Sensor-based path planning and intelligent steering control of nonholonomic mobile robots

This paper presents a control scheme for autonomous navigation of intelligent mobile robots under unstructured environments. Based on environmental information, which is acquired using external sensors or given by the upper level, the control scheme performs autonomous navigation composed of sensor-based path planning and tracking control in real time. First, using a simple genetic algorithm, the path planning module generates an obstacle-free path as a sequence of control vectors of orientation, considering kinematic constraints in steering control of wheeled mobile robots. Then, the tracking control module calculates the references for motion control of the mobile robot using a sensor fusion neural network. Simulation experiments of path planning under unstructured environments with several obstacles are illustrated. An experimental procedure for teaching the sensor fusion network is introduced, and basic characteristics of the internal and external sensors during straight-line and circular movements on the floor with black-striped markers are measured using our experimental small robot to show the feasibility of the proposed control scheme.