HTG-TA: Heterogenous Track Graph for Asynchronous Track-to-Track Association

Because of the different start times and the different update periods of local sensors, asynchronous measurements are ubiquitous in multisensor fusion systems. Therefore, it is urgent to solve the asynchronous track-to-track association (T2TA) problem. Time registration is widely used to align track to the unified time. However, the process of synchronization for track will lead to the accumulation of the estimation error. Some association methods without time registration need an additional point-by-point calculation of the tracks, which consumes a lot of time and can only focus on local track features. In light of the above problems, we propose an asynchronous T2TA method by using a heterogenous track graph (HTG-TA). The tracks in one scenario are represented by a heterogenous track graph to avoid time registration and traversal calculation. A multiscale heterogenous graphical neural network is used to focus on both local and global features. Experimental results demonstrate that HTG-TA can associate asynchronous tracks without time registration, surpass other association methods in both effectiveness and efficiency, and meet the demand of real-time association.