CAREER: Distributed Inference-Making via Crowdsensing

点击翻译按钮获取中文摘要

Crowdsensing systems allow human crowd participants carrying smart devices to contribute sensing measurements from built-in sensors in their devices toward a distributed inference-making task. Some crowdsensing systems also employ human crowd participants as humans-as-sensors where humans themselves observe a phenomenon and contribute subjective inferences obtained. The human-powered nature of crowdsensing makes the performance of such systems to become dependent, in addition to aspects of signal processing for extracting information about a phenomenon from observations made, on factors pertinent to human nature. This makes optimization of crowdsensing performance to require jointly addressing signal processing and human aspects, which this project aims to address by converging interdisciplinary perspectives. The project will significantly advance human-powered societal-scale distributed sensing technologies that can sustain smarter, safer, and more resilient communities as well as will advance human-in-the-loop signal processing capabilities. The project also aims to significantly enhance student engagement, learning, recruitment, and retention through an elaborate plan on integrating research and educational activities as well as will enhance participation of underrepresented populations, including women and minorities. The project develops models, analytical approaches, and optimization techniques that address signal processing aspects jointly with traits pertinent to the nature of human crowd participants to lay the foundations of crowdsensing-based distributed inference-making systems. Specifically, the project aims to: 1) produce novel game theoretic market-based crowdsensing mechanisms that jointly address signal processing and selfishness aspects to enable optimal information acquisition from selfish human crowd participants, under factors such as participatory cost uncertainties, resource constraints, privacy concerns, and dependency structures, in a competitive market environment while providing optimal (monetary or nonmonetary) incentives to induce their desired participation; 2) produce prospect theoretic models and methods for optimally employing cognitively biased human crowd participants for distributed inference-making tasks; 3) analytically unravel the impact of attacks from malicious human crowd participants, who can challenge the integrity of the contributed data, and develop mitigation techniques; and, 4) produce a crowdsensing testbed that supports a variety of smart devices and applications for analysis and performance evaluation of crowdsensing techniques under real-world operating conditions.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.