Machine intelligence for crisis handling in navigating vehicles using neuro-controllers

The paper addresses the design and development of an intelligent neuro controller for navigating vehicles, which can respond to a crisis when the human driver fails to react appropriately. Crisis situations were simulated analytically in a program with random changes in the road curvatures. Using transient dynamic equations and a vehicle model, the vehicle states, responses such as yaw rate, lateral velocity and variation of constructs such as the lateral position with respect to road centre (LPRC), heading error with respect to local curvature (HELC) under simulation trials were visually updated on the screen to enable manual manipulation of control inputs to guide the simulated navigation behaviour of the vehicle. A neuro controller was made to learn the inherent dynamics by an association of the vehicle states during such situations with the pattern of human reactions and his choice of inputs viz. the throttle, steer, brake and gear in controlling such crisis. Such a neuro controller then, has the ability to invoke such past learning of successful reaction patterns of the human control action during moments of human incapacity to react under crisis situations. The neuro controller guides the vehicle navigation under varying crisis conditions preventing road departures. The capability of the trained neuro controller to improvise effectively under unpresented crisis situations and maintain stability, controllability and safety of the vehicle has also been explored.