Pairwise is Not Enough: Hypergraph Neural Networks for Multi-Agent Pathfinding

HGN-MAPF addresses a significant theoretical bottleneck in distributed pathfinding: the 'pairwise limitation' of standard GNNs which fails to capture complex group dynamics in crowded environments. Despite the technical merit, the project scores a 2 on defensibility due to its status as an academic reference implementation with zero stars and minimal community engagement. It lacks a commercial moat or data gravity. The competitive landscape for MAPF is crowded with established baselines like PRIMAL, DHC, and more recent Transformer-based architectures. While frontier labs (OpenAI/Google) are unlikely to target this specific niche algorithm, specialized robotics firms (Amazon Robotics, Ocado) are the primary competitors/consumers. The displacement horizon is relatively short (1-2 years) because MAPF research is high-velocity, and more generalizable architectures (e.g., world-model based RL or specialized Transformers) are likely to supersede specific graph-based heuristics.