We evaluate the influence of multi-snapshot sensing and varying signal-to-noise ratio (SNR) on the overall performance of neural network (NN)-based joint communication and sensing (JCAS) systems. To enhance the training behavior, we decouple the loss functions from the respective SNR values and the number of sensing snapshots, using bounds of the sensing performance. Pre-processing is done through conventional sensing signal processing steps on the inputs to the sensing NN. The proposed method outperforms classical algorithms, such as a Neyman-Pearson-based power detector for object detection and ESPRIT for angle of arrival (AoA) estimation for quadrature amplitude modulation (QAM) at low SNRs.
This article introduces a bistatic joint radar-communication (RadCom) system based on orthogonal frequency-division multiplexing (OFDM). In this context, the adopted OFDM frame structure is described and system model encompassing time, frequency, and sampling synchronization mismatches between the transmitter and receiver of the bistatic system is outlined. Next, the signal processing approaches for synchronization and communication are discussed, and radar sensing processing approaches using either only pilots or a reconstructed OFDM frame based on the estimated receive communication data are presented. Finally, proof-of-concept measurement results are presented to validate the investigated system and a trade-off between frame size and the performance of the aforementioned processing steps is observed.
We investigate the potential of autoencoders (AEs) for building a joint communication and sensing (JCAS) system that enables communication with one user while detecting multiple radar targets and estimating their positions. Foremost, we develop a suitable encoding scheme for the training of the AE and for targeting a fixed false alarm rate of the target detection during training. We compare this encoding with the classification approach using one-hot encoding for radar target detection. Furthermore, we propose a new training method that complies with possible ambiguities in the target locations. We consider different options for training the detection of multiple targets. We can show that our proposed approach based on permuting and sorting can enhance the angle estimation performance so that single snapshot estimations with a low standard deviation become possible. We outperform an ESPRIT benchmark for small numbers of measurement samples.