|
Abstract:
|
Environment perception of a robot vehicle, including the determination of the objects around the robot vehicle and the position of the vehicle itself, is very important in autonomous robot vehicles. In this thesis is developed two methods for this purpose. One of them is a signal-fusion method which uses the vector-median/extended-vector-median (VM/EVM) filters to fuse the sensory data from different sensors or different moments. The method can reduce the measurement errors effectively and model the environment accurately. The second one is a map-matching method which matches the sensor range data to the a priori map of the working environment, based on the original position obtained from the odometry sensor, to update the position of the robot vehicle. The position updating can be achieved during or after the matching procedure. These two methods can be applied separately to fulfil different tasks. They can also be unified into a system: the signal produced from the fusion subsystem is used as the input signal to the matching sub-system. In this way, a faster and more accurate result of matching method are obtained. Meanwhile, a more accurate position of robot vehicle is also expected. The develop methods are shown to be robust and efficient. Trough computer simulation and experimental test, their abilities to reduce measurement uncertainties and errors are examined and more accurate positions for the robot vehicle are obtained. /Kir11 |