Interactive Multiple Model (IMM) Algorithm Implementation in MATLAB

This article presents a MATLAB implementation of the Interactive Multiple Model (IMM) algorithm, incorporating left-turn, right-turn, and constant-velocity motion models. The implementation employs Kalman filtering for predictive estimation to enhance tracking accuracy. Notably, the code includes multiple error evaluation methods and visualization capabilities for comprehensive performance assessment. The algorithm structure features model probability updates through Markov chain transitions and interactive mixing of model-conditioned estimates.

This implementation is suitable for target tracking applications where the state space comprises position and velocity components. The package includes modular files covering moving target tracking, model transition probability updates, and Kalman filter operations, providing complete algorithmic coverage for researchers and engineers. The implementation demonstrates excellent performance through proper handling of model switching via transition probability matrices and optimal state fusion, offering improved accuracy and efficiency for relevant research and practical applications. Key functions include model-specific prediction steps, innovation covariance calculations, and likelihood-based model probability updates.