Dynamically Configured λopt Heuristics for Bus Scheduling.pdfVIP

Dynamically Configured λλ-opt Heuristics for Bus λλ Scheduling Prapa Rattadilok and Raymond S K Kwan School of Computing, University of Leeds, UK {prapa, rsk}@comp.leeds.ac.uk Bus scheduling is a complex combinatorial optimization problem [3], [10]. The op- erations planning and scheduling process starts off with the designing of a timetable of trips that have to be served by buses. Each trip has a starting time/location and a destination time/location. Bus scheduling involves assigning a set of trips to a set of buses such that: 1. The sequence of the trips for each bus is time feasible: no trip precedes an ear- lier trip in the sequence 2. Each trip is served by exactly one bus There are two types of operational cost involved, these are layover (idle time) and dead running (relocating the bus between locations without passenger, this includes leaving and returning to the depot). The objective is to minimise the operational cost and the number of buses. Given the nature of the problem it is almost always impos- sible to reduce the layover and dead running cost to zero. The complexity of the problem quickly increases as the number of the depots and the types of vehicle in- creases. A solution is said to be λ-optimal (or simply λ-opt) if it is impossible to obtain a better solution by replacing any λ relation instances by any other set of λ relation instances. The λ-opt heuristic [5] is based on this concept of λ-optimality where in each trial, λ instances of the chosen relation (mapping between problem components, e.g. bus trips to bus’s timeslots) in the working solution are exchanged. The trial process continues