Obstacle recognition of indoor blind guide robot based on improved D-S evidence theory

Abstract

Abstract The ability to recognize obstacles with high accuracy is an important guarantee for the safe driving of indoor blind guide robots. In order to improve the accuracy of the indoor blind guide robot’s recognition of obstacles, a sensor data fusion method based on D-S evidence theory of the genetic algorithm is proposed. The system uses ultrasonic sensors, infrared sensors, and lidar to collect environmental information. Under the premise of determining the weight range of various sensors, the genetic algorithm is used to optimize each weight and the optimized weight is substituted into D-S evidence theory for data fusion. In application, the determination of the weight of evidence is the key to weighting and fusion of evidence. Through the comparison of the two fusion results, under the same conditions, the method proposed in this paper has an accuracy of 0.94 for the obstacle recognition of the indoor guide robot, which is 33.0% higher than the result of unweighted fusion. The algorithm can also be used for obstacle detection in other systems.