Decentralized Fuzzy Control of Multiple Nonholonomic Vehicles

This work considers the problem of controlling multiple nonholonomic vehicles so that they converge to a scent source without colliding with each other. Since the control is to be implemented on a simple 8-bit microcontroller, fuzzy control rules are used to simplify a linear quadratic regulator control design. The inputs to the fuzzy controllers for each vehicle are the noisy direction to the source, the distance to the closest neighbor vehicle, and the direction to the closest vehicle. These directions are discretized into four values: forward, behind, left, and right; and the distance into three values: near, far, and gone. The values of the control at these discrete values are obtained based on the collision-avoidance repulsive forces and an attractive force towards the goal. A fuzzy inference system is used to obtain control values from a small number of discrete input values. Simulation results are provided which demonstrate that the fuzzy control law performs well compared to the exact controller. In fact, the fuzzy controller demonstrates improved robustness to noise.