Connectivity-Influence Weighting: A Structural Bias for Graph-Based MARL
Graph neural networks used for multi-agent reinforcement learning typically treat every edge of the communication graph as equally important. This paper introduces …
Learning-Based Distributed Aerial Shepherding of UGV Swarms
A UAV acts as an aerial shepherd guiding a ground vehicle that leads a UGV swarm, while the UGVs autonomously select local leaders to improve cohesion and reduce communication …