Skip to content

Add sequential single-item auction for multi-robot task - #1462

Open
vishwagw wants to merge 2 commits into
AtsushiSakai:masterfrom
vishwagw:ssi-auction
Open

vishwagw wants to merge 2 commits into
AtsushiSakai:masterfrom
vishwagw:ssi-auction

Conversation

@vishwagw

@vishwagw vishwagw commented Oct 12, 2026 •

Copy link
Copy Markdown

Reference issue

Closes #: 1460

What does this implement/fix?

Adds a Sequential Single-Item (SSI) auction for multi-robot task allocation under Mission Planning (Lagoudakis et al., RSS 2005; Koenig et al., AAAI 2006).

Each round, every robot bids on every unassigned target and the lowest bid wins; the target is inserted into the winner's route at its cheapest position. Two team objectives:

  • MiniSum: bid = marginal cost → minimizes total distance
  • MiniMax: bid = new route cost → minimizes the longest route (completion time)

Travel costs are A* path lengths reusing PathPlanning/AStar/a_star.py, so allocations respect obstacles. Only existing requirements are used.

Files:

  • MissionPlanning/SequentialSingleItemAuction/sequential_single_item_auction.py: example with main() and show_animation
  • tests/test_sequential_single_item_auction.py: 9 tests
  • docs/modules/13_mission_planning/sequential_single_item_auction/sequential_single_item_auction_main.rst: problem, math, algorithm, results, comparison with related methods, autodoc links; added to the Mission Planning toctree

Results on the example map:

Objective Total distance Longest route Targets per robot
MiniSum 136 m 114 m 0, 9, 1
MiniMax 195 m 76 m 1, 7, 2

Additional information

  • Animation GIF PR:
  • Tests include obstacle awareness (a target behind a wall goes to the robot with the shorter path, even though another robot is closer in a straight line) and the MiniSum/MiniMax trade-off.
  • Locally on Python 3.13 with the pinned requirements: pytest tests -Werror → 165 passed, including the codestyle and mypy tests. mypy on the new module is also clean. MissionPlanning isn't in the mypy test's package list, so I ran it separately.
  • Docs built with Sphinx with no warnings for the new page. I checked the rendered HTML: the page is reachable from Mission Planning, and the table and parameter sections render.

CheckList

  • Did you add an unittest for your new example or defect fix?
  • Did you add documents for your new example?
  • All CIs are green?

@dajiaohuang

Copy link
Copy Markdown

Scoped review of open PR #1462

Captured current PR metadata and full 527-line diff at the task's capture time. This is an issue/PR contribution review, not CI/status polling or a full repository audit.

  1. Short wall segment crashes build_map. At the new helper's steps = int(max(abs(x1 - x0), abs(y1 - y0))), a valid-looking sub-unit segment such as ((1.0, 1.0), (1.5, 1.0)) yields steps == 0; the first loop iteration divides by steps, raising ZeroDivisionError. The function documents axis-aligned walls but no minimum length restriction. Clamp/sample at least one interval or reject unsupported lengths explicitly.
  2. A animation global is permanently changed.* TravelCostMap.__init__ assigns a_star.show_animation = False on the imported module and never restores it. Constructing a cost map in an application that uses A* elsewhere silently disables the other caller's animation. The local intent (quiet internal A* searches) should be scoped to searches or restore the previous setting.
  3. Recorded optimization plan versus implementation. On Issue Proposal: Sequential Single-Item (SSI) auction for multi-robot task allocation #1460, the PR author said they would use one Dijkstra flood fill per robot start/target to populate all pair costs. The submitted helper instead invokes A* once per distinct point pair and caches reverse pairs. This is not necessarily incorrect for the small example, but the PR should clarify the deliberate tradeoff or follow the stated optimization if larger instances are in scope.

No duplicate implementation was created. The feature already has PR #1462; this review stays on that contribution.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants