An Exact Algorithm for Public Transport Line Planning Considering Passenger and Operational Costs and Lost Demand

TL;DR

This paper introduces an exact algorithm for public transport line planning that optimizes passenger satisfaction and operational costs while considering demand levels and capacity constraints.

Contribution

It presents a novel mixed-integer programming model with an iterative solution method that efficiently handles complex, large-scale planning problems.

Findings

The method achieves significant speed-ups over direct CPLEX solutions.

Accounting for lost demand improves overall resource efficiency.

The approach effectively balances passenger service quality and operational costs.

Abstract

Line planning in public transport is the strategic problem of selecting lines and their operating frequencies. This problem is important as it defines the passenger service, based on available connections and expected travel times, and drives operational cost in terms of the number of vehicles required. This paper presents a line planning model that minimizes the weighted sum of passenger travel time, including in-vehicle time and frequency-dependent waiting and transfer times, and operating costs for the public transport agency. Unlike traditional approaches that assume demand to be fixed, our approach requires a minimum service level for demand to be captured, ensuring that services are provided only when they are attractive to users and cost-efficient to operate. The introduced capacity constraints ensure sufficient capacity on the lines and help guide the trade-off between expected demand on selected lines and their frequencies. The resulting mixed-integer program presents a challenging combinatorial problem as the number of passenger paths grows rapidly in relation to the number of lines and frequencies considered. To address this, we propose an iterative exact algorithm that utilizes a reduced problem representation and dynamically expands it with additional frequencies and paths. Evaluated on four networks with varying complexity and cost trade-offs, our method achieves significant speed-ups and tighter bounds compared to solving the complete model directly by CPLEX, particularly when operator and passenger costs are more evenly balanced in the objective. Furthermore, we demonstrate how accounting for lost demand leads to more efficient resource use from an overall perspective.