A key feature of the Autonomous Agents and Multiagent Systems (AAMAS) conference is its emphasis on ties to real-world applications. This emphasis of trying to marry theory and practice at AAMAS goes all the way back to the origins of its predecessor conferences, such as the first International Conference on Autonomous Agents (Johnson 1997). However, the effort to tie research into practical applications got a significant boost with the establishment of the industry track at AAMAS, which was more recently renamed as the innovative applications track.
Over the past few years, within this industry and innovative applications track at AAMAS and other related tracks at sister conferences including Innovative Applications of Al (IAAI), there have been presentations of several successful transitions of key component technologies of agents and multiagent systems. On the one hand, individual agents integrate multiple components and capabilities, for example, planning, learning, reactivity, goal orientedness, and they act autonomously while being situated in their environment--thus facilitating their application in real-world settings. On the other hand, multiagent systems and techniques focused on reasoning about multiple agents reflect the fact that there exist many autonomous agents (cooperative or self-interested) in the real world, and capturing their interaction establishes higher veracity of the model. This appropriateness of agent and multiagent systems to model complex real-world problems has led to successful transitions of practically applied technologies ranging from belief desire intention (BDI) frameworks, to game-theoretic approaches, to auction frameworks, to biologically inspired approaches. These previously successful applications have been reviewed in the literature and textbooks on multiagent systems (Wooldridge 2009, Shoham and Leyton-Brown 2008).
This article focuses on the more recent efforts to marry research with practical applications that have been reported at AAMAS over the past two years. Specifically, focusing on papers from AAMAS 2010 and AAMAS 2011, we will discuss the three broad areas that have been the focus of the transition from research into practice: security, sustainability, and safety. With respect to security, research at AAMAS has emphasized the use of game-theoretic techniques to schedule limited security resources to protect targets of economic and political importance. For example, ARMOR (Pita et al. 2008; Jain et al. 2010b) schedules checkpoints and canine patrols at the Los Angeles International Airport whereas IRIS (Tsai et al. 2009; Jain et al. 2010b) schedules federal air marshals on board international flights of U.S. air carriers. More game-theoretic scheduling assistants are being designed for other security agencies as well, such as GUARDS (Pita et al. 2011) for scheduling activities conducted by the Transportation Security Administration. GUARDS is being evaluated at an undisclosed airport for potential nationwide deployment. Finally, PROTECT (An et al. 2011a) is in use for scheduling the patrols of the U.S. Coast Guard in the port of Boston and beyond.
Multiagent systems have also been applied to research on the sustainable use of energy resources (Chalkiadakis et al. 2011; Kamboj, Kempton, and Decker 2011; Kok 2010). Sustainable production, delivery, and use of energy is an important challenge of today. One of the ways this can be done is by developing intelligent systems, like smart grids (Vytelingum et al. 2010b; Ramchurn et al. 2011), that can efficiently predict the use of energy and dynamically optimize its delivery. The distributed nature of the energy grid and the individual interests of users make multiagent modeling an appropriate approach for this problem. Multiagent research in this area has primarily focused on developing techniques based on game-theoretic approaches (including coalitional game theory) and auctions that help reduce the usage and wastage of energy (Vytelingum et al. …