Evaluation of two IMM-based algorithms in real radar tracking environments

This paper studies the performance of two IMM-based algorithms. The first is a traditional fixed structure IMM algorithm which has a constant velocity model and two constant acceleration models (IMMCVCA). The second is a variable structure IMM (VSIMM) algorithm. Radar data of maneuvering aircraft, generated by F-18 fighters and other commercial aircraft, were jointly collected by the defence departments of Canada and United States to support the practical maneuver tracking algorithm study defined in this paper. Root mean square prediction error (RMSPE) in position and computation time are used as the metrics of comparison between the two IMM-based algorithms. While the VSIMM algorithm demonstrates better root mean square performance, the IMMCVCA algorithm is shown to be more computationally efficient than the VSIMM algorithm.