With ever-increasing data-rate requirements for different applications, optimizing the existing wireless networks from sub-6 GHz to THz is needed. The wireless networks need to be driven by sensing the entire spectrum to achieve optimized performance. Even with my work on full-duplex radios, the radios would be limited to sensing a fraction of the spectrum and cannot cover the entire spectrum. My vision has been to design networks driven by global sensing of the spectrum, instead of individualized sensing. Cloud management of RAN can use the sensing information to optimize the entire network of base-stations. Thus, in the emerging spectrum landscape, sensing needs to achieve high agility in space, time, frequency, and dynamic range is necessary for next-generation network optimization. The core bottleneck is to enable a low-cost entire spectrum sensing with a high-time resolution, which led to no cognitive radio deployments to facilitate dynamic spectrum use. Furthermore, it is extremely hard to sense the entire spectrum with fine resolution in time, space, and simultaneously sense weak and the strong signals that are spectrally close.
My research has designed a low-cost, mobile sensor that sweeps the spectrum capturing the spectral activity for every frequency for a short period and reconstructs the entire spectral activity, demonstrating the world’s first low-cost, entire sub-6 GHz of spectrum captured in millisecond. My work introduced a sweeping receiver architecture that limits the sub-set of spectrum band and captures the ADC samples. Such an architecture leads to the short observation of each signal, which, combined with signal processing and machine learning, can detect accurately and preserve users’ privacy. It continuously sweeps the entire spectrum rapidly, allowing for signals on the different bands to naturally appear one at a time for a short period and then disappear. This technique effectively provides high resolution for spectrum sensing. It allows the receiver to sense strong signals when they are swept with sweeping architecture and blocks strong signals when they are sweep-sensing a weak signal. The next core bottleneck is shipping the sensed data in real-time to feedback to the core network (which turns out to be 800 Mbps of data from a single sensor) near impossible to ship. My research has developed SparSDR, which is an edge-compression technique that can compress (by 200x) the observation of spectrum and ship only helpful information to the cloud with low latency, building necessary tools to enable sensing-driven communication. The low-cost sensor that can sense the entire spectrum and quickly ship essential data from the sensor has evolved research on privacy and security of communications. Specifically, in our recent work, we showed that Bluetooth radios could uniquely be profiled with hardware signatures be and tracked by our sensors, even if they continuously change MAC address, leading to zero privacy for users. Our work on sensing exposes the privacy and security flaws of wireless networks.
This work from my group on sensing solves a long-standing problem, needed by recent calls from NSF Spectrum Innovation Initiative (SII) and many DoD organizations across the US are working to enable sensing of the incumbents like weather or military radar in the spectrum to allow the co-existence with commercial use. This also led to a multi-million dollar funded program by IARPA (SCISRS). Looking towards the future, we are working on designing networks that can react to the feedback, starting with cellular and then with decentralized WiFi networks. Based on a theoretical study, the expectation is that it would improve the spectral efficiency by 5000x. Sensing becomes fundamental to mm-wave frequency and sub-THz frequency as discovering the link itself is challenging with highly directional beams. My efforts are on generalizing this approach across mm-wave/THz to propose the design of low-cost, mobile spectrum sensors that can quickly gather global spectrum observations across the four dimensions and enable network optimization.