WASP: Adaptive Weighted Multi-Objective Scheduling in UAV-Assisted Edge Computing
This paper proposes WASP, a lightweight and training-free online scheduling framework that dynamically balances profit, latency, energy, and safety objectives through adaptive load-based scoring and robust task acceptance policies, demonstrating superior performance in UAV-assisted edge computing environments compared to existing state-of-the-art methods.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the world of modern computing, a quiet revolution is taking place as heavy processing tasks move away from massive, centralized data centers and closer to the devices that generate them. This shift, known as edge computing, allows for faster responses and less strain on distant networks. However, a new layer of complexity has emerged with the introduction of unmanned aerial vehicles, or drones, acting as flying servers. These drones offer incredible flexibility, hovering over disaster zones or remote farms to provide computing power where none exists. Yet, they face a unique and difficult challenge: they run on limited batteries and must make split-second decisions about which tasks to handle, how far to fly, and when to stop, all while balancing the need for speed, the desire for profit, and the strict requirement to keep their batteries from dying mid-flight.
Researchers have long struggled to create a set of rules that allows these drones to juggle these conflicting demands. Traditional methods often focus on just one goal, such as maximizing money earned or minimizing energy used, but this usually causes problems in other areas. A system that chases profit might fly too far and run out of power, while one that focuses only on speed might ignore valuable tasks. To solve this, a team of computer scientists has developed a new scheduling framework called WASP. This system does not rely on complex learning algorithms that require years of training; instead, it uses a smart, adaptive set of rules that change their priorities in real-time based on how busy the system is.
The core idea behind WASP is simple yet effective: the drone's behavior should shift depending on the workload. When the system is quiet and the drones have plenty of battery, the scheduler prioritizes tasks that bring in the most profit. However, as the number of waiting tasks grows and the system becomes crowded, the scheduler automatically shifts its focus. It begins to prioritize speed and proximity, choosing tasks that are closer and can be finished quickly to prevent the queue from backing up. This dynamic adjustment happens without any human intervention or pre-programmed training, allowing the drones to react instantly to changing conditions.
To ensure fairness, the system includes a mechanism to prevent any single task from being ignored forever. If a low-priority task is passed over too many times, the system flags it and forces the drone to handle it, ensuring that no job is left waiting indefinitely. Furthermore, the system is designed with a strict safety margin, always reserving a portion of the drone's battery to ensure it can return home safely and never runs completely dry. This approach creates a balanced environment where economic goals, speed, and safety are all considered simultaneously.
The researchers tested this new method through extensive computer simulations involving multiple drones and ground stations. They compared WASP against seven other existing scheduling methods, including some that are considered the current best in the field. The results showed that WASP outperformed all its competitors. It achieved the highest overall score by successfully completing nearly 95 percent of tasks before their deadlines, while also maintaining a fast average response time of just over 54 seconds. In contrast, other methods that focused heavily on profit often saw their safety rates drop significantly, with some drones failing to complete their missions because they ran out of energy.
The study also revealed that the adaptive nature of WASP was the key to its success. When the researchers tested the system under different conditions, such as varying battery sizes or changing the number of drones, WASP consistently adapted its strategy to maintain high performance. It managed to keep the drones safe and the tasks moving, even when the system was under heavy pressure. The researchers found that removing any single part of their system, such as the ability to change priorities based on load or the rule to prevent starvation, caused the overall performance to drop. This confirmed that the combination of these specific features was necessary to handle the complex reality of flying edge computing.
Ultimately, this work demonstrates that a drone fleet can be managed effectively without needing a supercomputer to learn from past mistakes. By using a clear, transparent set of rules that adjust to the moment, the system achieves a level of balance that previous methods could not. The findings suggest that for safety-critical applications, such as emergency response or public safety monitoring, a deterministic approach that guarantees fairness and safety while optimizing for profit and speed is not only possible but superior to more complex, learning-based alternatives. The researchers have shown that by carefully tuning how a system values different goals in real-time, it is possible to create a robust and reliable network of flying computers that can handle the demands of a dynamic world.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.