Objective:

Autonomously transition a wheeled vehicle from an initial position/orientation . . .

. . . to a desired position/orientation . . .

. . . and/or follow a desired path . . .

. . . without striking obstacles or endangering load or environment.

(Quicktime Video 1.7 MB or MPEG Video 1.8 MB)

This autonomous-wheelchair maneuver from a bookshelf, to a desk, and out a door (shown at twice natural speed) appears simple enough to execute, but it actually involves serious consideration of the following complex issues.

Issues/Outline

1. What is the relationship between movement of the two wheel rotations (which can be controlled separately and directly) and the vehicle's changing location and orientation on the floor? (Nonholonomic Kinematics)

2. How can one keep track of where the vehicle is on the floor and how it is oriented in the room to within the required precision? (Estimation)
   
3. How can one plan a path which can be realized and which terminates at the desired location and orientation? (Path Planning)

4. How should the motors be commanded in order to remain on this path, or to return to the path, if the vehicle strays from it? (Control)

5. What can be done to ensure adequate precision of the vehicle at critical junctures (such as when approaching a desk)? How can unforeseen obstacles be avoided? Can planned paths be combined? (Contingencies)

6. (Summary)

Part II of this monograph is organized according to the preceding outline. To best understand Estimation Based Navigation, begin this investigation with Nonholonomic Kinematics.

Continue to: Nonholonomic Kinematics

Return to: Vision-Based Robotics Using Estimation (Home Page)