Optimising LMAPF guidance graphs using Evolutionary algorithms: Advice needed [R]

Hello, I'm currently working on my dissertation and feel like I could really use some advice from someone who looks at the problem with fresh eyes. I appreciate all input. The Problem: Multi Agent Path Finding is the problem of finding paths for several agents to their destinations. Lifelong MAPF is the same, but upon task completion an agent is assigned a new task. For my dissertation (and usually in research) agents move on a grid-like graph