Research

1) Optimize operating systems with neuromorphic support for next-generation satellite computers.

2) Develop highly reliable aerospace flight simulators to ensure robust space missions.

3) Optimize full-stack robot operating systems for advanced Vision-and-Language Navigation (VLN).

4) Enhance cloud resource management by developing robust anomaly detection and allocation technologies for GPU clusters.

5) Advance OS technologies to support heterogeneous processors, CXL memory, and modern sensor systems.

6) Enable efficient and scalable embedded AI systems through future neuromorphic computing and sensor technologies.

We research operating system technologies to seamlessly integrate and optimize neuromorphic processors for next-generation satellite computers, addressing the critical challenges of strict power constraints and the growing demand for high-performance edge AI in space environments.

We research scalable satellite emulators targeting the GR740 processor, addressing the limitations of expensive, proprietary foreign simulators to establish an independent and highly efficient technological ecosystem. We focus on enhancing cycle-accurate modeling fidelity and overall simulation speed, while also developing a user-friendly, GUI-based Integrated Development Environment (IDE) that provides intuitive monitoring, comprehensive command execution, and scalable support for diverse development boards.

We research real-time operating system (RTOS) technologies to optimize fault tolerance management across safety-critical applications, addressing the challenges of schedulability degradation caused by uniform protection mechanisms. We study adaptive allocation techniques in RTEMS 6 that dynamically assign fault tolerance operations based on task criticality, iterative schedulability testing, and equitable overhead distribution to support high-reliability and stable real-time computing workloads.

We research Linux kernel technologies to optimize resource management across CPUs, GPUs, and NPUs, addressing the challenges of heterogeneous processor scheduling for AI and data-intensive applications. We also study memory management techniques for tiered systems using CXL-attached memory, focusing on efficient page placement, dynamic tiering, and minimizing migration overhead to support large-scale computing workloads.

We propose automatic SNN (Spiking Neural Network) generation methods optimized for neuromorphic hardware, supporting diverse IoT edge service requirements through profiling-based modeling and integration with platforms like Node-RED.

We build intelligent solutions across robotics and healthcare by utilizing advanced sensors such as LiDAR, 3D laser scanners, and AR-based systems, focusing on energy-efficient sensor management and real-time guidance technologies.