Master Thesis: Multimodal Sensor Fusion and Robust Odometry for XR Perception

Join our Team

About this opportunity:

We are looking for a motivated Master's student to develop a robust perception and motion-estimation system for an XR platform operating in visually degraded environments.

The thesis will investigate how complementary sensing modalities-such as RGB or stereo cameras, infrared or thermal cameras, LiDAR or depth sensors, and an IMU-can be combined to improve pose estimation, depth perception, and 3D reconstruction when individual sensors may fail.

The work will contribute to the NEXSoS XR project, with a focus on challenging conditions such as darkness, smoke, low texture, motion blur, and partial sensor degradation. You will build on existing algorithms and open-source frameworks where appropriate, while developing the integration, calibration, synchronisation, and evaluation methodology needed for a practical multimodal system.

What you will do:

• Review approaches for visual-inertial odometry, LiDAR odometry, SLAM, and multimodal sensor fusion, and define a suitable architecture for the NEXSoS XR platform.
• Integrate available sensing hardware and develop calibration, synchronisation, preprocessing, and coordinate-transformation components.
• Implement or adapt a sensor-fusion and odometry pipeline using frameworks such as ORB-SLAM3, RTAB-Map, or other relevant methods.
• Evaluate the system under different environmental conditions.
• Compare individual sensors with multimodal configurations in terms of accuracy, robustness, latency, and computational requirements.
• Analyse and visualise trajectories, point clouds, depth maps, and 3D reconstructions.
• Document system limitations and provide recommendations for future development.

The skills you bring:

• You are enrolled in or recently admitted to a Master's programme in Robotics, Computer Science, Electrical Engineering, or a related field.
• You have basic knowledge of computer vision, robotics, estimation, 3D perception, or signal processing.
• You have programming experience in Python and/or C++.
• You understand coordinate transformations, rigid-body motion, and sensor data processing at a basic level.
• You are interested in working with real sensors, experimental hardware, and research software.
• You can analyse experimental results and communicate technical findings clearly.

The following are considered a plus:

• Experience with visual, inertial, or LiDAR odometry, or SLAM.
• Familiarity with ROS/ROS 2, ORB-SLAM3, RTAB-Map, OpenCV, Open3D, PCL, or similar tools.
• Knowledge of Kalman filtering, factor graphs, nonlinear optimisation, or bundle adjustment.
• Experience with camera-IMU or LiDAR-camera calibration.
• Previous experience with thermal, infrared, depth, or LiDAR sensors.