MMWAVE WI-FI TRAJECTORY ESTIMATION WITH CONTINUOUS-TIME NEURAL DYNAMIC LEARNING

We leverage standards-compliant beam training measurements from commercial-of-the-shelf (COTS) 802.11ad/ay devices for moving object localization. Two technical challenges need to be addressed: one is the sparsity and irregularly sampled nature of the beam training measurements due to beam scanning overhead control and contention-based channel-time allocation, and the other is to exploit underlying object dynamics to assist the localization. To this end, we formulate the trajectory estimation as a sequence regression problem and propose a dual-decoder neural dynamic learning framework to simultaneously reconstruct Wi-Fi beam training measurements at irregular time instances and learn unknown dynamics over the latent space in a continuous-time fashion by enforcing strong supervision at the coordinate level. The proposed method was evaluated on an in-house mmWave Wi-Fi dataset and compared with a range of baseline methods including traditional machine learning methods and recurrent neural networks.