On Controlling Randomness in Transit Operations

In many transit systems, vehicles depart their originating stops at fixed intervals, but headways become increasingly irregular as they move along the line. This paper considers how best to control such randomness through a strategy based on holding individual vehicles at a chosen control stop. An algorithm is presented for constructing an approximate optimal dispatching strategy from the control point, where optimal is defined both in terms of average passenger wait at the control stop and average delay for riders already aboard the vehicle. The algorithm is used on actual operations data from a Boston subway line to propose service improvements.