Radar-Based Two-Dimensional Ego-Motion Estimation for Heavy Duty Vehicles
Abstract: The development of Advanced Driver Assistance Systems (ADASs) has during recent years paved the way for important improvements within the industry of long haulage driving. Never have heavy duty vehicles been that safe and efficient as they are today and one may argue that high-level competition among manufacturers and strongly regulated legislations are the major underlying causes of contribution. Further on, driver support is currently taking further steps towards autonomous driving which makes vehicle ego-motion estimation a more crucial task to deal with. Novel estimation techniques should guarantee robustness, precision and redundancy to further increase the level of performance in tomorrow’s driver support systems. This thesis presents a multi-sensor approach for two-dimensional ego-motion estimation based on a sensor set-up comprised by the above-ground type sensors Doppler radar and accelerometer along with the conventional inductive wheel encoder and yaw-rate sensor. A decentralized Kalman filter architecture with radar-based fusion state feedback and incorporated outlier detection has been implemented to gain robust long-term accuracy. Estimation of instantaneous longitudinal speed and yaw-rate are delivered by system, developed to suit a modern Scania produced heavy duty vehicle with standard specifications. The Proposed ego-motion estimation system has been, in estimation of longitudinal speed, shown to perform better than currently employed system. The proposed system is, based on tests in estimation of longitudinal speed on highway, shown capable to reduce Maximum Absolute Error (MAE) and Root Mean Square Error (RMSE) with up to 27 % and 45 %, respectively. Further on, increased sensitivity at low-speed start-stop driving has been achieved.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)