Oklahoma State UniversityAhmad SalehiyanIndustrial Engineer
← Back to Projects
Reinforcement Learning

Hierarchical Deep Reinforcement Learning for Warehouse Picker Routing

Dynamic programming within aisles, DDQN across aisles; 3.0%–15.4% improvement in travel distance over S-shape and largest-gap heuristics. SSRN preprint.

Hierarchical Deep Reinforcement Learning for Warehouse Picker Routing

Timeline

SSRN preprint

Impact Focus

3.0%–15.4% improvement in total travel distance over S-shape and largest-gap heuristics.

Core Tools

reinforcement learning, DDQN

Problem

Picker routing in two-block warehouses.

What I did

Dynamic programming within aisles; a hierarchical DDQN agent sequences aisles.

Result

3.0%–15.4% improvement in total travel distance over S-shape and largest-gap heuristics.

What changed

  • - Dynamic programming for aisle-level subproblems
  • - SSRN preprint No. 7052096

Deliverables

  • - 3.0%–15.4% improvement in total travel distance over S-shape and largest-gap heuristics.

Stack

reinforcement learning
DDQN
Contact