Advancing Practical Indoor Localization through Multi-Sensor Fusion and Deep Learning
Project Description
Indoor localization has many practical applications, from navigation in shopping malls and airports to asset tracking in warehouses. Although GPS works well for outdoor localization, it fails indoors because building structures block its signals. Indoor localization remains a challenging problem due to complex signal propagation, occlusions, and the constantly changing nature of indoor environments.
This project aims to address these challenges by developing a practical indoor localization system for HKUST that fuses data from multiple onboard smartphone sensors — including Wi-Fi, Bluetooth, and Inertial Measurement Units (IMUs) — with advanced deep learning techniques. By leveraging the complementary strengths of these modalities, we seek to achieve accurate, robust, and cost-effective indoor positioning without relying on exhaustive manual calibration. The system will be designed to support real-world navigation tasks, such as guiding users through large and complex buildings, and will be developed in collaboration with a logistics company to deploy the system in their warehouses.
This project aims to address these challenges by developing a practical indoor localization system for HKUST that fuses data from multiple onboard smartphone sensors — including Wi-Fi, Bluetooth, and Inertial Measurement Units (IMUs) — with advanced deep learning techniques. By leveraging the complementary strengths of these modalities, we seek to achieve accurate, robust, and cost-effective indoor positioning without relying on exhaustive manual calibration. The system will be designed to support real-world navigation tasks, such as guiding users through large and complex buildings, and will be developed in collaboration with a logistics company to deploy the system in their warehouses.
Supervisor
LI, Mo
Quota
2
Course type
UROP1000
UROP1100
Applicant's Roles
- A solid understanding of deep learning and prior experience with deep neural network (DNN) projects.
- A basic understanding about localization.
- Previous involvement in research projects or scholarly publications will be considered a plus.
- A basic understanding about localization.
- Previous involvement in research projects or scholarly publications will be considered a plus.
Applicant's Learning Objectives
- Develop fundamental research skills in the field of indoor localization and sensor fusion.
- Gain insights into the deployment of machine learning models in real scenarios.
- Gain insights into the deployment of machine learning models in real scenarios.
Complexity of the project
Moderate