Multi-Agent Systems / Multi-Robot Systems

Much of my research on multiagent / multirobot systems is specifically related to swarm intelligence, I have additionally conducted research related to other aspects of these systems, including market-based coordination, security engineering of multiagent systems, issues related to mobile agents on ad hoc wireless networks, power aware agents, and multiagent collaboration tools. The remainder of this page focuses on these things, while a separate page of my site focuses on my swarm intelligence research.

Software Engineering of Multiagent Systems

The Philadelphia Area Urban Wireless Network Testbed (PA-UWNT), among other things, developed a distributed and decentralized communications and coordination platform enabling the secure collaboration of security personnel and first responders (e.g., rescue workers, police, firefighters, paramedics) over a mobile ad hoc network (MANET) of a heterogeneous collection of nodes (e.g., tablets, laptops, etc) of limited resources. The article Designing Dependable Agent Systems for Mobile Wireless Networks, in IEEE Intelligent Systems, provides an overview of the project. During this project, we advanced the state-of-the-art in software engineering for multiagent systems operating on MANETs in such challenging urban environments.

We utilized a service-oriented architecture, as detailed in the article Service-Based Computing on MANETs: Enabling Dynamic Interoperability of First Responders in IEEE Intelligent Systems, which utilized an OWL-S web services registry, and mobile agents for service discovery on the MANET. We followed a comprehensive security engineering process in designing the system, as described in the paper Engineering Multi-Agent Systems, from the IEEE First Symposium on Multi-Agent Security and Survivability. And we developed design patterns for mobile swarming agents that traverse the MANET for a variety of tasks, such as network monitoring and service discovery, as presented in Designing Decentralized Software for a Wireless Network Environment: Evaluating Patterns of Mobility for a Mobile Agent Swarm from the Second IEEE Symposium on Multi-Agent Security and Survivability.

Environment-Aware Agents

MANETs are prone to network disruptions, due to a variety of factors that include limited resources on the devices that form the network, nodes that are dynamically in motion going in and out of range creating a dynamic network topology, unknown target environment with potential for structures to cause interference, among others. Thus, the PA-UWNT as discussed in the above section is a challenging one for systems design. For these reasons, the mobile agents of the PA-UWNT utilize environmental awareness. For example, the agents are power aware (see Agent Survivability Through Power Awareness) as well as network-aware (see Network Awareness and the Philadelphia Area Urban Wireless Network Testbed).

Market-Based Multiagent / Multirobot Coordination

The Federation of Intelligent Robotic Explorers (FIRE) project focused on a team of autonomous robots exploring a planetary environment, such as Mars. The team of robots of FIRE utilized a layered architecture, with a market-based approach for task allocation among the members of the team (Goldberg et al, 2003; Goldberg et al, 2002). The robots of the team bid on incoming tasks based on their capabilities, utilizing a simulated market model.

Selected Publications