Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Liner Shipping Network Design with Reinforcement Learning

Utsav Dutta, Yifan Lin, Zhaoyang Larry Jin

TL;DR

This paper proposes a novel reinforcement learning framework to address the Liner Shipping Network Design Problem (LSNDP), a challenging combinatorial optimization problem focused on designing cost-efficient maritime shipping routes to produce competitive results on the publicly available LINERLIB benchmark.

Abstract

This paper proposes a novel reinforcement learning framework to address the Liner Shipping Network Design Problem (LSNDP), a challenging combinatorial optimization problem focused on designing cost-efficient maritime shipping routes. Traditional methods for solving the LSNDP typically involve decomposing the problem into sub-problems, such as network design and multi-commodity flow, which are then tackled using approximate heuristics or large neighborhood search (LNS) techniques. In contrast, our approach employs a model-free reinforcement learning algorithm on the network design, integrated with a heuristic-based multi-commodity flow solver, to produce competitive results on the publicly available LINERLIB benchmark. Additionally, our method also demonstrates generalization capabilities by producing competitive solutions on the benchmark instances after training on perturbed instances.

Liner Shipping Network Design with Reinforcement Learning

TL;DR

This paper proposes a novel reinforcement learning framework to address the Liner Shipping Network Design Problem (LSNDP), a challenging combinatorial optimization problem focused on designing cost-efficient maritime shipping routes to produce competitive results on the publicly available LINERLIB benchmark.

Abstract

This paper proposes a novel reinforcement learning framework to address the Liner Shipping Network Design Problem (LSNDP), a challenging combinatorial optimization problem focused on designing cost-efficient maritime shipping routes. Traditional methods for solving the LSNDP typically involve decomposing the problem into sub-problems, such as network design and multi-commodity flow, which are then tackled using approximate heuristics or large neighborhood search (LNS) techniques. In contrast, our approach employs a model-free reinforcement learning algorithm on the network design, integrated with a heuristic-based multi-commodity flow solver, to produce competitive results on the publicly available LINERLIB benchmark. Additionally, our method also demonstrates generalization capabilities by producing competitive solutions on the benchmark instances after training on perturbed instances.

Paper Structure

This paper contains 24 sections, 29 equations, 11 figures, 9 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. LSNDP
  4. Policy Neural Network Design
  5. Encoder-Only Approach with One-Shot Rollout
  6. Encoder-Decoder Approach with Autoregressive Rollout
  7. Policy Optimization
  8. Experiments
  9. Result on Baltic Instance
  10. Experiments on Other Instances
  11. Solve Time Comparison
  12. Solution Robustness against Variations in the Problem Instance
  13. Discussion
  14. Conclusion and Future Work
  15. LSNDP details
  16. ...and 9 more sections

Figures (11)

  • Figure 1: Example of a butterfly service with a hub port at London (GBLON).
  • Figure 2: Encoder policy diagram for NDP.
  • Figure 3: LSTM-based decoder for NDP. The decoder takes in $\mathbf{x}_{t}$ generated from the previous steps and builds $\underline{\mathbf{x}}$ sequentially. The example shows how a full service $A_t = (v_2,p_1,p_3)$ is generated sequentially over $\tau=1,2,3,4$ within step $t$. Note that at $\tau=4$, $p_1$ is selected again, which closes the circle and ends $A_t$.
  • Figure 4: Expanded graph representation for an edge connecting ports $p$ and $q$. The expanded representation on the right side of the figure with proxy nodes fully captures all dynamics of commodity shipping.
  • Figure 5: MDP representation of the LSNDP
  • ...and 6 more figures