System architecture for AI-enabled corridor management

Abstract

Intelligent Transportation Systems (ITS) have become increasingly important over time demonstrating a need for smarter algorithms to be deployed in Integrated Corridor Management (ICM). Two Active Traffic Management (ATM) technologies that have shown promise for improving safety on the freeway are Variable Speed Limit (VSL) and Lane Control System (LCS). These devices can display configurations in response to congestion from incidents. Currently, Traffic Management Centers (TMCs) use operators to manage response plans for incidents. These operators could benefit from artificial intelligence algorithms to provide suggestions in forming these response plans. These algorithms would require lots of offline learning through simulation and real-time data. This thesis proposes an architecture by which an artificial intelligence decision support system (AI-DSS) can be constructed to bridge the gap between operators and AI algorithms for VSL and LCS suggestions. The architecture consists of six main components: process management, data management, event management, evaluation management, logging, and a network graph. These internal components can communicate to the TMC software through an API with TCP/IP connections. The architecture shown is then implemented in the Advanced Transportation and Congestion Management Technologies (ATCMTD) project with the Tennessee Department of Transportation. The project occurs on the I-24 Smart Corridor which runs from Nashville to Murfreesboro. The code is written in the Python programming language for ease of integration with rich libraries for artificial intelligence. The AI-DSS software has passed User Acceptance Testing (UAT) with TDOT allowing it to be deployed into production. This system will be the first system to support AI-enabled corridor management in real-time considering a human in the loop.