Research, Multi-sensor Information Fusion-based Autonomous Vehicle Navigation

The ability for the Automated Guided Vehicle (AGV) to perceive environment and the functions for AGV to avoid obstacles and to navigate have always been a research focus in the field of artificial intelligence and intelligent control. This thesis presents research on the avoiding obstacles and navigation based on multi-sensor information fusion and the motion control system of AGV based on PMAC, which aims at researching and developing a fast, accurate and reliable navigation and control system for AGV and upgrading the intelligence of AGV. The main research results and conclusions are as follows. Firstly, PMAC has been applied to the control of AGV successfully, and a novel motion control system based on IPC+PMAC has been developed. It has such advantages as high control accuracy, good real-time characteristic and stability, which will make AGV walk continuously at an accuracy speed. Secondly, a multi-sensor system for QDU-Ⅰ AGV has been designed by use of ultrasonic sensor array and infrared bump-prevention sensors, and the conditions how to choose the type and number of sensors and how to optimize the layout of all sensors have been presented. Thirdly, the least-square method based on relative error is proposed and used to fit the calibration curve of the ultrasonic sensors. Using this method, we can obtain higher calibration accuracy than using Gauss least-square method when calibrating close range, which is very practical in engineering. Fourthly, the detail algorithm for AGV to avoid obstacles and navigate is presented, which is based on Kohonen clustering fuzzy neural network and makes the environmental types of AGV simple. Fifthly, being under the VC++ 6.0 developing environment and using multi-thread technology and modular design method, the software of QDU-Ⅰ AGV has been developed, which includes such modules as avoiding obstacles and navigation simulation, interactive interface, algorithm for multi-sensor information fusion, path planning, and so on. Finally, the actual tests of avoiding obstacles and navigation carried out on QDU-Ⅰ AGV have proved that above-mentioned research results and conclusions are really efficient and feasible.