Ready, Bid, Go! On-Demand Delivery Using Fleets of Drones with Unknown, Heterogeneous Energy Storage Constraints Conference: Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025); May 19 – 23, 2025; Detroit, Michigan, USA <<fp1317>>
Mohamed S. Talamali, Genki Miyauchi, Thomas Watteyne, Micael S. Couceiro, Roderich Groß · 2025
Read Ready, Bid, Go! On-Demand Delivery Using Fleets of Drones with Unknown, Heterogeneous Energy Storage Constraints Conference: Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025); May 19 – 23, 2025; Detroit, Michigan, USA <<fp1317>> by Mohamed S. Talamali, Genki Miyauchi, Thomas Watteyne, Micael S. Couceiro, Roderich Groß (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-08585.txt.
Introduction
Mohamed Talamali and co-authors published a research paper in 2025 on sending delivery drones to orders as they arrive, when no one knows how much charge each drone truly holds. Fleets differ drone to drone, and their energy use is hard to model ahead. The team proposes a scheme where drones bid for jobs and learn from each flight.
Each drone bids only when its charge, the parcel's weight, and the trip length look safe, and it sharpens that judgment over time. In simulations with realistic energy models, giving jobs to the least sure bidders, oddly, cut wait times and raised completed orders. A forecasting variant lets low drones promise future slots.
Catalog record
| CREATOR | Mohamed S. Talamali, Genki Miyauchi, Thomas Watteyne, Micael S. Couceiro, Roderich Groß |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-08585
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Ready, Bid, Go! On-Demand Delivery Using Fleets of Drones with Unknown, Heterogeneous Energy Storage Constraints Conference: Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025); May 19 – 23, 2025; Detroit, Michigan, USA <<fp1317>> Mohamed S. Talamali, Genki Miyauchi, Thomas Watteyne, Micael S. Couceiro, Roderich Groß Abstract . Unmanned Aerial Vehicles (UAVs) are expected to transform logistics, reducing delivery time, costs, and emissions. This study addresses an on-demand delivery , in which fleets of UAVs are deployed to fulfil orders that arrive stochastically. Unlike previous work, it considers UAVs with heterogeneous, unknown energy storage capacities and assumes no knowledge of the energy consumption models. We propose a decentralised deployment strategy that combines auction-based task allocation with online learning. Each UAV independently decides whether to bid for orders based on its energy storage charge level, the parcel mass, and delivery distance. Over time, it refines its policy to bid only for orders within its capability. Simulations using realistic UAV energy models reveal that, counter-intuitively, assigning orders to the least confident bidders reduces delivery times and increases the number of successfully fulfilled orders. This strategy is shown to outperform threshold-based methods which require UAVs to exceed specific charge levels at deployment. We propose a variant of the strategy which uses learned policies for forecasting. This enables UAVs with insufficient charge levels to commit to fulfilling orders at specific future times, helping to prioritise early orders. Our work provides new insights into long-term deployment of UAV swarms, highlighting the advantages of decentralised energy-aware decision-making coupled with online learning in real-world dynamic environments. Y. Vorobeychik, S. Das, A. Nowé (eds.) 1. Introduction Unmanned Aerial Vehicle (UAV)-based on-demand delivery is expected to bring significant benefits to society, including reductions in delivery times, operational costs, and carbon emissions. By deploying a fleet of UAVs, multiple orders can be fulfilled simultaneously, enabling large-scale applications in parcel, meal, and medication deliveries. Various methodologies have been explored to address delivery problems with fleets of energy-constrained UAVs, including those that optimise order allocation and route planning. These include mixed-integer linear programming (MILP) ( Song et al. 2018 ; Agatz et al. 2018 ; Schermer et al. 2019 ; Kitjacharoenchai et al. 2019 ) , queuing theory ( Grippa et al. 2019 ) , auctions ( Rinaldi et al. 2022 ) , and various heuristic methods ( Dorling et al. 2016 ; Peng et al. 2019 ; Poikonen and Golden 2020 ) . An interesting problem variant is on-demand deliveries, where orders arrive stochastically over time ( Huang et al. 2021 ; Liu 2019 ; Chen et al. 2021 ) . Figure 1. On-demand delivery scenario with a fleet of UAVs committing to deliver orders arriving at a fulfilment centre. The fleet is heterogeneous, as the UAVs differ in battery health, and hence in their true energy storage capacities. Each UAV learns a policy for placing bids on incoming orders based on its current charge level, parcel mass and delivery distance. Optionally, it can use this policy for forecasting, enabling it to plan the fulfilment of orders in the future. <Illustration of a UAV-based on-demand delivery scenario> <Illustration of a UAV-based on-demand delivery scenario> Despite these advancements, studies often rely on several idealised assumptions that may not hold in real-world settings. First, they typically assume that the UAVs know their energy consumption models and can use them when planning their deliveries. This poses the risk of over-reliance on the given models: While various energy models have been proposed, it has been shown that different models can lead to divergent outcomes even under the same conditions ( Zhang et al. 2021 ) . Second, they commonly assume that UAVs are homogeneous. Yet in practice, even UAVs of the same make and model would differ due to variations in battery health, hardware wear, and other factors. These variations effectively result in a unique energy model for each UAV. Finally, they commonly assume the UAVs to charge instantaneously when visiting their base. This may hinder application in practical scenarios, as not reflecting the low duty cycles of most commercial delivery UAVs (e.g. 6.97% for DJI FlyCart 30 with one battery). Factoring in the significant proportion of time devoted to charging ultimately leads to a new class of problems. For example, UAVs opting to dispatch with only a partial charge could result in faster deliveries than those waiting to charge fully. The contributions of this paper are threefold: (1) Decentralised learning-based deployment strategy for UAVs with unknown battery health: We propose a decentralised learning-based deployment strategy for on-demand delivery using UAV fleets. Each UAV decides whether to bid on incoming orders based on its battery’s state of charge , parcel mass, and delivery distance. Unlike prior work, our approach requires no knowledge of the battery’s state of health (i.e., its true capacity) or the energy consumption model. Over time, each UAV learns to bid only on orders that it is likely to fulfil, enhancing the efficiency of the fleet. (2) Long-term evaluation of heterogeneous fleets: We evaluate the proposed strategy using simulated UAV fleets that vary in battery health due to factors such as ageing, usage, and storage patterns. Simulations using realistic UAV energy models are conducted over a period of eight weeks with orders arriving stochastically at different rates. The results show that deploying the least confident bidding UAVs leads to the highest number of fulfilled orders and the lowest delivery times. Furthermore, the strategy outperforms traditional approaches that dispatch UAVs only when their battery charge exceeds predefined thresholds. (3) Forecasting-enabled order commitment: We propose a second variant of the strategy where learned bidding policies are used for forecasting. This enables UAVs with insufficient charge levels to commit to fulfilling current orders at a specific future time, improving prioritisation of earlier orders and aligning better with first-come, first-served scheduling. The remainder of this paper is organised as follows. Section 2 presents the problem formulation. Section 3 details the proposed decentralised learning-based deployment strategy. Sections 4 and 5 present the simulator and results, respectively. Finally, Section 6 concludes the paper. 2. PROBLEM STATEMENT As illustrated in Figure 1 , a fulfilment centre (FC) operates a fleet of delivery UAVs, denoted as U=\{1,2,\dots,S\} . During an operational period of length T , the FC stochastically receives orders O=\{1,2,\dots\} , which arrive at times t_{1},t_{2},\dots , where 0\leq t_{1}<t_{2}<\dots<T . Each order j\in O corresponds to a delivery task involving a parcel of mass m_{j}\in[m_{\text{min}},m_{\text{max}}] , which is to be delivered to a location at a distance d_{j}\in[d_{\text{min}},d_{\text{max}}] from the FC. The FC maintains a communication infrastructure (e.g., a wireless network) to communicate with available UAVs (e.g., to announce delivery tasks) and allows UAVs to communicate with each other, facilitating collaborative decision-making. All UAVs are powered by an on-board battery with a theoretical capacity C_{\text{theoretical}} . The battery of UAV i\in U has a state of health (SoH) , denoted as \text{SoH}_{i}\in[0,1] , which determines its true (maximum) capacity. Specifically, the true battery capacity is given by C_{\text{true},i}=\text{SoH}_{i}\cdot C_{\text{theoretical}} . In this study, \text{SoH}_{i} is assumed to remain constant throughout the operational period. At any given time t , UAV i\in U possesses a state of charge (SoC) denoted by \text{SoC}_{i}(t)\in[0,100] which