Total: 1
Multistatic collaborative sensing eliminates self-interference, achieves spatial diversity gains, and enables wide-range seamless integrated sensing and communication (ISAC). However, conventional data fusion methods suffer from severe error amplification in geometry-sensitive regions. In addition, the conventional analog phased array solution introduces large beam sweeping overhead, whereas the fully digital arrays request high hardware cost. We propose a multistatic sensing framework enabled by a phase-time array (PTA). The rainbow beamforming maps spatial directions to orthogonal frequency division multiplexing (OFDM) subcarriers, achieving wide-angle coverage with a single radio frequency (RF) chain. We develop two parameter-level schemes-a geometry-aware analytical estimator (GDOP-WLS) and a lightweight multilayer perceptron (PF-MLP)-to mitigate the effects of topological singularities. Additionally, an end-to-end signal-level convolutional neural network (SF-CNN) directly estimates target coordinates from raw signals, avoiding cascaded estimation errors. The results demonstrate that the parameter-level schemes ensure robust convergence under adverse geometric conditions with minimal computational latency. Conversely, the signal-level scheme achieves sub-meter precision but requires an increased computational load. Consequently, the proposed framework establishes a scalable solution for collaborative surveillance of unmanned aerial vehicles (UAVs), providing flexible trade-offs among hardware complexity, latency, and accuracy.