DEC-MAPF researcher documentation¶

Use DEC-MAPF to run decentralized and centralized methods on shared scenarios, validate their trajectories and compare results. The Python API and CLI support the full experiment workflow; the optional GUI reads the same saved runs. Start with the workflow below that matches your task. Alpha status describes compatibility and scientific scope.

Start here¶

Your starting point	Follow this path	What you will have at the end
I want to see the system	Gallery → Installation → GUI walkthrough	A validated simulation, inspected decisions and a portable replay
I want to run an experiment	Installation → Headless tutorial → Reproducibility	A saved study specification, one row for every planned trial and matched comparisons
I want to compare decentralized and centralized methods	Eight-trial example → Solver guide → Metrics	Matched scenarios, explicit budgets and independently validated outcomes
I want to add a solver or coordination protocol	Extension guide → Public API → Architecture	A clear adapter, configuration and telemetry integration path
I have my own maps and agents	Scenarios → Parameters → Solvers	Validated inputs with recorded hashes and supported settings
I want to use Python or HTTP	Python and API → Contracts	A programmatic run using the same application services
I want to contribute	Contributing → Extending → Workspace architecture	A tested change with clear scientific and API consequences

Understand what the evidence means¶

Simulation and scientific interpretation: ticks, movement settings, negotiations, commitments, validation, costs and denominators.
Metric definitions: measured communication, unavailable values, detours and qualified comparison quantities.
Experiment reference: budget enforcement, attempt lineage, stopping reasons and source checks.
Reproducibility: what to freeze, what to report, and how article-derived profiles differ.
TAOP conformance and centralized conformance: exact implementation boundaries and comparison assumptions.
Solver identities, algorithm semantics and native baselines: the algorithm behind each name.
Citation provenance: stable publication metadata; citations are also in the README.

Reference and recovery¶

Use Troubleshooting for installation, missing replay layers, timeouts, source mismatches and pairing errors. The parameter reference is generated from the current request models. Live HTTP schemas are available at /docs and /openapi.json on your local server; the versioned OpenAPI file supports client generation.

The gallery provenance identifies the actual runs and image checksums. Documentation verification records the bounded checks performed for this guide. The examples index distinguishes runnable experiment specifications from direct Python API illustrations.

Maintainer and method references¶

Use the project map for repository contents and workflow ownership, related tools for the engineering references, and CI/distribution checks for package qualification and future release boundaries.

Read Architecture and extension boundaries for domain/service ownership and resource-aware batches for optional measured CPU/RAM admission in the ordinary CLI and GUI workflows.

The GUI contracts, integration policy, GUI development environment, Java mapping and design rationale describe current responsibilities, contracts and scientific interpretation. The benchmark guide distinguishes software regression fixtures from empirical studies.

Solved modern instance from the article settings with 80 agents and approximately 20 percent obstacles

32×32 cells, 205 blocked cells (~20%), 80 agents, HeatMap with ZC and FoV 5. A selected valid modern solution from the article-verification population. Selection and provenance describe how this example was selected and checked.

New-study tools and searchable reading¶

Create a study, run a complete capsule, follow one recorded negotiation, or build the searchable local site. The public API, compatibility matrix, glossary and maintenance scope make the supported boundaries explicit.

For a first code contribution, follow the code reading order and design conventions, then the contribution checks.