Antwerp N177: a semi-flexible feeder service on a real corridor

The earlier work on the semi-flexible feeder service was tested on generated benchmark instances. This case study asks what happens on an actual corridor, using real road-network data and a real incumbent network as the benchmark.

The setting is the N177 between Boom and central Antwerp, a suburban corridor that is poorly served relative to its proximity to the city. The simulated service places 8 mandatory and 56 optional bus stops along it, serves 100 passenger requests over a two-hour window, and guarantees a maximum headway of 20 minutes with a vehicle capacity of 20. Requests were generated on a realistic road network rather than assumed to be uniformly spread, and the results are compared against more than 20 existing De Lijn bus lines serving the same area.

The findings:

• Service quality improved by 31.6% on the global objective, and average user ride time dropped by 22% compared with the existing transit options.
• Acceptance saturates at around 90% once roughly 12 buses are available. Beyond that, extra vehicles add little.
• There is a genuine trade-off in how much freedom the optimizer is given: with only 3 buses a near-perfect solution is achievable for accepted requests, but the global objective suffers. The sweet spot in the degree of optimization freedom is intermediate, not extreme. The same Goldilocks logic that runs through the service design itself.

The instance generation for this study was done with a custom generator built on OpenStreetMap road networks, so the demand and travel times reflect the actual corridor rather than an abstract network.