Vehicle Motion Estimation using Low-Cost Optical Flow and Sensor Fusion

This paper explores the use of low-cost optical flow sensors and fusion with other commonly-used sensors for automotive vehicle motion estimation. In this research, 3 types of sensors are used. A set of custom-made optical flow sensors using low-cost optical mouse chips provide velocity in 2D at low speeds and yaw rate indirectly. An inertial measurement unit, which is commonly used for vehicle motion detection, provides velocity and yaw rate of the vehicle at high speeds. Lastly, the vehicle's own wheel sensor, via On-Board Diagnostics-II, is used to provide low resolution forward speed. A Kalman filter is designed to fuse the three types of sensors and provide a more robust and accurate sensor system. Simulations and testing on an actual outdoor vehicle show that sensor fusion significantly improves the result compared to when each type of sensor is used alone.