agv op• Data kinds: image2d → table
• Call: import fullseye as fs; fs.ledger.mapf_ecbs(grid, starts, goals, w=1.2, max_nodes=50000, horizon=None) (to call the implementation directly, import agvfleet; agvfleet.mapf_ecbs(grid, starts, goals, w=1.2, max_nodes=50000, horizon=None); from the registry, opsdrive.get("mapf_ecbs"))
CBS with focal search at the high level (the same idea as the high level of ECBS, Barer et al. 2014). Sum of costs is within w times the optimum.
The minimum total cost LB among the open nodes is at most the optimum (the optimal solution lies under some node). So, among the nodes with コスト ≤ w · LB (the focal set), the one with the fewest conflicts is expanded. The returned plan satisfies cost ≤ w · LB ≤ w · 最適 (gate). The return value is the dict of mapf_cbs plus lower_bound and w. Where optimal CBS stalls in narrow corridors by examining every pair of equally long yielding orders, this escapes by leaning toward the side with fewer conflicts.
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• poc_agv_fleet — py -3.11 examples/poc_agv_fleet.py
table as input)course_layout · course_occupancy · course_contains · world_build · world_camera · world_move · lidar_scan · rss_longitudinal_same
agv)warehouse_grid · grid_distances · mapf_cbs · mapf_joint_astar · mapf_prioritized · plan_conflicts · plan_cost · adg_build
*Provenance: agvfleet.py — DRIVE operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.