Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

One Rule to Bring Them All: Investigating Transport Connectivity in Public Transport Route Generation for Equitable Access

Aleksandr Morozov, Ruslan Kozliak, Georgii Kontsevik, Sergey Mityagin

TL;DR

This work introduces a transport connectivity-aware accessibility metric that bases optimization on principles of equitable accessibility rather than traditional trade-offs between passenger and operator costs, and shows a noticeable improvement in network resilience.

Abstract

Designing a city-wide public transport network poses a dual challenge: achieving computational efficiency while ensuring spatial equity for different population groups. We investigate whether AI-based optimization hybrid neuroevolutionary methods combining graph neural networks with evolutionary algorithms - can scale Transit Network Design Problem (TNDP) solutions from synthetic tests to real urban networks while preserving social fairness. Our contribution is to introduce a transport connectivity-aware accessibility metric that bases optimization on principles of equitable accessibility rather than traditional trade-offs between passenger and operator costs. The results show a noticeable improvement in network resilience by improving algebraic connectivity on synthetic datasets, and highlight the ambiguity of applying network generation to real data.

One Rule to Bring Them All: Investigating Transport Connectivity in Public Transport Route Generation for Equitable Access

TL;DR

This work introduces a transport connectivity-aware accessibility metric that bases optimization on principles of equitable accessibility rather than traditional trade-offs between passenger and operator costs, and shows a noticeable improvement in network resilience.

Abstract

Designing a city-wide public transport network poses a dual challenge: achieving computational efficiency while ensuring spatial equity for different population groups. We investigate whether AI-based optimization hybrid neuroevolutionary methods combining graph neural networks with evolutionary algorithms - can scale Transit Network Design Problem (TNDP) solutions from synthetic tests to real urban networks while preserving social fairness. Our contribution is to introduce a transport connectivity-aware accessibility metric that bases optimization on principles of equitable accessibility rather than traditional trade-offs between passenger and operator costs. The results show a noticeable improvement in network resilience by improving algebraic connectivity on synthetic datasets, and highlight the ambiguity of applying network generation to real data.
Paper Structure (15 sections, 14 equations, 3 figures, 3 tables)

This paper contains 15 sections, 14 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Problem Formulation
  5. Neuro-evolutionary algorithm
  6. Transport connectivity
  7. Experiments and Results
  8. Data
  9. Assessment
  10. Experiments on Benchmark Data
  11. Comparative Analysis on Pareto Front and Accessibility Metrics
  12. Routes visualization
  13. Experiments on City Dataset
  14. Discussion and Conclusions
  15. Acknowledgments

Figures (3)

  • Figure 1: Pareto fronts of passenger cost $C_p$ and operator cost $C_o$ across benchmark instances Mandl and Mumford0--3. The rhomb marks our solution for $w>0$.
  • Figure 2: Transit route sets in the Mandl benchmark optimized under three objectives: (left) connectivity, (center) operator cost, and (right) passenger convenience. The operator-perspective solution (center) uniquely forms a spanning tree of the full Mandl graph, containing exactly 14 edges, full vertex coverage, and guaranteed connectivity without redundancy.
  • Figure 3: Visualization of results on the Tartu network: (a) Preprocessed urban blocks with stops, (b) OD flows (visualized using Flowmap City* platform), (c) Generated routes on road network for configuration $(0,0,1)$.