Hybrid Belief Reinforcement Learning for Efficient Coordinated Spatial Exploration
The Hybrid Belief Reinforcement Learning (HBRL) framework coordinates multiple autonomous agents to explore and serve areas with unknown demand by jointly learning spatial patterns and planning trajectories. It uses a Log-Gaussian Cox Process for spatial modeling and Pathwise Mutual Information for informed exploration paths. The system combines belief state transfer with demonstration seeding to accelerate learning, achieving better performance in wireless service provisioning.