Neural Routing for Disaster Logistics
Budget: $750 – $1,500 USD
Project Description
Overview:
I am seeking a highly skilled researcher and developer with expertise in Deep Learning and Operations Research to assist in the completion of a Master’s degree project. The research focuses on developing a Nested-Learning Architecture to solve the Multi-Depot Heterogeneous Vehicle Routing Problem with Time Windows (MDHVRPTW) specifically for disaster response scenarios.
The project involves using Neural Combinatorial Optimization (NCO) to create an adaptable routing system that leverages real-world GIS data (OpenStreetMap) and can handle dynamic environment changes (e.g., road closures, varying vehicle types).
Key Responsibilities:
Model Implementation: Refine and implement a Nested-Learning module (Fast Inner Loop / Slow Outer Loop) using Graph Neural Networks (GNN) or Transformers.
Data Integration: Work with GIS data and routing APIs (OSRM/OpenRouteService) to ensure the model operates on real city topologies.
Benchmarking: Test the architecture against standard VRP benchmarks (Cordeau, Solomon) and real-world datasets.
Academic Writing: Assist in drafting technical sections of the thesis, including methodology, experimental results, and comparative analysis.
Optimization: Improve the model’s ability to handle heterogeneous fleets (drones, vans, trucks) and strict time-window constraints.
Required Skills:
Advanced Python: Proficiency in PyTorch or TensorFlow.
Deep Learning: Experience with Graph Neural Networks (GNNs), Attention mechanisms, and Meta-Learning.
Operations Research: Strong understanding of VRP, MDHVRPTW, and Combinatorial Optimization.
GIS Tools: Experience with OSMnx, NetworkX, or similar spatial data libraries.
Academic Excellence: Ability to write at a Master’s/PhD level in English.
Current Project State:
I have already established the core architecture proposal, a master sheet of parameters, and a curated list of relevant literature. You will be building upon this foundation to move from theory to a fully functional, validated simulation.
Deliverables:
Clean, documented Python code for the Nested-Learning model.
Experimental results and performance visualizations (charts/tables).
A comprehensive draft of the Methodology and Results chapters.
Overview:
I am seeking a highly skilled researcher and developer with expertise in Deep Learning and Operations Research to assist in the completion of a Master’s degree project. The research focuses on developing a Nested-Learning Architecture to solve the Multi-Depot Heterogeneous Vehicle Routing Problem with Time Windows (MDHVRPTW) specifically for disaster response scenarios.
The project involves using Neural Combinatorial Optimization (NCO) to create an adaptable routing system that leverages real-world GIS data (OpenStreetMap) and can handle dynamic environment changes (e.g., road closures, varying vehicle types).
Key Responsibilities:
Model Implementation: Refine and implement a Nested-Learning module (Fast Inner Loop / Slow Outer Loop) using Graph Neural Networks (GNN) or Transformers.
Data Integration: Work with GIS data and routing APIs (OSRM/OpenRouteService) to ensure the model operates on real city topologies.
Benchmarking: Test the architecture against standard VRP benchmarks (Cordeau, Solomon) and real-world datasets.
Academic Writing: Assist in drafting technical sections of the thesis, including methodology, experimental results, and comparative analysis.
Optimization: Improve the model’s ability to handle heterogeneous fleets (drones, vans, trucks) and strict time-window constraints.
Required Skills:
Advanced Python: Proficiency in PyTorch or TensorFlow.
Deep Learning: Experience with Graph Neural Networks (GNNs), Attention mechanisms, and Meta-Learning.
Operations Research: Strong understanding of VRP, MDHVRPTW, and Combinatorial Optimization.
GIS Tools: Experience with OSMnx, NetworkX, or similar spatial data libraries.
Academic Excellence: Ability to write at a Master’s/PhD level in English.
Current Project State:
I have already established the core architecture proposal, a master sheet of parameters, and a curated list of relevant literature. You will be building upon this foundation to move from theory to a fully functional, validated simulation.
Deliverables:
Clean, documented Python code for the Nested-Learning model.
Experimental results and performance visualizations (charts/tables).
A comprehensive draft of the Methodology and Results chapters.