Swapping based joint estimation of uniform state model

The paper presents an algorithm for the on-line joint parameter and state estimation of the state model whose innovations are uniformly distributed. We use a Bayesian approach and evaluate a maximum a posteriori probability (MAP) estimates in discrete time instants. As the model innovations have a bounded support, the searched estimates lie within a set that is described by the system of inequations. In consequence, the problem of MAP estimation can be easily converted to the problem of linear programming. A joint state and parameter estimation is performed as the alternating subtasks of state filtration and parameter estimation. The resulting estimation algorithm is applied to the traffic data.