In this thesis a vehicle tracking problem using an ultra-wideband radar sensor is considered. Prior research is heavily focused on specific applications, such as highway driving, where tracked vehicle motion is confined and limited. The target application of this thesis is one of low speed but high variability in tracked-vehicle’s entry and exit points.
After analysis of common nonlinear estimation techniques, and with the target application in mind, the tracker is developed within a Particle Filter framework. Given the cluttered nature of the radar-sensor data, pruning and gating methods are formulated for use in the measurement update procedure. Considering the quality and separation of vehicle data points within the radar-sensor data, a simple data association step is developed that facilitates the tracking of multiple vehicles simultaneously and independently. The system is extended to a moving platform via developed mappings from the radar frame-of-reference to an inertial frame, and vice versa. An
Extended Kalman Filter is developed to estimate the platform’s state from limited, noisy sensor measurements.
The results show that the developed system is successful in detecting and tracking single and multiple vehicles when using real-world data from the radar sensor. The Extended Kalman Filter is also shown to provide a suitable state estimate when using real-world data. Testing of the two systems jointly is advised for future research.