Microcomputer Based Adaptive Control for an Electric Wheelchair Guidance System

The control problem associated with an electric wheelchair guidance system involves the generation of the actuator commands that are required to achieve certain specified wheelchair motions. Complications arise because of the time variations which occur in the frictional and inertial characteristics of the system. This mandates the use of an adaptive controller to maintain the stability of the system and to guarantee that performance remains within prespecified bounds.

An adaptive control concept known as the variable structure controller allows control coefficients to be stored in memory and appropriately retrieved; thus computational requirements are reduced and implementation on the electric wheelchair is feasible. A new design procedure is presented for the variable structure controller. Optimal control theory is utilized to develop a design for the time-invariant case. Cluster analysis techniques from pattern recognition theory are then used to develop regions in the time-variable parameter space throughout which a small set of controllers will maintain both stability and performance. Linear discriminant functions are then utilized as separating boundaries between the derived regions.

An analytical model of the electric wheelchair is developed to provide a means of testing the adaptive controller. Also, a proposed implementation using state-ofthe- art microcomputer technology is presented