Agent-Based Real-Time Computing and Its Applications in Traffic Detection and Management Systems

Abstract

This paper studies the generic architecture and coordination mechanism of an FIPA (Foundation for Intelligent Physical Agents) compliant heterogeneous multi-agent system (MAS) in a distributed control scenario. Proposed generic MAS framework has been applied to a real-time traffic detection and management system. The agent-based real-time traffic detection and management system (ABRTTDMS) is designed for the information fusion of different traffic detection systems on highways to optimize real-time traffic detection and management. ABRTTDMS is a multi-level agent system. The lowest level agents are responsible for processing the real-time traffic data and monitoring traffic conditions. The middle level agents coordinate all of the lowest level agents in a subnetwork. The highest-level agent is located in the transportation management center (TMC) to accept human commands and delegate tasks to lower level agents. The re-configurable hybrid control architecture of the system provides the capability of dynamically grouping distributed heterogeneous agents (intelligent control systems) into virtual clusters to accomplish tasks related to different geographical areas and different purposes. A platform independent agent execution engine has been implemented using an embeddable C/C++ interpreter called Ch. The communication between agents that reside on different hosts has been simulated in the Integration Engineering Laboratory at the University of California, Davis. The use of agent technology greatly enhances the distributed computing and cooperation capabilities of traffic detection and management systems.