# Chang and Harabor propose Goal-Branching LaCAM for dense multi-agent pathfinding [Canonical reading](https://escapemeridian.com/readings/chang-and-harabor-propose-goal-branching-lacam-for-dense-multi-fb07518e) Authors Stephen Chang and Daniel Harabor propose Goal-Branching for multi-agent pathfinding (MAPF), where a solver searches for paths for multiple agents sharing a space. They say LaCAM can struggle in dense, corridor-heavy instances because resolving deadlocks may require several agents to move together, sometimes away from their goals. Their approach gives blocking agents temporary goals and tries each as an additional branch of LaCAM’s search; they report that completeness is preserved for any rule used to choose those goals. On dense instances across three maze and room maps, the authors report that Goal-Branching solves 89% of instances, compared with 44% for LaCAM\*. On instances both methods solve, they report first solutions 10–14 times faster on two named maps, and nearly 200 times faster in median time at 75% density on one map. They also report a trade-off: solution cost is 15–22% higher on instances both methods solve. On open maps and the full MovingAI benchmark, they report no higher solution cost than LaCAM\* and at most a 3% slowdown. These results describe the tested maps and benchmark; the abstract does not establish how the approach performs on other dense maps or in real-world deployments. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Nov 5, 2026 (2026-11-05). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-08T17:48:29.337Z. First observed: 2026-11-06T05:10:33.178Z. Retrieved: 2026-11-08T17:47:43.765Z. - [Goal-Branching LaCAM: Temporary Goals as Search Branches in Dense Multi-Agent Pathfinding](https://arxiv.org/abs/2611.07528) ## Limits The reported dense-instance results cover three maze and room maps; the abstract does not report performance on other dense maps. The source evidence provided is the abstract, so details of the experimental setup and results beyond its summary are unavailable. The abstract reports benchmark results, not real-world deployments. ## Open question How does Goal-Branching perform on dense maps beyond the three maze and room maps reported here? ## Source-stated Chang and Harabor propose adding temporary goals for blocking agents as additional branches in LaCAM’s search; they state that completeness is preserved for any rule for choosing those goals. > We examine random and corridor-aware rules for choosing these temporary goals and show that completeness is preserved for any such rule. - [Supporting source](https://arxiv.org/abs/2611.07528) ## Source-stated On dense instances across three maze and room maps, the authors report that Goal-Branching solves more instances than LaCAM\*. > On dense instances (60-90% of free cells occupied) of three maze and room maps, Goal-Branching solves 89% of instances against 44% for LaCAM\* - [Supporting source](https://arxiv.org/abs/2611.07528) ## Source-stated On instances both methods solve, the authors report faster first solutions on named maps and higher solution cost; the abstract also reports no higher solution cost on open maps and the full MovingAI benchmark, with at most a small slowdown. > Solution cost increases by 15-22% on instances both solve, while on open maps and the full MovingAI benchmark its solutions cost no more than those of LaCAM\* and it is at most 3% slower. - [Supporting source](https://arxiv.org/abs/2611.07528) ## Source and contributor context Chang, Stephen, Harabor, Daniel · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Benchmark A defined set of tests used to compare systems. Results depend on the tasks and measures selected. - [Definition reference](https://arxiv.org/abs/2211.09110) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## Editorial provenance Version 1. Publication ID: ceacea5f-c1a2-4fa3-8acd-441c254c85ae. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T17:48:29.337Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: b0567c6ca43a13d7e7d2b7c235af0bc8fa4f80e905e470f00b52212c04f4f63c. Source text hash: 37b945968830e26e36b6109f3325ce13aaf945aaaf854e57081cc7cdd685cb0f. ## Revision and source-check history ### Version 1 — 2026-11-08T17:48:29.337Z Authors Stephen Chang and Daniel Harabor propose Goal-Branching for multi-agent pathfinding (MAPF), where a solver searches for paths for multiple agents sharing a space. They say LaCAM can struggle in dense, corridor-heavy instances because resolving deadlocks may require several agents to move together, sometimes away from their goals. Their approach gives blocking agents temporary goals and tries each as an additional branch of LaCAM’s search; they report that completeness is preserved for any rule used to choose those goals. On dense instances across three maze and room maps, the authors report that Goal-Branching solves 89% of instances, compared with 44% for LaCAM\*. On instances both methods solve, they report first solutions 10–14 times faster on two named maps, and nearly 200 times faster in median time at 75% density on one map. They also report a trade-off: solution cost is 15–22% higher on instances both methods solve. On open maps and the full MovingAI benchmark, they report no higher solution cost than LaCAM\* and at most a 3% slowdown. These results describe the tested maps and benchmark; the abstract does not establish how the approach performs on other dense maps or in real-world deployments. ## Original source arXiv cs.AI · Article Original source date: Nov 5, 2026 (2026-11-05). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-08T17:48:29.337Z. First observed: 2026-11-06T05:10:33.178Z. Retrieved: 2026-11-08T17:47:43.765Z. - [Goal-Branching LaCAM: Temporary Goals as Search Branches in Dense Multi-Agent Pathfinding](https://arxiv.org/abs/2611.07528) ## Limits The reported dense-instance results cover three maze and room maps; the abstract does not report performance on other dense maps. The source evidence provided is the abstract, so details of the experimental setup and results beyond its summary are unavailable. The abstract reports benchmark results, not real-world deployments. ## Open question How does Goal-Branching perform on dense maps beyond the three maze and room maps reported here? ## Source-stated Chang and Harabor propose adding temporary goals for blocking agents as additional branches in LaCAM’s search; they state that completeness is preserved for any rule for choosing those goals. > We examine random and corridor-aware rules for choosing these temporary goals and show that completeness is preserved for any such rule. - [Supporting source](https://arxiv.org/abs/2611.07528) ## Source-stated On dense instances across three maze and room maps, the authors report that Goal-Branching solves more instances than LaCAM\*. > On dense instances (60-90% of free cells occupied) of three maze and room maps, Goal-Branching solves 89% of instances against 44% for LaCAM\* - [Supporting source](https://arxiv.org/abs/2611.07528) ## Source-stated On instances both methods solve, the authors report faster first solutions on named maps and higher solution cost; the abstract also reports no higher solution cost on open maps and the full MovingAI benchmark, with at most a small slowdown. > Solution cost increases by 15-22% on instances both solve, while on open maps and the full MovingAI benchmark its solutions cost no more than those of LaCAM\* and it is at most 3% slower. - [Supporting source](https://arxiv.org/abs/2611.07528) ## Source and contributor context Chang, Stephen, Harabor, Daniel · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Benchmark A defined set of tests used to compare systems. Results depend on the tasks and measures selected. - [Definition reference](https://arxiv.org/abs/2211.09110) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## Editorial provenance Version 1. Publication ID: ceacea5f-c1a2-4fa3-8acd-441c254c85ae. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T17:48:29.337Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: b0567c6ca43a13d7e7d2b7c235af0bc8fa4f80e905e470f00b52212c04f4f63c. Source text hash: 37b945968830e26e36b6109f3325ce13aaf945aaaf854e57081cc7cdd685cb0f. ### uncertain — 2026-11-08T17:48:29.337Z — anthropic/claude-opus-5.5 Publication: ceacea5f-c1a2-4fa3-8acd-441c254c85ae attribution — pass: The proposal, the completeness result and all benchmark figures are attributed to authors Stephen Chang and Daniel Harabor. Nothing is presented as independently verified. > Authors:Stephen Chang, Daniel Harabor summary — uncertain: The summary explains MAPF, but it never says LaCAM is an existing search-based MAPF solver that the paper modifies. It also does not distinguish LaCAM from LaCAM\*, the baseline used in comparisons. Smallest repair: add one clause that LaCAM is an existing search-based solver (a tooltip for LaCAM\* would also help). The terms completeness and solution cost need term-library tooltips. A minor softening also appears: the source says coordinated moves are needed and are often away from goals; the draft says they 'may' be needed and are 'sometimes' away from goals. The draft could restore 'often'. > LaCAM is a search-based solver for multi-agent pathfinding (MAPF) that runs a depth-first search over joint configurations date — pass: The originalPublished date of 2026-11-05 matches the arXiv v1 submission date. It does not use the later feed date. > \[Submitted on 5 Nov 2026\] support — pass: Each claim matches the abstract: the 89% versus 44% solve rates, the 10-14x and nearly 200x speedups, the 15-22% cost increase and the at most 3% slowdown. The limitations and the question correctly note scope (three maps, benchmark only, abstract only). The draft adds no invented benefits or deployments. > Goal-Branching solves 89% of instances against 44% for LaCAM\* media — pass: The selected media is the verified direct link to the arXiv abstract page for this paper. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/fb07518e012a0c4e33a4b994b606bf3d33bdaf629c8f6d61bb28651f8cad8829/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/chang-and-harabor-propose-goal-branching-lacam-for-dense-multi-fb07518e.json)