Autonomous Vehicle Platform And Key Technique Research Focusing On Urban Environment

The Intelligent vehicles have been considered an effective method tosolve the more and more serious urban transportation problem. Thisthesis proposed an intelligent vehicle platform focused on urbanenvironment, and do some further researches in the key techniques, whichare necessary during autonomous navigation of intelligent vehicles. Thereare four essentials: the vehicle control scheme design, the GPS-INS basedvehicle positioning method, the road navigation method using multisensors, the map based intersection navigation method.In the vehicle control scheme design section, the requirementanalysis is given. A scheme design comes up to fulfill these requirements.In the design, Chery Tiggo is used as the basic platform. The steeringsystem, the throttle and the brake part are refitted into electroniccontrolled ones. Specific control methods are also provided. To get exactvalue of relatively low speed, extra sensors shall be added. A new methodusing hall sensor is proposed, the service life of which is much longerthan the widely used optical encoder and flexible shaft, because of thenon-mechanical and non-contact features.In the GPS-INS based vehicle positioning method section, thehigh-precision GPS is considered inappropriate in urban environment forits sensitivity to the signal quality which is easily disturbed by thebuildings, high ways and trees. Thus, low precision GPS shall be adopted.To increase the positioning accuracy, an Inertial Navigation System(INS)is used to give a predicted position in the extended Kalman filter. And themeasured value in it is from the low-precision GPS. Experiments are conducted in the campus to prove the effectiveness.In the road navigation method using multi sensors section, aprogram structure with two layers is proposed. The upper layer gives anideal control command mainly based on the lane detection result from thecamera. The position and coursing angle of the vehicle determine thecontrol command when the camera missed the lane. When given thecontrol command, the lower layer adjusted it if necessary to keep thevehicle safe. The available area, or safe region as well, combined both theroad bound result calculated from the two pitched-down laser scannersand the obstacle detecting result from the horizontal-mounted laserscanner. A quadratic model is applied to approximate the intersection lineof the scanning plane and the road surface instead of the traditional linearone, for the road surface is normally a camber one in the consideration ofdewatering. The road bound detecting method is tested with the datacollected inside and out sided campus. An autonomous navigation in a16km long countryside road is proceeded to confirm the navigationmethod.In the map based intersection navigation method section, an extraintersection map is built up to provide more information making up forthe lost of lane lines. The map gives the entry point and exit point of theintersection and the road directions as well. From these, the intelligentvehicle can get through the intersection only with the position andcoursing angle. To further improve the smooth of the track, several virtualtarget point is added during the progress. Also, an autonomous navigationexperiment is conducted in the urban area of Chifeng, which prove thewhole system a practicable one.