Research On Stair-Climbing Robot Control System Based On Information Fusion

Based on analyzing the multi-information fusion and neural network technology，inview of insufficient amount of information of the single sensor in Stair-Climbing robot, thispaper mainly uses ultrasonic sensors, infrared sensors as the environment perception system,and uses the multi-information fusion technology to improve the ability to adapt tounknown environments; aiming at the difficult problems of setting parameters of thetraditional PID controller, the paper designs the PID controller based on BP neural networkfor the Stair-Climbing robot, and does the simulation analysis for the traditional PID controland BP neural network PID control method using MATLAB software, and the results showthat the BP neural network PID control method with signal tracking speed, stronganti-interference ability, high stability and strong adaptability and other advantages.The main research contents of this paper are:1. The paper has studied the theories of multi-information fusion technology and neuralnetwork technology, analyzed the structure and learning algorithm of BP neural network,and proposed the corresponding solutions for the shortcomings of the algorithm.2. Build Stair-Climbing robot’s hardware system with STC15F2K60S2MCU as themain control chip, including design the triangular-star-wheel moving structure and thedumping gear mechanism. Build the environment perception system with the ultrasonicsensor supplemented infrared sensor, there are seven road ultrasonic sensor onStair-Climbing robot, measuring left and right in front of the front, front and the robot’s left and right sides of the obstacle of information, arranging two ultrasonic sensors on each sideof the robot, and choosing PSD infrared sensors to compensate the blind zone of ultrasonicsensor. Design ultrasonic sensors’ transceiver circuit, and finally define the sensors’information fusion rules.3. This paper constructs the dynamic model of the triangular-star-wheel turnoverstructure based on the Lagrange method. According to the related machine parameters ofStair-Climbing robot, built kinematics model with the triangle wheel’s line speed for output.In view of the traditional PID controller’s the control parameters are difficult to setting,large overshoot and poor anti-interference ability in the Stair-Climbing robot, and this paperproposes that the BP neural network algorithm is introduced into the conventional PIDcontroller, through neural network’s self-learning capacity and ability to approximate anyfunction, adjust the PID control parameters online to satisfy output of the neural networkunder certain optimal control.4. Using MATLAB software to compare the BP neural network PID control andconventional PID control. And the results show that tracking signalcharacteristics,anti-interference ability and adaptive ability of the Stair-Climbing robot’s PIDcontroller are improved significantly. The whole design has greatly increased the autonomousability of the robot in unknown environment.