Smart, but do they work together?
A collaborative architecture for intelligent agents is the key
Stovepipes old “legacy” software you’ve bought and can’t easily replace that can’t work together are one of the biggest obstacles to getting the most out of information technology. This is especially true in Naval operations. Sailors and Marines use many systems that have been acquired over the years to changing requirements and different specifications.
Planning an operation lends itself to automation, because of the large amount of raw information that goes into sound decision-making, but unfortunately the automated systems that handle different aspects of the problem tend to have difficulty working together.
Intelligent agents can broker information among such systems, and take on a lot of the grunt work that used to burden human operators.
This is a new way of exploiting information technology for operational missions. Teams of intelligent agents that respond autonomously to commanders and staffs, and that perform delegated tasks on behalf of users will revolutionize joint mission planning and execution. ONR -supported research on a reusable multi-agent infrastructure (the “RETSINA agent architecture,” for “Reusable Environment for Task-Structured Intelligent Networked Agents”) has led to a basic architecture for agent systems. (And no, it has nothing to do with an homage to the Greek national drink.)
This work has shown that any agent system must have four types of agents: interface agents, information/sensor agents, task agents, and middle agents. This agent architecture has led to the Defense Advanced Research Projects Agency (DARPA)”s Agent Markup Language (DAML) used by DARPA CoABS GRID and Sandia Labs. RETSINA has been further enhanced to include agents that adaptively form heterogeneous teams on demand to gather information and collaboratively plan. Northrop Grumman, General Motors, PNC Bank, Nokia, and Samsung have adapted these agents.
Most recently, ONR-supported researchers have achieved interoperability between two major agent communities: SRI’s Open Agent Architecture (OAA), and CMU’s RETSINA. (OAA agents speak a Prolog-based language, while RETSINA agents use KQML.)
Researchers are now developing methods for mixed initiative collaborations among humans and agents, generalization of interoperability among any agent communities, and design of scalable agent system architectures. These are critical issues in the development of agent technology the key to realizing FORCEnet (a new initiative that stresses the need to network command and control, weapon, and sensor systems among ships, aircraft, and ground units), pervasive computing, and distributed systems, where massive amounts of information from sensors and databases must be collected, processed, integrated, and disseminated to the right users.
For more information on RETSINA if you are media, please contact John Petrik at 703-696-5031, or email onrpao@onr.navy.mil.
