Scheduling Aerial Vehicles in an Urban Air Mobility Scheme

TL;DR

The paper addresses scheduling autonomous aerial vehicles for urban mobility by aiming to maximize served customer requests while minimizing energy use through flight-level control and collision avoidance. It introduces an energy-aware optimization framework with a fully connected station graph, discrete time slots, charging rules, and edge-intersection constraints to prevent mid-air conflicts. Two ILP-based approaches are proposed: an Optimal offline algorithm and a scalable Incremental ILP that schedules one AV at a time, achieving near-optimal performance with dramatically reduced computation time. Empirical results show that Incremental ILP reaches around 96% of the optimal task completion with modest energy overhead and provides orders-of-magnitude faster runtimes, enabling practical scheduling over large numbers of locations, vehicles, and tasks.

Abstract

Highly populated cities face several challenges, one of them being the intense traffic congestion. In recent years, the concept of Urban Air Mobility has been put forward by large companies and organizations as a way to address this problem, and this approach has been rapidly gaining ground. This disruptive technology involves aerial vehicles (AVs) for hire than can be utilized by customers to travel between locations within large cities. This concept has the potential to drastically decrease traffic congestion and reduce air pollution, since these vehicles typically use electric motors powered by batteries. This work studies the problem of scheduling the assignment of AVs to customers, having as a goal to maximize the serviced customers and minimize the energy consumption of the AVs by forcing them to fly at the lowest possible altitude. Initially, an Integer Linear Program (ILP) formulation is presented, that is solved offline and optimally, followed by a near-optimal algorithm, that solves the problem incrementally, one AV at a time, to address scalability issues, allowing scheduling in problems involving large numbers of locations, AVs, and customer requests.

Figures (10)

• Figure 1: Task execution vs number of tasks ($5$ drones).
• Figure 2: Energy units used vs number of tasks ($5$ drones).
• Figure 3: Example execution - flight levels.
• Figure 4: Total number of time points for all AVs at each flight level.
• Figure 5: Task execution vs number of drones ($50$ tasks).