logo motion mind
Mapping and Vacuuming

MotionMind/Swan is supposed to understand a vacuuming area that is partially bounded by walls.

First a human master, Andrew, seven years old in the following video, teaches the boundary of this area by pulling Swan and by stopping at each vertex of the area boundary. At the end of the teaching phase, he explicitly tells Swan with a hand sign that they are back home. Watch the video, Vacuuming-Teaching. Here, the pull function plays the crucial role of spatial teaching. Swan learns a polygon boundary with a special vertex, home. MotionMind/Swan creates a map, 0403vt_andrew.map, whose size is only 1227 bytes.

In an execution phase MotionMind/Swan first loads this map, 0403vt_andrew.map, for a simulated vacuuming. Then it starts sweeping the boundary, keeping a 10 cm gap between the boundary and itself. Then it makes an inward spiral to sweep the entire area. A new inner path is shifted by 27 cm, which is the width of Swan. When the simulated sweeping is finished, it goes back home. Watch the video, Vacuuming-Execution

Normally only one teaching is necessary. Then, using the map, you can execute vacuuming as many times as you want. This map can be used for other MotionMind-based robots in other buildings if the area is the same.

If you need to define another vacuuming area for a new task, you can teach Swan again to create another map with a distinct name. Swan can create and keep as many maps as needed.

In both teaching and execution phases, several localization algorithms work. Without them, we cannot expect Swan to function appropriately, even for this simple vacuuming area. Normally Swan localizes itself when Swan is facing reference walls.

No landmarks or use of optical/magnetic tape pasted to the floor are necessary. Using this human-guided teaching method, creating a map costs robot users nothing.

© 2010 MotionLab, LLC. Website Design & Web Development: www.businessprodesigns.com