Holistic Framework for Multi-Drone Data Collection in IoT Systems

Rotary-type unmanned aerial vehicles (commonly referred to as drones) are increasingly deployed as mobile data collectors for large-scale Internet-of-Things (IoT) sensor networks. However, wireless transmission rates are inherently stochastic, causing drones to adapt their trajectories during deployment. This adaptation alters flight time and energy consumption, creating a structural mismatch between deterministic offline route optimization and real-world execution under tight battery constraints. As a result, routes that appear feasible offline may become infeasible during deployment. Existing multi-drone IoT routing methods assume deterministic communication models and static energy budgets, leading to systematic infeasibility during deployment. In this article, we present a closed-loop planning framework that empirically mitigates energy exhaustion under stochastic communication. Our architecture first computes energy-constrained routes using a global offline planner that jointly optimizes drone-to-sensor assignments, visit order, and service waypoint locations. It then evaluates route feasibility using Monte Carlo simulation with an online adaptive planner that adjusts trajectories in response to real-time networking performance. Execution statistics are fed back to iteratively calibrate planning energy allocation, yielding robust routes. We further incorporate an empirically derived travel-to-transmission tradeoff that captures how optimal drone-sensor distance depends on drone speed and data volume, and exploit this tradeoff through a polynomial-time routing and convex waypoint refinement algorithm. Extensive numerical experiments and a proof-of-concept implementation on a physical drone and wireless sensor testbed demonstrate that the proposed framework significantly reduces deployment failures while maintaining low data latency compared to fixed-margin planning approaches.