ABSTRACT
Abstract
Doc2Control connects natural-language task descriptions with scheduling and control decisions for UAV-assisted campus vehicle systems. It models uncertainty in LLM-derived priorities, maintains minimum service frequencies for control-critical tasks, and jointly allocates scheduling, communication, computing, and battery resources through a relaxed mixed-integer formulation with receding-horizon control. Reported simulations show improvements in semantic-control utility, critical-task service, safety, and energy efficiency under constrained resources.