An ALNS Heuristic for Large-Scale Line Planning with Mode Choice and Line Generation

Demand responsiveness is an important consideration in public transport line planning, as network design and service quality influence passenger demand.

However, accounting for this interaction further complicates an already challenging combinatorial optimization problem.

To address this challenge, we propose a scalable Adaptive Large Neighborhood Search (ALNS) algorithm for large-scale line planning with endogenous demand.

The algorithm jointly optimizes lines and frequencies while accounting for passenger mode choice, passenger assignment, and vehicle capacities.

Candidate lines are generated dynamically throughout the search, and solutions are evaluated using an embedded evaluation procedure for passenger assignment and demand estimation, together with a dedicated local search procedure for frequency optimization.

The proposed methodology is evaluated on the public transport network of Odense, Denmark, comprising approximately 1,800 origin-destination pairs.

Computational results demonstrate the applicability of the approach to realistic, large-scale instances.

The optimized networks concentrate resources on fewer, higher-frequency services, reducing average headways from approximately 41 minutes to 6.8-13 minutes while substantially increasing public transport ridership.

Furthermore, the results show that network design is highly sensitive to assumptions regarding passenger behavior, highlighting the importance of carefully calibrated demand models when incorporating demand responsiveness into line planning.