Researchers led by Haoze Dong have presented a way to coordinate teams of robots that must pass through narrow passages, according to a paper on arXiv submitted to IEEE Robotics and Automation Letters. Fixed global schedules break down when robots deviate from them, while purely local methods cannot plan far ahead, the authors write.

Their framework plans routes that take account of how passages are used and in which direction, penalises plans in which robots would meet in the same place at the same time, and leaves the final collision avoidance to a controller that runs online. Robots that fall behind or move ahead are tracked flexibly, and routes are replanned only where needed instead of reserving fixed trajectories.

Across 1,200 simulated cases in four environments with 5 to 30 robots, it succeeded almost every time and beat the other methods tested on completion time. In hardware experiments, eight robots completed all trials without a collision.

Source: arXiv Robotics (cs.RO)

Background

Haoze Dong

Robotics researcher

Robotics researcher.