Alternatives to Monte-Carlo simulation evaluations of two multisensor fusion algorithms (Q1129665)

The authors present parallel and sequential implementations of the MultiSensor Joint Probabilistic Data Association (MSJPDA) algorithm for multitarget tracking. They extend the hybrid approximation of the covariance propagation to multiple sensors. This leads to a non-simulation comparison of the parallel and sequential implementations of the multisensor probabilistic data association filter (MSPDAF) in terms of either the average RMS position, or the expected track lifetime, or any other measure of the error. Extensive comparisons between the computational complexity and tracking performance of the sequential and parallel MSJPDA are reported. The authors' non-simulation approach provides reliable comparison for a fraction of the time that would be necessary for simulation experiments.