EN
Back to the archive

The encyclopedia · Software & IT · Technical decision · 2020–2024

Washington DC's paratransit cut operating cost ~15% with nested decomposition

MetroAccess paratransit modernized its scheduling engine and folded in accessible taxis, cutting operating costs around 15% for door-to-door riders.

MetroAccess (Washington D.C. paratransit) · IT Curves

The solution

Paratransit provides door-to-door shared rides for riders with travel-limiting disabilities, but the service loses money and the software that schedules it has weak optimization engines. Operators also restricted themselves to their own fleet, ignoring cheaper alternatives.

A university–industry collaboration with software maker IT Curves rebuilt the engine: a nested decomposition that embeds column generation within a temporal decomposition framework handles large ride networks in overlapping time windows. The team also integrated alternative service types, such as accessible taxis, into the scheduling decision, and added a reoptimization procedure for changes that arrive during the day.

Deployed in paratransit operations in the Washington D.C. metro area, the new optimization relative to the legacy system produced significant improvements in key operational metrics and roughly 15% operating cost reductions.

Why it worked

  • Column generation inside a temporal decomposition makes ride networks that used to be unsolvable tractable, so more rides get matched efficiently.
  • Allowing accessible taxis to compete for trips turns a hidden pool of capacity into cheaper service without reducing quality of ride.
  • The reoptimization procedure keeps the plan good when rides change mid-day, where legacy systems just degraded.
What it achievedSplit scheduling in time; taxis join the poolclever

What can be applied

Two modest changes — a stronger solver and a broader vehicle pool — can beat one heroic one: letting cheaper trip types compete inside the model is where transit savings hide.

Aftermath

Published in the INFORMS Journal on Applied Analytics in 2025. The new scheduling engine is deployed in the Washington D.C. metro area with documented operating cost reductions around 15%, and the authors frame the nested decomposition as transferable to other paratransit and demand-responsive transit systems.

Sources

spotted an error? The archive wants to know.

Related cases