Digital Twin-Based Task-Driven Resource Management in Intelligent UAV Swarms

Abstract

UAV swarms offer substantial opportunities for Search and Rescue (SAR) applications. Confronted with numerous concurrent sensing tasks in complicated environment, resource-scarce UAV networks need a dynamic, task-driven deployment and resource configuration strategy for multi-UAV swarm coordination to ensure the efficient execution of sensing tasks. This paper introduces a Digital Twin (DT)-based collaboration architecture for resource management in UAV swarms, connecting realistic task crowdsourcing and virtual traffic flow scheduling to achieve a complementary multi-UAV swarm allocation. We propose an intelligent dynamic task crowdsourcing scheme that manages the swarm scale and membership configuration of multiple UAV swarms based on theoretical evaluation results. The architecture constructs DTs of UAV swarms and shifts the scheduling of traffic flow paths to the virtual world, thereby sidestepping the overhead of routing configuration and network reorganisation. With the aid of a traffic flow allocation algorithm based on Stochastic Network Calculus (SNC), the virtual swarm pre-schedules traffic flows and assesses end-to-end delay theoretically, so as to achieve a collaborative deployment of sensing, computational, and communication resources within the swarm. The simulation results substantiate that our architecture can uphold a 90% achievement ratio for task requirements while keeping UAV costs comparable to other algorithms.