Research On Obstacle Avoidance Of Mobile Robot Based On Multi-sensor Information Fusion

With the development of science, the research of mobile robot is also deepenedcontinuously, among which the robot’s behavior of obstacle avoidance in an unknownenvironment is always the research focus. Multi-sensor information fusion as theresearch direction is needed to carry on, the accurate detection of obstacles in theenvironment and to make the corresponding obstacle-avoiding behavior. Comparedwith the single sensor information, the credibility of the multi-sensor is greatlyimproved after the fusion, and the control precision and speed are highlighted whendealing with the nonlinear problem of robot obstacle avoidance. In this paper, thetraveler II mobile robot is treated as the research platform, the structure and algorithmof multilayer information fusion in the application of robot obstacle avoidance arestudied, the main research contents:Firstly, for the fact that the environment can’t be described accurately andcomprehensively by a single sensor, the method of multiple sensors measuring jointlyis used, on the basis of the hardware platform of the traveler II mobile robot, and theultrasonic and infrared switch is selected as the sensor of the ranging system, theCCD camera is selected as the visual sensor. A structure of a fusion of two levels ofinformation is designed, that the external information obtained by the sensor can befully used and the requirement of real-time obstacle avoidance is met.Secondly, the improved BP neural network is used to fuse the informationacquired by the distance measuring sensor. Aiming at the drawbacks of having a slowconvergence speed and being easy to fall into the local extremum of the BP neuralnetwork, the momentum-adaptive factor is attached to improve the BP networklearning algorithm, and the output results of the BP neural network and the improvedBP neural network are compared through the simulation. it’s found that the improvedneural network is much better than the unimproved one, the fused result and theunfused result are also compared, which discovers that the fused result is closer to thereal value.Thirdly, for the problem of many interference factors in the image backgroundsacquired by the visual sensor, a method of combing two results of considering fromthe color, extracting an area, then by the gray, high-pass filtering, denoising and theedge detection is put forward, that the correct edge information is acquired,and thedistance between the left and right edge and the image center line is extracted.Fourthly, an information fusion method of fuzzy neural network basing on theoptimized T-S model is used for the control of the mobile robot obstacle avoidance,the replacement factor is used to improve the problems of shocks and slow speed that r may occur during the parameters learning and updating process, it has been verifiedby the simulation that the improved fuzzy neural network has a quick convergencespeed, and the output value of the network and the real value of the test sample arecompared to find that its precision can satisfy the control requirements of the obstacleavoidance.Finally, with the traveler II mobile robot hardware development platform, a set ofmobile robot motion control system is designed, VC++6.0is developed by systemapplication layer, Using mixed programming with VC and MATLAB in thedecision-making layer, By calling MATLAB engine function, the obstacle avoidancecontroller based on two layers of multi-sensor information fusion designed inMATLAB is applied to the mobile robot motion control system. By using thedesigned mobile robot motion control system, the robot movement in a realenvironment is realized, obstacles that exist in the environment are avoidedautomatically, the validity of the designed algorithm is verified.