Kalman filter process models for urban vehicle tracking

Faced with increasing congestion on urban roads, authorities need better real-time traffic information to manage traffic. Kalman filters are efficient algorithms that can be adapted to track vehicles in urban traffic given noisy sensor data. A Kalman filter process model that approximates dynamic vehicle behaviour is a reusable subsystem for modelling the dynamics of a multi-vehicle traffic system. The challenge is choosing an appropriate process model that produces the smallest estimation errors. This paper provides a comparative analysis and evaluation of linear and unscented Kalman filters process models for urban traffic applications.