An evidential approach to environment sensing for autonomous robot

Abstract

In order to improve the correctness of the environment sensing, multiple ultrasonic sensors are applied to acquire the information of robot's surrounding environment and a modified evidential theory is applied to analyze and fuse sensor data. Sensor model and definite discrete beliefs to ranging data are presented according to the physical characteristics of ultrasonic sensors. Some issues such as the data fusion sequence, the data fusion level, the definition of the frame of discernment, the choices of evidences, the modified fusion algorithm and the establishment of decision rules are studied profoundly which all belong to the domain of the application of evidential theory to environment sensing. Computer simulation results indicate that although approximation and neglect during algorithm inference and simulations will lead recognition error, the error has little influence on robot navigation. So the proposed method has very perfect effects on environment sensing.