Navigation of an autonomous vehicle using a combined electrostatic potential field/fuzzy inference approach

A two layered fuzzy controller, combined with an electrostatic potential field path planner is created to implement navigation of an autonomous mobile vehicle in a dynamic two-dimensional environment. The environment is mapped into a resistor network and an artificial potential field is created through injection of current into the network. The path of maximum current, through the network, is the path of minimum distance in the environment. The potential field guarantees a near-optimal path in a static environment. The fuzzy control implements sensor fusion as well as low level control to ensure the path generated is followed as long as collisions with both moving and non-moving obstacles are avoided. The first layer fuzzy inference engine provides a model for fusion of the multiple sensor inputs. The second layer fuzzy controller implements collision avoidance with dynamic obstacles while following the trajectory defined by the potential field. The proposed method is illustrated through various examples.