Wireless spectrum is a precious resource that can only be meaningfully used when access is controlled. Current mechanisms for control are largely legislative, granting organizations and individuals the right to exclude others. Exclusive use of frequency ranges over time and space can be highly inefficient - denying non-interfering, would-be users access despite under-utilization by rights holders. Spectrum collaboration is advanced as an alternative control mechanism in which intelligence on the part of transceiver equipment is used to simultaneously govern access and maximize utilization. Such intelligence is manifested in the form of sophisticated algorithms and supporting flexible radio and computing hardware. This proposal takes a systems approach to spectrum collaboration. The PIs propose new approaches to understand and classify activities within a range of frequencies and to study and evolve channel access protocols that are both proactive (getting data through) and polite (assisting others in the mission of getting their data through). More broadly, the investigators seek to create novel protocols and tools to facilitate spectrum sharing.
The proposed research is built on three pillars: (a) improved spectrum sharing through PHY-layer optimization techniques such as beamforming and interference cancellation, (b) mechanisms to learn the RF environment and spectrum usage patterns using reinforcement learning and signal classification, and (c) defining a model for spectrum sharing, called channel-as-society (CAS). The investigators contend that advances in these three domains have direct applicability to the practice of spectrum collaboration and the DARPA Spectrum Collaboration Challenge (SC2) in particular. The investigators' ongoing joint work on Low-Power Wide-Area Networking (LP-WAN) serves as a testbed for some of the concepts relevant to SC2, both informing and being informed by the proposed research. In addition, the investigators plan to co-teach a Special Topics course in Wireless Networks and Mobile Computing specifically targeting the SC2 problem statement, engaging the students in the creation of novel SC2 concepts and techniques.
