Practical 3D Reconstruction using mmWave radars
This research project aims to leverage machine learning techniques to enhance and reconstruct 3D objects in practical scenarios for autonomous driving.
Wireless signals offer a significant advantage for sensing and imaging, especially in situations where traditional modalities like cameras or LiDAR fall short. Optical sensors often struggle in conditions with occlusions, such as fog, rain, or snow. However, wireless signals, particularly mmWave signals, can penetrate these obstacles, enabling us to perceive environments beyond the limitations of visual sensors.
Despite these advantages, mmWave radar sensors face hardware limitations that affect their resolution and field of view, making their signals difficult to interpret. This project seeks to extend the typical two-dimensional field of view into three dimensions without compromising spatial resolution. Additionally, we explore various methods to enhance these signals by combining machine learning techniques with advanced signal processing.
Active student projects
- Multi-Radar SLAM platform
- Deploying a SLAM pipeline using a single mmWave Cascaded radar
- Study of sensors synchronization and calibration with mmwave radar
- Enhancing mmWave radar heatmaps empirically using deconvolution
Completed student projects
-
Multi-Sensor Assisted Mobile Millimeter-Wave Mapping Platform
Student: Cyril Golaz
-
3D partial and complete point clouds dataset using mmWave radar
Student: Swathi Narashiman
People
