Abstract

Planning optimal and, in some cases, even approximately optimal collision-free paths for robots is NP-hard. Yet, one must often find high-quality collision-free paths for them in real time. Artificial intelligence has made good progress on increasing the efficiency of such multi-agent pathfinding algorithms over the past ten years.

In this high-level talk, I’ll use multi-agent pathfinding as an example to explain current efforts in the United States to further the collaboration of AI and OR researchers. I’ll also explain progress toward opening up new applications of multi-agent pathfinding algorithms that go beyond the automated warehouse applications that are often used to motivate research on multi-agent pathfinding.