Collaborative Planning for UAV-UGV Teams under Operational Constraints via Masked Error-bounded Neuro-Symbolic Diffusion

This research proposes MEND, a neuro-symbolic diffusion framework for collaborative planning in UAV-UGV teams under operational constraints. The method addresses limitations in existing masked diffusion models by incorporating error-bounded inference scheduling and logic-induced admissibility indicators to ensure constraint compliance during execution. Experimental results demonstrate improved planning optimality in simulation and real-world settings.