Performance Evaluation of Interacting Multiple Model Kalman Filter

Abstract

Conventional Kalman filter having constant velocity model (KFcv), conventional Kalman filter having constant acceleration model (KFca) and Interacting Multiple Model Kalman Alter (IMMKF) having constant velocity & constant acceleration models were implemented and their performance was studied. Overall performance of IMMKF was better than the other two. The performance of IMMKF with different process noise variance, measurement noise variance and model transition matrix was explored. It was concluded that Q1 and Q2 higher than their true would give better performance. The measurement noise variance R lesser than the true value gives improved results. The diagonal elements of model transition matrix derived from sojourn time shows superior performance. Results showed that, IMMKF exhibits overall better performance and was more suitable for tracking maneuvering targets than conventional Kalman filters.