowledge of the building, using only a GPS fix on the building center to draw elliptical target points for the trajectory planner. The larger goal is a foundation for decentralized multi-UAV exploration with shared maps and coordinated task allocation.","feed_headline":"UAV pipeline explores unknown 3D buildings, dodges obstacles","feed_subtitle":"A live 3D map, GPS-drawn flight targets, and repulsive obstacle fields let a UAV explore a building without a prior map.","key_machinery":"The load-bearing mechanism is the OctoMap: an octree of voxels, each holding a probability of being occupied, built from point clouds and used as the single map representation for trajectory planning. Around it sits a four-stage pipeline—local mapping from sensor data, OctoMap generation, trajectory planning to GPS-ellipse target points, and potential-field obstacle avoidance for local correction. For the 2D methods it reviews, the paper frames exploration through the cost-utility score $\\mathrm{BCU}(a) = U(a) - \\lambda_{CU} C(a)$, trading expected information gain against travel cost; in the 3D pipeline this trade-off is replaced by fixed geometric targets, with safety handled separately by the potential field.","core_discovery":"On its own terms, the paper's central claim is that the 3D exploration problem is best handled by separating map representation from planning: a SLAM system produces point clouds, an OctoMap stores those points as probabilistic free/occupied voxels in a memory-efficient octree, and a trajectory planner steers the robot toward GPS-derived ellipse targets while a potential-field layer locally deforms the trajectory around obstacles detected by laser data. The paper positions this as a response to traditional map-dependent algorithms failing in changing environments, because the map is built online rather than assumed in advance. It also claims that the same architecture is the natural base for multi-robot and multi-UAV extensions through decentralized map merging and coordinated exploration.","pith_inferences":["A natural next test the paper does not run is to replace the GPS-ellipse targets with frontier or next-best-view targets generated from the OctoMap itself; that would remove the dependency on GPS and likely improve interior coverage in cluttered buildings.","The fire-detection capability is asserted but never measured; attaching a thermal camera and reporting detection latency and coverage during the same exploration would turn the claim into a benchmarkable one.","Because trajectory planning does not replan online, the potential-field layer is the only safety net; one could test how often local corrections are needed and whether they suffice in dense obstacle fields, which the paper leaves open.","The modular framing suggests the survey contribution may outlive the specific implementation: the same OctoMap-plus-obstacle-avoidance architecture could host any target generator, so a planner swap is an obvious extension."],"forward_implications":["A robot can start exploring a building with no prior map: the map is built online from point-cloud data, so stale or unavailable maps do not invalidate the system.","The potential-field obstacle-avoidance layer lets the robot react to obstacles the planner did not know about, because trajectory execution is locally validated against live laser data.","The OctoMap keeps memory use low enough for onboard 3D planning, since only free and occupied voxels are stored in a tree rather than a dense volumetric grid.","The same modular design is a plausible base for a multi-UAV team, with decentralized map creation, map sharing, and coordinated task allocation as the paper's stated next steps.","For 2D teams, cost-utility frontier allocation remains the reference method, but the review's comparison implies 3D systems should be judged by memory and computation as much as by exploration time."],"supporting_citations":[{"why":"Defines the OctoMap octree voxel representation used by the implemented 3D pipeline.","marker":"[16]"},{"why":"Introduces frontier-based exploration, the concept that anchors the reviewed 2D strategies.","marker":"[10]"},{"why":"Supplies the cost-utility formulation and Hungarian-method task allocation for coordinated multi-robot exploration.","marker":"[13]"},{"why":"Presents the target-point-based MAV 3D exploration method the paper's GPS-ellipse trajectory planning builds on.","marker":"[24]"},{"why":"Provides the information-potential-field 3D exploration strategy whose UAV/OctoMap trajectory illustration the paper uses.","marker":"[17]"},{"why":"Defines autonomous exploration and the centralized-versus-decentralized comparison used throughout the survey.","marker":"[5]"},{"why":"Offers the receding-horizon next-best-view planner that motivates online tree-based 3D path planning.","marker":"[23]"}],"fun_headline_variants":["Multi-robot 3D mapping adapts to changing buildings","3D exploration without predefined maps for robot teams","Robot swarms build live 3D maps for fire detection","OctoMap-based planning keeps UAVs safe in unknown spaces","SLAM + potential fields: autonomous 3D building scouts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the GPS coordinates of the building center are available and accurate enough that an ellipse drawn around them provides target points that lead the robot through the parts of the building that need exploring; the paper does not test this in GPS-denied or cluttered environments.","fun_headline_variants_meta":{"raw":{"variants":["Multi-robot 3D mapping adapts to changing buildings","3D exploration without predefined maps for robot teams","Robot swarms build live 3D maps for fire detection","OctoMap-based planning keeps UAVs safe in unknown spaces","SLAM + potential fields: autonomous 3D building scouts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001124,"raw_usage":{"total_tokens":4617,"prompt_tokens":832,"completion_tokens":3785,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":448,"completion_tokens_details":{"reasoning_tokens":3702}},"tokens_in":448,"tokens_out":3785,"duration_ms":23272,"temperature":1.0,"reasoning_tokens":3702,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:41:05.481800+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fly the described system in a building with no GPS signal or with a non-convex layout, and check whether the ellipse-derived trajectories reach and map the interior rooms. If the robot repeatedly hits walls, flies outside the building, or leaves large interior spaces unmapped, then the central exploration claim fails.","supporting_citations":[{"cited_title":"Techniques for multi-robot coordination and navigation,","cited_arxiv_id":null,"evidence_quote":"Defines the OctoMap octree voxel representation used by the implemented 3D pipeline."},{"cited_title":"A frontier-based approach for autonomous exploration,","cited_arxiv_id":null,"evidence_quote":"Introduces frontier-based exploration, the concept that anchors the reviewed 2D strategies."},{"cited_title":"Coordinated multi-robot exploration,","cited_arxiv_id":null,"evidence_quote":"Supplies the cost-utility formulation and Hungarian-method task allocation for coordinated multi-robot exploration."},{"cited_title":"A target point based MAV 3d exploration method,","cited_arxiv_id":null,"evidence_quote":"Presents the target-point-based MAV 3D exploration method the paper's GPS-ellipse trajectory planning builds on."},{"cited_title":"Efficient autonomous robotic exploration with semantic road map in indoor environments,","cited_arxiv_id":null,"evidence_quote":"Provides the information-potential-field 3D exploration strategy whose UAV/OctoMap trajectory illustration the paper uses."},{"cited_title":"A comparison of path planning strategies for autonomous exploration and mapping of unknown environments,","cited_arxiv_id":null,"evidence_quote":"Defines autonomous exploration and the centralized-versus-decentralized comparison used throughout the survey."}],"review_version":1}