logo motion mind
Mapping and Navigation

This is another example of spatial-understanding practice by robots. This scenario would be closer to practical applications.

(I) We define a part of the MotionLab office space as operation area "labA." To demonstrate this area, "labA", we prepared a movie, Panorama. Here, each wine bottle represents a featured point to which a name is given: "home," "sf," "corner," "copier," "ny," "library," and "fridge," in that order.

(II) We teach Swan "labA" using the human-guided teaching/learning method. As a result, Swan creates a map, labA.map, which is only 3522 bytes. The video of this teaching/learning phase is not available.

(III) Now we come to a navigation-executing phase. Suppose Swan is initially "home" and is instructed to go to "copier," then to "library," and then go back "home." This name database is included in the map and is shared by both the master and Swan. This is the start of human-robot intelligent communication.

(IV) Next, MotionMind/Swan finds the following minimum-cost paths for the given instructions, using labA.map and Dijkstra's algorithm:
1. <"home" "sf" "corner" "copier">
2. <"copier" "corner" "sf" "ny" "library">
3. <"library" "ny" "sf" "home">

(V) Finally, Swan executes the navigation task. Watch the video, Map-Based Navigation. See the trajectory.

(VI) During the navigation, MotionMind/Swan executes several localization algorithms with the loaded map, labA.map. Without these, Swan could not have completed the trip.

As you can see, Swan can navigate itself without the assistance of artificial landmarks or tape pasted on the floor.

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