A multiagent framework for collaborative airlift planning using commercial air assets
✍ Scribed by G.A Godfrey; C Hellings; A Knutsen
- Publisher
- Elsevier Science
- Year
- 2004
- Tongue
- English
- Weight
- 947 KB
- Volume
- 39
- Category
- Article
- ISSN
- 0895-7177
No coin nor oath required. For personal study only.
✦ Synopsis
Future operations by the U.S. military services will require greater collaboration within the government and with the private sector. Commercial enterprises that normally compete with one another will have to cooperate to satisfy the goals of the operation. For example, the military uses commercial airlift assets to support the movement of soldiers and cargo to the theater. Ideally, the military would like to receive adequate commercial airlift capacity at a reasonable cost, while the commercial air carriers would like to balance their workload and minimize the disruption to their daily operations. We present a distributed optimization approach that uses software agents--representing the interests of the military and commercial carriers--to collaboratively plan the airlift. By auctioning the missions and allowing carriers to swap missions when mutually beneficial, this approach cuts the controllable operating costs and schedule disruption costs by more than half compared with a centralized planning approach currently used. (~) 2004 Elsevier Ltd. All rights reserved.
Keywords--Strategic airlift, Commercial aviation, Multiagent systems, Game theory, Auctions.
1. BACKGROUND
Future operations by the U.S. military services will requir e intense collaboration within each service, across the services, with other departments and agencies (e.g., the State Department and the CIA), and with its allies. Successful collaboration will also need to occur between the government and the private sector. Within the private sector, enterprises that normally compete with one another will have to cooperate to satisfy the goals of the operation. In fact, it will often be the case that, while the parties agree on the basic goals of the operation, each party may have its own subagenda and operating constraints. Furthermore, all of the collaborating parties may not fully trust each other, and some may be in competition, economic or otherwise.
Consequently, these dynamic resource allocation problems cannot be formulated as a traditional, centralized optimization problem because the players involved have a greater degree of autonomy and potentially conflicting goals. Instead, we consider a distributed optimization approach to what are semiautonomous agents negotiating the division of tasks and resources. The disadvantage of this agent approach, however, is that there is no guarantee that solutions based This research has been supported by the contract F30602-00-C-0175 from the Defense Advanced Research Projects Agency (Taskable Agent Software Kit program). The authors would also like to thank Mr. S.D. Draper for his assistance in developing the military demand data sets used in this paper and Dr. T. Mifflin for introducing the concept and potential benefits of collaborative airlift planning to the authors.