Multi-sensor Data Fusion And Its Application In Obstacle Avoidance Of Cleaning Robot

Automatic detection and automatic control of the complex system can not be separated from multi-sensor data acquisition and information integrated use. Therefore, as an important information processing method, heterogeneous sensor data fusion has become a research hotspot. In this paper, the traditional mathematical algorithms and computational intelligence algorithms are used to study the data fusion of ultrasonic and infrared sensors.On the basis of programming simulation, a fuzzy neural network is constructed, and the application of multi-sensor data fusion in the autonomous obstacle avoidance of cleaning robot is studied.This article has done the following work:(1) The traditional mathematical algorithm is used to study the data fusion of heterogeneous sensors. In this paper, an adaptive weighting fusion algorithm is proposed to fuse the distance information of ultrasonic and infrared sensor ranging system. Using MATLAB to program the proposed algorithm simulation. The simulation results show that the adaptive weighting fusion algorithm proposed in this paper is stable and the convergence speed is fast, and the weights of each sensor can be allocated according to the variance of the sensor.(2) The BP neural network is used to study the data fusion of heterogeneous sensors. In this paper, a three-layer BP neural network is constructed for data fusion of ultrasonic and infrared sensors. The BP neural network of the design is simulated by MATLAB. Aiming at the problem that the standard BP neural network in the simulation results is slow and the fusion result is not ideal, a training algorithm for adding additional momentum terms is proposed. The improvement effect is compared and analyzed, and the results show that the BP neural network with additional momentum term converges faster than the standard network, and the fusion result is more accurate.(3) Research on the application of data fusion technology in obstacle avoidance of cleaning robot. A five-layer fuzzy neural network control system is constructed for theprototype system of cleaning robot with five ultrasonic distance measuring sensors and an angle sensor.The distance measured by the five ultrasonic sensors and the target azimuth angle measured by an angle sensor are taken as input variables of the fuzzy neural network system. And the network output is calculated by systematic reasoning. The fuzzy neural network system is simulated by MATLAB and the results show that when the environment is free of obstacle, the cleaning robot can move from the set starting point along the straight line to the target point. When there is an obstacle in the environment, the cleaning robot can be effective to avoid the obstacle to reach the target point.