CPS:Small:Collaborative Research: Incentivizing Desirable User Behavior in a Class of CPS
Lead PI:
Mingyan Liu
Abstract
In Cyberphysical Systems (CPS) such as large scale infrastructure systems, individual users are shifting from being passive consumers of services to active participants. This shift promises societal, economic, and environmental benefits. For example, turning consumers into "prosumers" through distributed renewable energy integration can improve sustainability, and turning users into sources of data about traffic and road conditions can help alleviate congestion. However, an explosion of decision makers also leads to heterogeneity in concerns, aims, and quality of decision making. Participants will optimize their decisions on data collection, data sharing, and actuation based on their own self-interest. In an interconnected system, these decisions will affect the system state, and hence, the utilities other participants derive from the system. In extreme cases, undesirable and unintended consequences such as instability can also ensue from actions of a few decision makers. Thus, there is an urgent need to understand and modulate decision making by various participants so that they internalize the impact on system performance and utilities of the other participants that their decisions produce; the proposed research solves basic problems in this domain. The proposed research has potentially substantial social and economic impact. The general trend in CPS systems, such as infrastructure systems, is to allow users to become active participants. However, this will challenge the traditional algorithms and organization for such systems. Since most CPS systems provide basic services to the population, designs that guarantee reliable operation in spite of strategic behavior of self-interested participants are needed. The research is coupled with an education plan that includes integration of research and education, curriculum design, student advising and training, as well as outreach. Existing approaches such as cooperative control or competitive game theory in which participants are price takers are insufficient to consider strategic behavior by participants. Motivated by this observation, this proposal will advance the science of CPS by (i) abstracting this problem into one of designing contracts by which the decisions of strategic users, who may anticipate the effect of their actions and misreport any information, can be aligned to obtain desired system performance, (ii) presenting new mechanisms for this model under several assumptions inspired by large scale CPS such as smart infrastructure systems, and (iii) demonstrating the efficacy of this framework by applying it to several CPS case studies. The proposed approach couples optimization and systems theory with economics and mechanism design as well as models and constraints from specific infrastructure systems.
Performance Period: 10/01/2017 - 09/30/2020
Institution: University of Michigan Ann Arbor
Sponsor: National Science Foundation
Award Number: 1739517