Neural Network Applications in Sensor Fusion for a Mobile Robot Motion

To fully utilize the information from sensors, this paper proposes a technique for autonomous robot safe moving in a partially structured environment. We used fuzzy logic and neural networks to solving the motion planning problem in mobile robot. The method of path construction is based on two neural networks. The first T-S fuzzy neural network is used to fuse the ultrasonic sensor information and to determine the free space. The outputs of the first network and other location information are as inputs of the second multilayer perceptron neural network. The second network is used to finds a safe direction for the next robot section while avoiding the obstacles. The mobile robot could recognize obstacles and environment types online, and decide its action by using the information fusion method put forward. The simulation proved that we realized mobile robot´s movement without collision in an obstacles environment. The method above mentioned is also helpful to solve the motion of mobile robot in complex environment comprising mobile obstacles.