Multi-View Depth Estimation for Planar Light-Field Arrays
Research Immersion Lab project evaluating multi-view depth estimation algorithms on planar light-field camera arrays, as part of the FROST project.
I am a Ph.D. student at the Telecommunications Lab of Saarland University, which I joined in February 2025. My Ph.D. is funded by the Saarland/Intel Joint Program on the Future of Graphics and Media. I work on 3D computer vision for next-generation visual media within the FROST project: light-field scene representations, light-field rendering and multi-view depth estimation.
My background combines computer vision and robotics, with a focus on multi-modal 3D perception and RGB–depth sensor fusion: fusing LiDAR point clouds with camera images for object detection, RGB-D fusion for autonomous drone perching, active vision under limited fields of view, and visual-inertial state estimation for multi-robot aerial systems.
I received my M.Sc. in Robotics from the University of Bristol in 2024 and my B.Sc. in Robotics from Beijing Institute of Technology in 2022, with a Distinguished Bachelor's Thesis on LiDAR–camera active vision, supervised by Prof. Yushu Yu, Prof. Vincenzo Lippiello and Dr. Mauro Bellone.
Saarland University · Telecommunications Lab · Saarbrücken, Germany
University of Bristol · Bristol, UK
Beijing Institute of Technology · Beijing, China
Froxels for Production: Ray-Level Inspection and Consensus-Guided Light-Field Rendering
ACM European Conference on Visual Media Production (CVMP), London, UK, 2026. Accepted
This work will be presented by my co-authors at CVMP in London. Please feel welcome to reach out to them during the conference.
Multi-Agent Visual-Inertial Localization for Integrated Aerial Systems With Loose Fusion of Odometry and Kinematics
IEEE Robotics and Automation Letters, vol. 9, no. 7, pp. 6504–6511, 2024.
Tight Fusion of Odometry and Kinematic Constraints for Multiple Aerial Vehicles in Physical Interconnection
IEEE International Conference on Robotics and Automation (ICRA), pp. 3891–3897, 2024.
Research Immersion Lab project evaluating multi-view depth estimation algorithms on planar light-field camera arrays, as part of the FROST project.
Turned LiDAR point clouds into multi-channel images and fused them with camera images for CNN object detection. Derived perception constraints that keep the region of interest in both sensors' limited fields of view, built them into a model predictive controller, and flew it on a real small-scale UAV.
Constrained control for systems on manifolds with uncertainties and input/state constraints, via tube-based MPC on Euclidean space. Proved input-to-state stability and recursive feasibility, and implemented the controller on real aerial robots.
Modelled multi-UAV cooperative transportation systems and designed a hierarchical control scheme that handles uncertainties and state/input constraints, compared against existing controllers in MATLAB simulation.
Real-time YOLOv3 detection of cars and blocks on a Raspberry Pi camera, combined with ultrasonic and infrared sensing and Raspberry Pi-to-Arduino serial motor control for autonomous obstacle avoidance.
Modelled an amphibious quadrotor that keeps a submarine-style hull, with a retractable four-rotor structure for switching between flight and diving, plus onboard sensors and cameras for environmental monitoring.
Czech Technical University in Prague · Multi-robot Systems (MRS) Group · Prague, Czech Republic
University of Bristol · Bio-Inspired Flight Lab · Bristol, UK
Megvii Technology · Beijing, China
North Carolina State University · Raleigh, NC, USA
Massachusetts Institute of Technology · Cambridge, MA, USA