In the summer 2002 and 2003, the Michigan DOT rebuilt a 13.5-mile section of I-94 in the suburbs of Detroit. To help smooth traffic flow, the DOT deployed a dynamic lane merge (DLM) system. The system was designed to regulate merging movements in transition areas where lane configurations changed from three lanes to two lanes. A series of five DLM trailers equipped with microwave traffic sensors, wireless communications, and "Do Not Pass When Flashing" static signs with flashing beacons were installed upstream from the work zone at 1,500 foot increments. When traffic queues developed on the approach and traffic conditions degraded beyond pre-set threshold limits, the system automatically activated flashers on DLM signs to promote early merging. Once activated, the flashers continued operating for a minimum of five minutes. In addition, a state patrol officer was on site to promote compliance, several static signs showing the message "Do Not Pass" and "Form Two Lanes" were posted on the approach, and a portable dynamic message sign with a large arrow and a "Merge" message was strategically placed in the transition area.

The DLM system was most effective at peak hour traffic volumes of 3,000 to 3,500 vehicles per hour. The study found that if traffic volumes were consistently below or above these levels the system would have limited effectiveness. With low traffic volumes, merging control would most likely not be needed; and with higher traffic volumes, the resulting queues lengths would be so long that the flashers would operate continuously and the automated system would not be needed.

Probe vehicle runs and traffic data were collected by Michigan DOT and Wayne State University. The following impacts were highlighted.

• During the morning peak period the average travel speed increased from approximately 40 mi/h to 46 mi/h. During the afternoon peak period there was no change.
• During the morning peak period, the average number of stops (per probe vehicle run) in the work zone decreased from 1.75 to 0.96. During the afternoon peak period, there was no change.
• During both peak periods there was no change in the stopped time delay (per probe vehicle run).
• During the morning peak period, the average delay decreased from 95 seconds per vehicle to approximately 69 seconds per vehicle for every 10,000 feet of travel.