Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Neural-Evolutionary Algorithm for Autonomous Transit Network Design

Andrew Holliday, Gregory Dudek

TL;DR

A novel algorithm for planning networks of routes for autonomous buses is proposed, which outperforms the learned policy alone by up to 20% and a plain evolutionary algorithm approach by up to 53% on realistic benchmark instances.

Abstract

Planning a public transit network is a challenging optimization problem, but essential in order to realize the benefits of autonomous buses. We propose a novel algorithm for planning networks of routes for autonomous buses. We first train a graph neural net model as a policy for constructing route networks, and then use the policy as one of several mutation operators in a evolutionary algorithm. We evaluate this algorithm on a standard set of benchmarks for transit network design, and find that it outperforms the learned policy alone by up to 20% and a plain evolutionary algorithm approach by up to 53% on realistic benchmark instances.

A Neural-Evolutionary Algorithm for Autonomous Transit Network Design

TL;DR

A novel algorithm for planning networks of routes for autonomous buses is proposed, which outperforms the learned policy alone by up to 20% and a plain evolutionary algorithm approach by up to 53% on realistic benchmark instances.

Abstract

Planning a public transit network is a challenging optimization problem, but essential in order to realize the benefits of autonomous buses. We propose a novel algorithm for planning networks of routes for autonomous buses. We first train a graph neural net model as a policy for constructing route networks, and then use the policy as one of several mutation operators in a evolutionary algorithm. We evaluate this algorithm on a standard set of benchmarks for transit network design, and find that it outperforms the learned policy alone by up to 20% and a plain evolutionary algorithm approach by up to 53% on realistic benchmark instances.
Paper Structure (16 sections, 7 equations, 2 figures, 3 tables)

This paper contains 16 sections, 7 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Graph Nets and Reinforcement Learning for Optimization Problems
  4. Optimization of Public Transit
  5. The Transit Network Design Problem
  6. Markov Decision Process Formulation
  7. Cost Function
  8. Methodology
  9. Learned Constructor
  10. Training
  11. Evolutionary Algorithm
  12. Experiments
  13. Results
  14. Trade-offs Between Passenger and Operator Costs
  15. Comparison with Other Methods
  16. ...and 1 more sections

Figures (2)

  • Figure 1: A flowchart of the transit network construction process defined by our MDP. Blue boxes indicate points where the timestep $t$ is incremented and the neural net policy selects an action.
  • Figure 2: Trade-offs achieved by different methods between passenger cost $C_p$ (on the x-axis) and operator cost $C_o$ (on the y-axis), across values of $\alpha$ evenly spaced over the range $[0, 1]$, averaged over 10 random seeds. Both axes have units of minutes. Error bars show one standard deviation over the ten random seeds for each point. We wish to minimize both values, so the lower-left direction in each plot represents improvement. A line links two points if they have adjacent $\alpha$ values, so these curves show a smooth progression from low $C_o$ (at the right) to low $C_p$ (at the left) as $\alpha$ increases.