Improving Bus Service Reliability with Stochastic Optimization

Bus route performance, typically expressed in terms of excess waiting time (EWT), is often unstable and suffers from bunching which results in lower regularity. Governments are introducing regularity-based contracts for operators, where monetary incentives or penalties are introduced depending on performance. Optimization of bus regularity requires the recursive coordination of several buses, hindering the solution scalability in real-time. In this work we propose a bus headway balancer based on stochastic search and branch hopping/merging algorithm which optimizes schedules to minimize the EWT. This algorithm balances bus headway deviations by introducing dwell intervals in the schedule, which can be applied off line or in real-time. A test-case implementation of the approach used 3-month AVL data from a bus operator in Asia and showcased an improvement of EWT by up to 50% with reduction in computational complexity to almost linear time and at least 2x times increase at solution space search.