Zaide Zhang and Zheyuan Zhang have built a docking method for differential-drive mobile robots that combines lidar and a camera to line up precisely with workstations, according to a paper on arXiv. Such robots must correct sideways offsets through combined turning and driving in a short approach.

A line fitted to the front of a cabinet in the lidar data gives the distance and heading, and an AprilTag identifies the workstation and, combined with the lidar, its position along the cabinet. Positions taught in advance set the docking target, the robot lines up sideways before the final approach while steering by camera, and it stops early to allow for lag at low speed.

In two sessions at five workstations, all 111 docking tasks were completed, 101 of them at the first attempt. Measured on board against the taught positions, the root-mean-square errors were 1.51 mm along the approach, 5.93 mm sideways and 0.238 degrees in heading, and 96.4% of dockings were within 10 mm and 1 degree.

Source: arXiv Robotics (cs.RO)

Background

Zaide Zhang

Robotics researcher

Robotics researcher.