Combinatorial Auction-based Multiple Dynamic Mission Assignment Jin-Hee Cho, Ananthram Swami, and Trevor Cook U.S. Army Research Laboratory Computational and Information Sciences Directorate Adelphi, MD 20783 {jinhee.cho, ananthram.swami, trevor.j.cook}.civ@mail.mil ! " # $ #$ % $ $ #$ & ’ ’ I. INTRODUCTION Tactical networks often have assigned for execution by multiple members where the missions arrive at different times and have to be completed within given, possibly different deadlines. For example, in the theatre of military operations, multiple missions may dynamically arise that require unmanned aerial vehicles (UAVs) or (static/mobile) sensors or other assets, and since they are a constrained resource, we must decide how best to allocate these assets to the missions to best support the army’s goals. Furthermore, we must perform the allocations in an efficient way, so as to reliably deliver high quality services with minimum overhead. This work proposes a multiple mission assignment mechanism based on combinatorial auction theory for tactical mobile ad hoc networks (MANETs). Each node will execute multiple missions during its lifetime and each mission may be performed by multiple members. In this tactical network, each node has a goal of maximizing its utilization by reducing idle time. On the other hand, in terms of the mission leaders’ perspective, missions should be optimally assigned to maximize successful mission completion. The task assignment problem has been well studied in the areas of UAVs, robot-to-robot networks, and sensor-mission (or task) matching in wireless sensor networks (WSNs) [1], [6], [7], [10]. A market-based approach to model dynamic, distributed resource allocation problems was proposed in [11]. The underlying idea is that market-based mechanisms are able to facilitate decentralized resource allocation in terms of complex tradeoffs of goods and services in sensor networks just as in human societies [11]. Mainland et al. [10] used an auction theory to propose a decentralized resource allocation algorithm in sensor networks. Choi et al. [6] proposed consensus-based algorithms to coordinate a fleet of mobile autonomous robots. Ahmed et al. [1] proposed a dynamic auctioning scheme for UAV search and rescue mission. However, all these works [1], [6], [10] only considered static missions (i.e., all missions are given, with the same timeframe of arrival and completion) and each node performs only one mission or task during its entire lifetime. In our formulation, a node may perform multiple missions. Dynamic auction-based resource allocation schemes are also used in cloud computing [2], [8]. Lin et al. [8] proposed a dynamic auction mechanism maximizing profit of the cloud service provider via efficient allocation of its computational resources. An et al. [2] studied resource allocation algorithms to maximize the utility of service providers in cloud computing. Both works assume a static setting of service providers and customers. However, we consider the dynamics of node failure or disconnection/reconnection in this work. The performance of auction algorithms has been investigated via metrics such as delay or network lifetime [12], [13]. Park et al. [12] used a delay metric that combines the overall computation time plus coordination delay. Tei et al. [13] proposed an auction-based route allocation approach to maximize network lifetime in terms of energy consumption in MANETs. Different from [12], [13], our work uses the metrics of communication overhead, a node’s resource utilization (i.e., busy time) and the fraction of completed missions in order to reflect relevant performance aspects of military network environments. Recently market-based approaches were proposed [7] to model dynamic missions or a mission-asset matching problem, which seems similar to this work. Johnson et al. [7] studied a multiple mission assignment problem for wireless stationary sensor networks where missions are modeled as static (i.e., all missions start and finish at the same time) or dynamic (i.e., missions arrive and should be completed by different times). For static missions, they also used combinatorial auction theory, modeling sensors as bid items and missions as bidders. In the The 2011 Military Communications Conference - Track 3 - Cyber Security and Network Operations U.S. Government work not protected by U.S. copyright 1327