This project develops, from the ground up, a new theoretical framework for analyzing and designing algorithms for dynamic ad-hoc wireless networks. This proposal embraces network dynamics as an opportunity to be exploited, not an adversity to be overcome.
The approach is based on four inter-related thrusts: 1. Incremental Topology Learning: Tracking changes in the network much more efficiently than re-learning entire topology, using sparse "error graph" representations. 2. Topology and Traffic Shaping: Controlling the "effective" wireless network topology so that (i) at any instant of time it appears to be highly disconnected to scheduling algorithms, but retains global connectivity over time; and (ii) modifying traffic statistics to ensure statistical spatial correlation decay. 3. Warm-starting Distributed Algorithms: Message-passing algorithms that can warm-start the optimization based on local knowledge of past solutions. 4. Proteus - A Mobile Robot Testbed: This project validates its approach via implementation on a mobile robot testbed called Proteus, which is used to optimize algorithms in a practical setting.
Broader Impact: Industry is involved in this research from the start, via the WNCG Affiliates program at UT. The research will be disseminated via publications in top-tier venues, industry interactions, and specially organized workshops. Both graduate students and undergraduate students, via a REU program at UT (with emphasis on recruiting women and minorities), get exposure to both real-world wireless networks (via the testbed), and cutting edge theory.
