AI-Driven Drone Simulation and Planning for Onsite Construction Resources Delivery
Project Description
This project explores the integration of multi-drone systems to automate construction resources delivery in high-density urban environments, with a focus on Hong Kong’s emerging low-altitude economy. Traditional construction logistics involves a labor-intensive process: materials are unloaded from trucks, transported to near-site staging areas, and manually lifted using cranes. In recent years, Hong Kong has increasingly adopted Modular Integrated Construction (MiC), which integrates high-quality factory prefabrication of modular components and fast on-site assembly. This transition has heightened the demand for efficient and automated on-site construction solutions, further emphasizing this research's impact in revolutionizing construction logistics.
On the other hand, drones provide a streamlined alternative by taking off directly from trucks and delivering materials to workers in a single, efficient operation, thereby bypassing ground-level inefficiencies. However, enabling drones to autonomously navigate complex construction environments with obstacles such as scaffolding, buildings, and moving workers—remains a significant challenge. Within this context, our work leverages advancements in AI methods, particularly reinforcement learning and vision large language models, to equip drones with advanced intelligence. These technologies enable drones to interpret real-time data, optimize flight paths, and execute precision deliveries for construction resources, including various building elements such as structural modules, walls, beams, columns, slabs, etc. The project will be conducted in physically-realistic simulated environments and extended to real-world applications.
On the other hand, drones provide a streamlined alternative by taking off directly from trucks and delivering materials to workers in a single, efficient operation, thereby bypassing ground-level inefficiencies. However, enabling drones to autonomously navigate complex construction environments with obstacles such as scaffolding, buildings, and moving workers—remains a significant challenge. Within this context, our work leverages advancements in AI methods, particularly reinforcement learning and vision large language models, to equip drones with advanced intelligence. These technologies enable drones to interpret real-time data, optimize flight paths, and execute precision deliveries for construction resources, including various building elements such as structural modules, walls, beams, columns, slabs, etc. The project will be conducted in physically-realistic simulated environments and extended to real-world applications.
Supervisor
WANG, Ziqi
Quota
2
Course type
UROP4100
Applicant's Roles
1. Develop a database of construction recourses suitable for transport by one or multiple drones alongside digital twins of real-world construction sites.
2. Develop innovative path-planning and swing suppression algorithms to enable drones to lift, transport, and place construction resources at high precision, ensuring they navigate around static and dynamic obstacles within our physically realistic simulation environments.
2. Develop innovative path-planning and swing suppression algorithms to enable drones to lift, transport, and place construction resources at high precision, ensuring they navigate around static and dynamic obstacles within our physically realistic simulation environments.
Applicant's Learning Objectives
1. Learn to use drone simulator.
2. Learn to plan the trajectory of drones.
3. Learn to write academic papers.
2. Learn to plan the trajectory of drones.
3. Learn to write academic papers.
Complexity of the project
Challenging