M 2 TrackFormer: Transformer-Based mmWave Tracking with Lost Target Re-Acquisition Capability

This paper considers a millimeter wave (mmWave) integrated sensing and communication (ISAC) system, where a base station (BS) equipped with a large number of antennas emits dedicatedly designed beams for persistent tracking of a moving target. Conventional tracking techniques assume that the echo signals from the moving target are always available. However, the mmWave beams generated by the BS are very narrow, and the tracking lost issue arising from the misalignment between the pencil-like beam and the true target position is inevitable, especially when the trajectory of the target is complex. To deal with this issue, we propose a Transformer-based mmWave tracking framework, namely m2TrackFormer, with a re-acquisition mechanism, such that even if the echo signal from the target is too weak to extract sensing information, we are able to re-acquire the target location quickly with small beam sweeping overhead. Specifically, we utilize the parallel processing capability of the Transformer to predict multiple possible trajectories of the target in the future. When the tracking lost event occurs, the BS can scan beams to just several sites where the target is most likely to be. Numerical experiments demonstrate that the proposed framework performs robust tracking even when the trajectory is complex.