CIF: Small: Statistical Learning Methods for Communications, Sensing and Control in Actuated Wireless Networks

点击翻译按钮获取中文摘要

As the development of unmanned aerial vehicles (UAVs) or drones advances, there is potential for such aerial vehicles to be used to enhance and extend the cellular wireless communication infrastructure. As humans, we already engage in decisions about communication and actuation when using wireless communications. If we feel our connection is bad, we move in the hopes of experiencing a better wireless channel between our cellular phone and a base station. Wireless communications suffer from several deleterious propagation effects such as random fluctuations in signal strength, signal blockage due to obstructions and multipath effects which are due to signals bouncing off of buildings and other large objects. A mobile base station could serve to mitigate these effects by moving where additional coverage is needed, including the tracking of mobile hotspots (where there is extra traffic or congestion). To achieve this UAV enhanced communications infrastructure, this year-long project seeks to answer some key questions about cooperative and decentralized estimation as well as localization. In this exploratory project, a collection of UAVs is envisioned to work in concert. To this end, both centralized and distributed methods for estimation among the UAVs will be considered. Building on prior work on observation driven scheduling, the case of unknown underlying distributions of the sensed signal will be considered. In particular, data driven methods which enable the implicit learning of those distributions in a decentralized fashion will be developed. Observation driven scheduling will also be fully decentralized and different consensus-based strategies will be theoretically analyzed in terms of their performance. Tensor factorization-based non-parametric localization methods will be made more practical. In particular, a method for determining the number of targets in a field to be localized will be developed; prior work assumed that this number was known. Channel estimation based on feature maps and the non-parametric tensor factorization methods will be developed. A performance analysis of feature map-based methods will conducted.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.