Related tools and engineering references¶
Documentation index · Project map
These projects illustrate useful research-software practices. A venue, star count or polished demo alone does not establish implementation correctness. The lessons below concern usability, contracts and evidence; DEC-MAPF does not import their implementations or claim to reproduce their results.
| Primary source | Useful pattern | Where it appears here |
|---|---|---|
| SMART project and demo · documentation | Discoverable examples with map dimensions/agent counts; detailed visualization controls | Gallery, GUI walkthrough and saved-run library. DEC-MAPF remains a discrete grid simulator; no physics integration is implied. |
| PRIMAL GRID · PRIMAL code | A direct path from explanation to an interactive example | Independent GUI/headless entry points, scenario editing and a small validated first run. No RL training or legacy dependencies are adopted. |
| POGEMA benchmark · environment | Explicit configuration, seeds, benchmark families and linked evaluation evidence | Versioned presets, immutable plans and separation of tutorial, current-run and historical evidence |
| PettingZoo environment tests | Conformance tests and versioned behavior beyond a single happy-path example | Cross-interface preset tests, lifecycle failure tests and concrete trajectory witnesses |
| PBS, EECBS, CBSH2-RTC | Explicit map/scenario/agent/cutoff commands; native-build checks; distinction between current and paper code | Solver identities, native profiles, platform CI and article boundaries |
| MovingAI MAPF benchmarks · MAPF definitions | Make collision rules, objectives and target behavior explicit | Problem-assumptions card, scientific semantics and independent validation |
| SILLM · implementation | Tell readers which code/configuration belongs to which experiment | Algorithm/configuration provenance and historical-result qualifications. Lifelong MAPF remains outside this modernization. |
| ITA-CBS2 and MAPF-LRR2023 | Runnable experiment recipes, solver flags and visualization guidance | Bounded examples, generated parameter reference and explicit resource policies; no new task-assignment track |
| AMLGym, MAcq, libMultiRobotPlanning | Clear method/adapter boundaries and testable interfaces | Architecture, extension contracts and small independent acceptance cases |
| MAPFAST | Visible correction notices tied to affected results | Explicit method identity, bounded scientific claims and preserved experiment evidence |
| League of Robot Runners and MAPF-Mechanism | Separate problem setup, evaluation and configuration from presentation | Science, metrics and reproducibility; their lifelong/mechanism objectives are not DEC-MAPF benchmark claims |
SMART's official documentation cites RA-L 2026. POGEMA's ICLR 2025 paper and PettingZoo's NeurIPS 2021 paper support their publication context. This is a curated engineering reference list, not a ranking of repositories or a recommendation to combine incompatible benchmark objectives.