Analyzing and Improving the Performance of Dynamic Message Sign Reporting Work Zone–Related Congestion

Dynamic Message Sign (DMS) systems aim to provide realistic, reliable, and real-time traffic information to roadway users. This study investigates performance of DMS present at work zone which warn drivers of any imminent congestion. A work zone in Davenport, Iowa is used as the test site to evaluate performance and the proposed algorithm improvements. A typical automated DMS used a dedicated sensor measuring speed or occupancy and simple thresholds to post messages warning driver of the congestion using a set of predefined messages including Traffic Delays Possible, Slow Traffic Ahead, and Stopped Traffic Ahead. Based on the field observation, it was found that these simplistic algorithms lead to a significant number of very short messages, erroneous messages during the night time, and groups of messages that continuously alternate between different displays. This paper first develops performance metrics to report the issues with the existing DMS automated programming logic and then proposes a machine learning based real-time algorithm for improved operations.