Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large-scale Urban Facility Location Selection with Knowledge-informed Reinforcement Learning

Hongyuan Su, Yu Zheng, Jingtao Ding, Depeng Jin, Yong Li

TL;DR

This paper distill the essential swap operation from local search, and simulate it by intelligently selecting edges on a graph of urban regions, guided by a knowledge-informed graph neural network, thus sidestepping the need for heavy computation of local search.

Abstract

The facility location problem (FLP) is a classical combinatorial optimization challenge aimed at strategically laying out facilities to maximize their accessibility. In this paper, we propose a reinforcement learning method tailored to solve large-scale urban FLP, capable of producing near-optimal solutions at superfast inference speed. We distill the essential swap operation from local search, and simulate it by intelligently selecting edges on a graph of urban regions, guided by a knowledge-informed graph neural network, thus sidestepping the need for heavy computation of local search. Extensive experiments on four US cities with different geospatial conditions demonstrate that our approach can achieve comparable performance to commercial solvers with less than 5\% accessibility loss, while displaying up to 1000 times speedup. We deploy our model as an online geospatial application at https://huggingface.co/spaces/randommmm/MFLP.

Large-scale Urban Facility Location Selection with Knowledge-informed Reinforcement Learning

TL;DR

This paper distill the essential swap operation from local search, and simulate it by intelligently selecting edges on a graph of urban regions, guided by a knowledge-informed graph neural network, thus sidestepping the need for heavy computation of local search.

Abstract

The facility location problem (FLP) is a classical combinatorial optimization challenge aimed at strategically laying out facilities to maximize their accessibility. In this paper, we propose a reinforcement learning method tailored to solve large-scale urban FLP, capable of producing near-optimal solutions at superfast inference speed. We distill the essential swap operation from local search, and simulate it by intelligently selecting edges on a graph of urban regions, guided by a knowledge-informed graph neural network, thus sidestepping the need for heavy computation of local search. Extensive experiments on four US cities with different geospatial conditions demonstrate that our approach can achieve comparable performance to commercial solvers with less than 5\% accessibility loss, while displaying up to 1000 times speedup. We deploy our model as an online geospatial application at https://huggingface.co/spaces/randommmm/MFLP.
Paper Structure (9 sections, 4 equations, 3 figures, 1 table)

This paper contains 9 sections, 4 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Method
  4. The Universal SWAP operator
  5. Knowledge-Informed RL
  6. Divide-and-Conquer Framework
  7. Experiments
  8. Overall Performance
  9. Conclusion

Figures (3)

  • Figure 1: An example of MFLP. Red pentagons and blue diamonds are the placed two types of facilities. Red and blue edges indicate that residents of each region visit their closest facilities to access basic services. Best viewed in color.
  • Figure 2: The overall RL framework solving the large-scale FLP. (a) FLP as a sequential decision-making problem on the swap graph, where one single facility is moved from its origin region (green) to one of the vacant regions (white) through the SWAP operator at each step. (b) The knowledge-informed RL agent for edge selection leverages the dynamic wiring technique to filter out both Tabu Edges and Negative Benefit Edges. Best viewed in color.
  • Figure 3: FLP performance comparison. For each pair (N,p), we generate 1000 simulated cities and calculate (a) the average relative gap% between each method and Gurobi (optimal). (b) the average speed up compared to Gurobi (optimal).