□ DGIST (President Kunwoo Lee) announced on December 13 that two of its research teams have been selected for the NVIDIA Academic Grant Program, a research support initiative run by the global AI company NVIDIA, through which the teams shall receive access to high-performance GPU computing resources for research in next-generation artificial intelligence (AI) and robotics.
□ NVIDIA’s Academic Grant Program is an initiative that selects innovative research projects from universities and research institutions worldwide and provides them with GPU hardware or large-scale computing resources. Universities and research institutions face difficulties in continuously securing the required level of computing infrastructure because of practical constraints, such as budget limitations, power capacity, and deployment timelines. To bridge this gap and support cutting-edge academic research and talent development, NVIDIA operates this program that provides GPU resources for competitive research projects.
□ The DGIST projects selected under this program are: 1)research on AI-driven, physics-based video generation models and 2) research on human–robot collaborative intelligence based on Vision–Language–Action.
□ Professor Sunghoon Im’s research team shall receive 32,000 A100 GPU-hours to support “Research on Physics-Compliant Video Generation Models Using Chain-of-Thought Reasoning.” The team is studying a model in which AI sequentially infers object motion in accordance with physical laws, enabling the generation of natural videos with minimal distortion.
□ Professor Im stated, “We aim to enhance model performance by leveraging approximately 32,000 hours of computing resources and to expand the scope of our research to robot simulation and autonomous driving by realizing high-fidelity video generation technology that minimizes physical errors.”
□ The joint research team led by Professors Daehee Park and Jisoo Mok shall receive four RTX PRO 6000 Blackwell Server Edition GPUs (hardware) to support “Extended Research on Vision-Language-Action (VLA) Models for Dynamic Human-Robot Collaboration.” The team is developing physical AI-based robotic intelligence that enables robots working alongside humans to understand their surroundings and human movements and select safe collaborative behavior.
□ Professors Park and Mok stated, “We will accelerate large-scale training and validation by leveraging the high-performance GPUs provided and further develop collaborative robotic intelligence that can be applied in practice to expand our research outcomes.”
□ The selection of the DGIST research teams for NVIDIA’s Academic Grant Program is significant, as it highlights the global recognition of the teams’ research directions and potential. Since the training and validation of large-scale AI models are highly dependent on the availability of computing resources, which directly determine the scope and speed of research outcomes, both research teams are expected to expand the scale of their experiments based on this support and further enhance the precision of model performance validation.
