Kookmin University Presents RayOcc at IROS 2026

Korea Times reports that a paper from a Kookmin University research team has been accepted for presentation at the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), set for Sept. 27 to Oct. 1 in Pittsburgh. The paper is titled "RayOcc: Occlusion-Agnostic Ray Occupancy Estimation via Gaussian Mixture Intensity," with undergraduate senior Kim Jun-ho as first author and professor Lee Seong-won as the research lead, per Korea Times. Korea Times states the research was performed in Kookmin's Computer Intelligence Lab. The university noted the method aims to improve 3D occupancy estimation when observations are limited or parts of a scene are occluded, with applications to autonomous driving and robotics.

The paper applies Gaussian mixture intensity to ray-based occupancy estimation - a probabilistic approach that can model partial observability and occlusion inherent in real-world sensor data. IROS is a premier international robotics and intelligent systems venue, and acceptance signals peer review of the approach's merit. Per Korea Times, this is also the second lead-author acceptance for Kim Jun-ho: an earlier paper, "VG3T: Visual Geometry Grounded Gaussian Transformer," was accepted at ICRA 2026 (Vienna, June 1-5). "I find it very meaningful that the research I took part in as an undergraduate has been recognized at IROS following its acceptance at ICRA," Kim said, per Korea Times.

• •Whether a preprint or code release appears ahead of IROS in September, allowing independent replication and benchmarking.
• •Performance results on standard 3D perception benchmarks, particularly under occlusion conditions relevant to autonomous driving datasets.