The Use of AI Technology in Drone Delivery

(Image Source: MeiTuan)


With the flourishing development of e-commerce and increasing consumer demand for faster delivery, traditional delivery methods are no longer sufficient to meet customer expectations. Drone delivery, with advantages such as efficiency, low cost, environmental friendliness, and flexibility, has gradually become a prominent delivery tool[1]. In August 2022, the Civil Aviation Administration of China released the “Civil Unmanned Aerial Vehicle Development Roadmap V1.0 (Draft for Comments),” outlining the development timetable for drones in the logistics sector. The draft indicates that short-distance, low-speed, light-weight drone delivery in urban areas will mature, be applied, and promoted before 2035. Concurrently, the rapid development of big data, cloud computing, and artificial intelligence has provided robust support for drone delivery.

Application of AI Technology in Drone Delivery

Modern advanced technologies often initially develop in the military domain before being applied in civilian sectors, and drone development is no exception. Initially designed for military purposes, drones aimed to replace manned aircraft, reduce casualties, and address extreme situations, as shown in Figure 1. Today, they find widespread application in scientific research and civilian fields, especially in developed countries with higher labor costs.

Figure 1. Drone Development Process[2]

In contemporary society, commercial drone applications continue to expand, with companies actively exploring drone delivery services. Initially, this trend aimed to address the high delivery costs in rural and remote areas. Due to inconvenient transportation and dispersed populations in rural areas, traditional delivery methods cost approximately five times more than in urban areas. Drones, with their convenience, low cost, high mobility, and adaptability, have become an ideal solution. Companies like SF Express,, and Antwork have initiated drone logistics pilot projects in rural areas. For example, successfully reduced the delivery cost per order from 20-30 Yuan to 5 Yuan[3]. While urban drone logistics started later due to densely populated areas, complex terrain, and policy restrictions, it is crucial for improving urban logistics, including enhancing delivery speed, reducing costs, and meeting various demands. Currently, urban drone deliveries typically employ a hub-and-spoke distribution model, where drones transport goods from one station to another, rather than directly delivering to customers. Companies like EHang have promoted this model in some regions, significantly reducing delivery time from 40 minutes to 8 minutes and cutting operational costs by nearly 80%.

In recent years, with the rapid development of artificial intelligence technology, researchers have started applying AI to drone delivery, providing key support in the following areas:

  1. Path Planning and Obstacle Avoidance: Machine learning algorithms assist drones in planning optimal routes, avoiding obstacles, and navigating through hazardous areas. Real-time monitoring of weather, traffic, and other drone positions enables AI systems to intelligently choose flight paths, enhancing safety and efficiency[4][5]. Academic journal[6] proposed a sequential route network planning method for drone delivery, and Figure 2 demonstrates its application in real-world scenarios. While traditional manual route design takes 2 to 4 hours, this method can automatically design 40 routes within 1 hour, significantly reducing route design time.
  2. Flight Control and Stability: AI technology enhances drone flight control and stability. By monitoring wind speed, atmospheric pressure, and other environmental factors, AI systems adjust drone flight attitudes to ensure the safe delivery of goods.
  3. Data Analysis and Optimization: AI algorithms analyze large amounts of data to optimize delivery routes and schedules. By collecting and analyzing historical delivery data, AI systems determine the best delivery time windows and paths, improving efficiency and reducing costs. For instance, Academic journal[7] designed a novel competition-based route planning algorithm, maximizing individual route performance while optimizing overall system performance.

Figure 2. Drone Delivery Path Automatically Designed by Algorithm[6]

Case Study: Meituan Drone Delivery

In recent years, Meituan Food delivery has faced increasing pressure on delivery costs. According to Zhiyouji and Guotai Junan Securities research, from 2017 to 2021, the average monthly salary of Meituan delivery riders increased from nearly 6,000 Yuan to almost 10,000 Yuan, prompting Meituan to actively engage in the research and application of drone delivery technology to explore possible solutions. Figure 3 illustrates the Meituan drone delivery process. After a user place an order, a delivery rider picks up the goods and places them at a nearby community delivery station. The drone then picks up the goods and delivers them to the station, allowing users to retrieve their orders by scanning a QR code with their smartphones. As of now, Meituan drones have been operational in multiple cities, completing 170,000 delivery tasks covering nearly 20,000 types of products.

Figure 3. Meituan Drone Delivery Process Diagram.

The safety and efficient delivery of Meituan drones rely on the application of AI technology, as follows:

  1. Visual Navigation System: Given the high-rise buildings prevalent in Chinese urban environments, Meituan developed a computer vision-based navigation system for drones to navigate through high-risk environments. The system autonomously perceives the environment and plans flight routes. Meituan released a new algorithm at the 2022 ICRA conference, increasing the positioning accuracy of drones during visual flight by nearly 30%[8], significantly enhancing flight safety. Facing limited computing power, Meituan adopted a “walk and look” strategy to improve recognition accuracy and reduce computational consumption.
  2. Unmanned Aircraft Systems Traffic Management (UTM) System: Meituan’s UTM system connects various units, utilizing AI technology for order processing and route scheduling to enhance delivery efficiency. Its innovative four-dimensional space-time capsule system coordinates multi-drone operations efficiently and reliably. The system can effectively manage the scheduling of thousands of drones per square kilometer, monitor status in real-time, intelligently identify anomalies, and take corresponding measures to reduce flight risks.

Outlook and Challenges in the Future

AI technology will play a more crucial role in the drone delivery field to address future challenges. Future trends include continuous optimization of path planning through real-time adaptation to factors like traffic and weather, reducing delivery times and saving energy. Simultaneously, AI will coordinate drone flights for more precise environmental perception and flight control, reducing accident risks. The ongoing improvement of data analysis and predictive capabilities will optimize delivery efficiency, lower costs, and enhance the autonomous flight and collaborative operation capabilities of drones during peak hours and large-scale deliveries, thereby improving the scalability of urban logistics.

AI technology has made significant progress in the drone delivery field, as demonstrated by successful cases like Meituan. Despite the current higher cost of urban drone delivery compared to traditional human delivery and the need for further policy refinement, it is expected that more companies will enter the drone delivery sector as the demographic dividend gradually diminishes and policy support becomes more robust. However, this field still faces various challenges, including complex airspace management, safety issues in urban environments, the impact of dynamic conditions on route planning, as well as challenges related to data privacy and security. Regarding the futuristic scenario of drones delivering directly to homes as depicted in science fiction novels, Meituan’s head of drone delivery unit, Yinan Mao, believes it could become a reality but may require another 20 to 30 years.[9]


[1] Wei, H., Shi, J., Liu, Z., Zhang, J., & Liu, Z. (2023) “Overview of Drone Delivery Models and Path Planning for Last-Mile Delivery.” Computer Systems & Applications, 32(9): 1-18.

[2] YTO Research Institute, “Analysis and Prospects of Drones Now and in the Future and Their Application in the Express Logistics Industry.”

[3] Bao, X., Luo, P., & Wang, K. (2017). “Analysis of the Advantages and Obstacles of Drone Delivery.” Modern Business, (23), 13-14.

[4]He, X., Jiang, C., Li, L., & Blom, H. (2022). A Simulation Study of Risk-Aware Path Planning in Mitigating the Third-Party Risk of a Commercial UAS Operation in an Urban Area. Aerospace, 9(11), 682.

[5]Li, L., He, X., Mo, Y., Sun, Z., & Qin, S. J. (2023). Route Network Planning for Urban Drone Delivery: Network Flow Theory or Graph Search Algorithms?. Available at SSRN 4589264.

[6] He, X., He, F., Li, L., Zhang, L., & Xiao, G. (2022). A route network planning method for urban air delivery. Transportation Research Part E: Logistics and Transportation Review, 166, 102872.

[7] He, X., Mo, Y., Huang, J., Li, L., & Qin, S. J. A Competition-Based Route Network Planning Method for Drone Delivery Services in Cities. Available at SSRN 4370169.

[8] Hu, J., Hu, J., Shen, Y., Lang, X., Zang, B., Huang, G., & Mao, Y. (2022, May). 1d-lrf aided visual-inertial odometry for high-altitude mav flight. In 2022 International Conference on Robotics and Automation (ICRA) (pp. 5858-5864). IEEE.

[9] Yang, Z. (2023). Food delivery by drone is just part of daily life in Shenzhen.

The work described in this article was supported by InnoHK initiative, The Government of the HKSAR, and Laboratory for AI-Powered Financial Technologies.
(AIFT strives but cannot guarantee the accuracy and reliability of the content, and will not be responsible for any loss or damage caused by any inaccuracy or omission.)

Share this content


Units 1101-1102 & 1121-1123,
Building 19W Science Park West Avenue,
Hong Kong Science Park,
Shatin, Hong Kong

Products & Solutions


About Us


Copyright © 2024 Laboratory for AI-Powered Financial Technologies Ltd. All Rights Reserved.