The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9502

Title:Experiments in Rule Based Adaptive Control
Date: 1995
Abstract:We present the rule based adaptive (RBA) control approach which uses rules for accomplishing tasks in a what might be changing environment and for adapting to physical changes in the system itself. We propose a robot organisation under the approach consisting of arbitrated stimulus-response modules and a low-level controller which abstracts the motor system. We illustrate the RBA approach with two experiments. The first experiment involved two real LEGO mobile robots where we solved some of the implementation issues concerning the robot organisation. We tested the mobile robot controller for adaptivity by reversing the motors while in operation for which the controller adapted quickly. We tested program portability by porting the program to another robot with a different structure and, consequently, most of the changes were in the RBA controller only thus preserving the structure of the solution. The second experiment was a robot arm simulation using the Simderella simulator. The controller's task was to position the end effector on the target in preparation for grasping by issuing joint angle changes based on visual information alone. The joint angle values and visual information consisting of the target and end effector positions were available to the controller through the simulator. The controller did not require any other prior information about the link lengths or carried out kinematic calculations. We present experimental results showing how the controller handled link length changes and random motor reversals.

[Search These Pages] [DAI Home Page] [Comment]