Parallelized particle filtering for freeway traffic state tracking

We consider parallelized particle filters for state tracking (estimation) of freeway traffic networks. Particle filters can accurately solve the state estimation problem for general nonlinear systems with non-Gaussian noises. However, this high accuracy may come at the cost of high computational demand. We present two parallelized particle filtering algorithms where the calculations are divided over several processing units (PUs) which reduces the computational demand per processing unit. Existing parallelization approaches typically assign sets of particles to PUs such that each full particle resides at one PU. In contrast, we partition each particle according to a partitioning of the network into subnetworks based on the topology of the network. The centralized case and the two proposed approaches are evaluated with a benchmark problem by comparing the estimation accuracy, computational complexity and communication needs. This approach is in general applicable to systems where it is possible to partition the overall state into subsets of states, such that most of the interaction takes place within the subsets.