Abstract:

The deep learning trend has recently impacted a variety of fields, including communication systems, where various approaches have explored the application of neural networks in place of traditional designs. Neural networks flexibly allow for data-driven optimization, but are often employed as black boxes detached from direct application of domain knowledge. Our work considers learning-based approaches to end-to-end design of modulation and signal detection for the non-coherent multi-input multi-output (MIMO) channels. We demonstrate that simulation-driven optimization can outperform traditional Grassmann designs. Additionally, we show the feasibility of noncoherent MIMO communications over extremely short channel coherence time, with as few as two time slots, which have never been explored in existing literature due to design hardness.