New DP algorithm predicts worst-case bus bunching in transit systems

Vehicle bunching is a persistent problem in public transit: when buses cluster together, the lead bus handles most passengers, while trailing buses run nearly empty, wasting fuel and driver time. After the last bus in a bunch passes, the gap (headway) until the next bus grows, causing long waits. Transit agencies combat bunching by holding buses at stops, but existing research focuses only on average-case improvements, ignoring the extreme headway scenarios that cause rider frustration and operational inefficiencies.

Now, a team led by Michael Yuhas at Vanderbilt University has published a dynamic programming method that computes the absolute minimum and maximum headway times for any bus route—even with arbitrary control points, vehicle counts, and holding policies. Their bounds are “tight,” meaning they can identify the exact sequence of events leading to worst-case bunching. When tested on Nashville’s transit system (11 pages, 9 figures), the algorithm reveals how different holding strategies and stop placements affect worst-case performance. This gives planners a concrete tool to redesign routes and schedules to reduce both bunching and wait times without sacrificing rider experience.