Towards Data Science is a community-powered publication that showcases work in data science, machine learning and artificial intelligence. Every day newcomers, seasoned researchers and industry practitioners publish tutorials, research notes and real-world case studies that help the field move forward.

Towards Data Science

A data scientist shares a two-year project building a hybrid MARL-LP (Multi-Agent Reinforcement Learning + Linear Programming) system for logistics scheduling. The post explains why standard solvers, genetic algorithms, and pure RL approaches were insufficient, then describes a MARL architecture where decentralized hub agents act as fleet managers deciding truck dispatch quantities, while an LP solver handles bin-packing and vehicle selection. Two agent versions are detailed: V1 where agents slice a priority queue, and V2 where agents directly control truck counts per destination. Key design choices include scale-invariant histogram observations for zero-shot generalization across network topologies, and emergent LTL consolidation behavior where agents learned to wait for fuller trucks as shipping costs increased.

A Generalizable MARL-LP Approach for Scheduling in Logistics