A Fast Implementation of Interactive-Model Generalized Labeled Multi-Bernoulli Filter for Interval Measurements

Abstract Aiming at tracking multi-maneuvering targets for interval measurements, we introduce the multi-model concept into the Generalized Labeled Multi-Bernoulli (GLMB) algorithm and propose an Interactive Multi-Model GLMB (IMM-GLMB) tracking algorithm for interval measurements. The sampled particles are predicted by the Interactive Multi-Model (IMM) algorithm. Afterwards, we update the predicted particles by introducing a generalized likelihood function in conjunction with the GLMB filter update strategy. With the combination of the characteristics of GLMB filtering and the IMM method, this algorithm can effectively improve the target state prediction accuracy and avoid tracking failure caused by model mismatch in the maneuvering process of targets, but it requires different truncation step for each component in the prediction and update. In order to improve the efficiency of the algorithm, we adopt the idea of a fast algorithm proposed in the literature and propose the IMM Generalized Labeled Multi-Bernoulli Fast (IMM-GLMBF) algorithm. This IMM-GLMBF algorithm integrates the prediction and the update steps of the IMM-GLMB algorithm, so that it requires only one truncation. Compared to the IMM-GLMB algorithm, the proposed IMM-GLMBF algorithm is more accurate in estimating the number of targets and their states, and it greatly reduces the computation cost. In addition, we present the Sequential Monte Carlo (SMC) implementation of the IMM-GLMB and IMM-GLMBF algorithms.