Cooperative Detour Planning for Dual-Task Drone Fleets

A team of researchers has developed a novel decentralized AI framework that enables commercial drone fleets to perform two critical urban functions simultaneously: executing delivery tasks and acting as a mobile sensor network for real-time traffic monitoring. The system, detailed in a paper submitted to the 2026 IEEE Conference on Decision and Control, models the city as a network of road segments with dynamic 'information values' that accumulate uncertainty about traffic conditions over time. When a drone visits a segment, it resets this value by collecting data. The core challenge is formulated as a mixed-integer linear programming problem where each drone must maximize the total traffic information it gathers while strictly adhering to two hard constraints: a maximum allowable detour for its primary delivery mission and its total battery budget.

Unlike computationally heavy centralized control systems that struggle with large fleets, this method employs a 'meet-and-merge' strategy based on dynamic local clustering. Drones operate independently but, when they enter communication range of each other, they exchange their beliefs about traffic status and temporarily form a local cluster. They then transition from isolated planning to a joint optimization mode to resolve their coupled constraints and compute new, efficient paths for each member. Simulation results on the real road network of Barcelona show the framework successfully utilizes available battery and detour budgets to explore the urban area and gather comprehensive traffic data. The decentralized approach achieves coverage quality close to a global optimum but with a fraction of the computational overhead, making it scalable for future large-scale urban air mobility operations.