Abstract: Using UAVs for goods delivery can reduce costs, increase speed, and cut emissions. To achieve these benefits, optimising UAV deployment is crucial. Here, we focus on on-demand delivery where orders arrive stochastically and should be processed promptly—neither too late, which would prolong delivery times, nor too early, which could increase the risk of failed deliveries due to insufficient battery level. Unlike previous studies, we do not rely on a specific UAV energy model and consider unknown hardware conditions. We propose a decentralised auction-based strategy that enables UAVs, given their current battery level, to decide whether and how much to bid for an order. Our multi-agent simulations demonstrate that this approach outperforms methods that require UAVs to reach a specific battery level before deployment.

Read the article: https://inria.hal.science/hal-04723692/document