Comparison of augmented state track fusion methods for non-full-rate communication

For linear-Gaussian non-deterministic dynamics, that is, systems with non-zero process noise, it is well known that tracklet fusion based on equivalent measurement is optimal only for full communication rate, i.e., if the local posterior probabilities or estimates are communicated and fused after each observation and update time. Despite this constraint, tracklet fusion has become very popular because it performs well in many real world problems even when communication is not at full rate. By including local state estimates at multiple times, augmented state (AS) tracklet fusion computes the optimal global estimate despite this communication constraint. A similar method with this property is distributed accumulated state density (DASD) fusion, which computes decorrelated local pseudo estimates by means of a relaxed evolution model. This paper compares these two methods by examining their underlying principles. Numerical results compare their performance and also with that of a centralized Kalman filter. The results show that they have many properties such as the estimation accuracy in common despite their different derivations.