In our work on design of next-gen communication and sensing systems (including joint design to take advantage of a common infrastructure), our emphasis is on the millimeter wave (mmWave) bands. The high carrier frequencies (small carrier wavelengths) in these bands enable the realization of antenna arrays with a large number of elements in compact form factors, leading to enhanced spatial reuse (communication) and spatial resolution (sensing). The orders of magnitude higher available bandwidths in these bands translate to corresponding increases in data rates (communication) and ranging accuracy (sensing). We have been working in this area since about 2005, but the research challenges keep getting more and more interesting.  Here’s a recent presentation which provides some research snapshots and perspective.  A more detailed account of our research in this area is provided here. 

UCSB has established leadership in this exciting area through interdisciplinary collaborations spanning system and algorithm design and experiments, circuit design, and device research.  We are grateful for the funding our group has received in the area of mmWave systems from a series of NSF projects, and through participation in large centers funded by DARPA and the Semiconductor Research Consortium (SRC), with interdisciplinary collaborations at the boundaries of signal processing, circuits, and networks forming the intellectual core of our efforts.  Interdisciplinary NSF-funded projects include GigaNets (completed), a project on signal processing/hardware co-design of extended arrays funded by the RINGS program (ongoing), and 4D100, a project on Joint Communication and Imaging (ongoing).  DARPA/SRC funding for our group comes from our participation in three successive centers: SONIC (2013-17), a center under the STARNET program led by the University of Illinois (2013-17), ComSenTer (2018-22), a center under the JUMP program led by our UCSB collaborator Prof. Mark Rodwell, and CUbiC (2023-27), a center under the JUMP 2.0 program led by Columbia University.

Since about 2010, we have seen dramatic advances in AI, with deep neural networks (DNNs) rapidly becoming the state of the art in an ever-expanding array of fields (computer vision, speech processing, natural language processing, scientific discovery, …), learning to infer, to recommend, and to generate.  While these developments are truly stunning and are fueling huge R&D efforts in industry and academia, we are pursuing a slightly contrarian agenda in our group.  We are intrigued by the lack of robustness and transparency of DNNs, and wonder whether ideas from communication theory and neuroscience can help.  We are worried about data-driven AI reinforcing biases in the society from which the data is drawn, and wonder whether we can come up with proactive means of using automated decision-making to increase societal fairness. Here’s a presentation on recent work on robustness and fairness.  And when it comes to applying DNNs to specific applications, we worry about whether they are actually learning what we mean them to: here’s a presentation striking a cautionary note on using DNNs for wireless fingerprinting.

We gratefully acknowledge funding for our work in this area from NSF, ARO and the DARPA RFMLS program.