Multi-agent learning in extensive games with complete information
AAMAS '03 Proceedings of the second international joint conference on Autonomous agents and multiagent systems
Planning and Acting in Partially Observable Stochastic Domains
Planning and Acting in Partially Observable Stochastic Domains
Distributed Implementations of Vickrey-Clarke-Groves Mechanisms
AAMAS '04 Proceedings of the Third International Joint Conference on Autonomous Agents and Multiagent Systems - Volume 1
ABLE: a toolkit for building multiagent autonomic systems
IBM Systems Journal
| Hi-index | 0.00 |
In a large-scale dynamic system with multiple distributed entities, each with their own set of interests, there is a need to find a globally acceptable and optimal solution state. This solution state is, by definition, efficient to all entities with respect to their own individual goals and to the system as a whole. In these dynamic environments, this solution state can be achieved by utilizing software techniques from the field of game theory in order to make optimal decisions. We present an application built upon a generalized optimization framework that can be applied to a number of domains, such as cargo or network traffic algorithms. In this research, we used a market-based approach to air traffic flow management through a modeling and simulation environment. The aim is to allow individual aircraft a certain degree of local autonomy, much like cars on a highway. Our system is able to cope in real time with failures such as node loss and adjust system parameters accordingly to optimize results based on the goals of the involved agents. We describe tradeoffs between different agent interaction frameworks with respect to their performance in market mechanism auctions. We also discuss lessons learned while implementing this application. This research has built upon our previously reported work [20, 21] on route optimizations and airspace sector design in an air traffic control network, by adding in the goals of interested entities, e.g. airlines, aircraft, and airports, maximizing the “payoff” to each player (agent). It is intended that the results of our work will be directly used in this domain. In addition, we envision our work being leveraged for other optimization tasks such as data traffic on a network, first responder / disaster relief efforts, and other tasks where rapid solving of large-scale optimization problems is essential.