Sensor-data-fusion for an autonomous vehicle using a Kalman-filter

This paper presents a method to estimate the system-state, especially the full position, of an autonomous vehicle using sensor data fusion of redundant position signals based on an extended Kalman-filter. The position is detected with the help of magnet sensors attached at the vehicle and a global camera signal with low resolution, similar to GPS. A lane marked with permanent magnets and an infrared camera are used for this purpose. The vehiclepsilas driving dynamics are described using a nonlinear single-track model.