Regulating Self-Adaptive Multi-Agent Systems with Real-Time Interventions

Abstract

Dynamic resource allocation in multi-agent systems has numerous critical applications, especially in solving some very challenging problems related to sustainability. These applications include traffic management in transportation systems, energy management in building systems and demand side management in power grids and water supply network. Since resources in these systems are bounded or limited, agents need to share and compete for the limited resources with peers. The agents are autonomous or self-motivated. They are not directly controlled by the regulatory entities, though they are subject to regulations implemented by the regulators. However, the preferences of the agents are subject to change and not fully available to the regulators due to privacy or other considerations. These systems may not meet societal goals without proper regulations or management. Therefore, it is beneficial to study the principles of interventions in these systems. In this work, we investigate whether, under what conditions and to what extent people are able to create effective interventions to manage these multi-agent systems in real-time. We present an abstract transportation framework named Jiao Tong to model the dynamic resource allocation problems. We then develop a user interface based on Jiao Tong, which enables users to create real-time interventions. We propose six hypotheses with respect to Jiao Tong and then perform and analyze user studies to test these hypotheses.

Date of Award	Jun 2013
Original language	American English
Supervisor	Jacob Crandall (Supervisor)