Gridless processing for multi-band MIMO sensing

In the field of multiple-input multiple-output (MIMO) sensing, traditional studies assume the probing signals are narrowband. With modern 5G/6G and sensing systems operating on multi-bands, models and methods tailored for multi-band active sensing are necessary. This thesis develops the model and method tailored for MIMO sensing with multi-band transmitted signals. The multi-band array can jointly sense gridless direction-of-arrivals (DoAs) and spectral responses of targets. In this thesis, we use the greatest common divisor (GCD) of multi-band for joint signal processing. The spectral responses can be different at different frequencies. We show that the gridless DoA and spectral response can be estimated jointly in numerical simulations. Due to the multi-band, the number of targets can be more than the antennas even for uniform linear arrays (ULA), which has exceeded the traditional limit of maximum resolvable targets. Subsequently, this thesis discusses the DoAs and ranges sensing in frequency diversity arrays (FDA). We formulate the problem as low rank matrix recovery problem, and solve it. We show that DoAs and ranges can be jointly sensed with a limited number of snapshots and low computational complexity. Finally, this thesis further discuss nonconvex nonsmooth optimisation problems. We analyse the algorithms: proximal gradient method, PDOM and ADMM.