Improvement of Optimization using Learning Based Models in Mixed Integer Linear Programming Tasks
This addresses computational bottlenecks in MILP for industries like construction and logistics, but it is incremental as it builds on existing warm-starting techniques with learning methods.
The paper tackles the problem of long computational times in Mixed Integer Linear Programming (MILP) for planning and scheduling by using a learning-based framework with Behavior Cloning and Reinforcement Learning on Graph Neural Networks to generate initial solutions, reducing optimization time and variance while maintaining solution quality.
Mixed Integer Linear Programs (MILPs) are essential tools for solving planning and scheduling problems across critical industries such as construction, manufacturing, and logistics. However, their widespread adoption is limited by long computational times, especially in large-scale, real-time scenarios. To address this, we present a learning-based framework that leverages Behavior Cloning (BC) and Reinforcement Learning (RL) to train Graph Neural Networks (GNNs), producing high-quality initial solutions for warm-starting MILP solvers in Multi-Agent Task Allocation and Scheduling Problems. Experimental results demonstrate that our method reduces optimization time and variance compared to traditional techniques while maintaining solution quality and feasibility.