Researchers led by Xi Jin have built FOCBS, a planner for fleets in which some robots have tasks and others have none, according to a paper on arXiv. Idle robots can block busy ones, but sending them to set parking spots adds needless travel.
FOCBS searches over the paths of robots with tasks using conflict-based search and moves the idle robots with a minimum-cost flow method. When the chosen paths leave idle robots no way to avoid collisions, it turns the blockage into constraints that guide the search without ruling out the best solution. The authors prove that it finishes and minimises the total service time of the robots with tasks.
In benchmarks it scaled well as the number of idle robots grew, and in simulation and on physical robots its plans could be carried out. In laboratory tests it cut the fleet's travel by 40.8% compared with coordination that sends idle robots to parking spots.



