Speakers

Even as AI accelerates partner discovery and connections, shared outcomes require people and organizations that can design how they work together. Drawing on differences in pace, responsibility, and definitions of success in international joint R&D, this presentation proposes a global collaboration strategy for GAIA through The Collaboration Equation. It introduces the Interfacialist, three collaboration protocols—context translation, buffer-zone building, and reaching shared agreement—and 15 core competencies. It also explores how individual experience can become shared records, operating standards, talent development, and structured handovers, creating a foundation for collaboration beyond the funded program.

This session examines the current state of autonomous driving, from SAE Levels and system architecture—including sensors, perception, planning, and control—to industry case studies from Tesla, Waymo, NVIDIA, and Hyundai Motor Group. It explores the technical, regulatory, ethical, and economic challenges between today’s Level 2–3 systems and full autonomy, and considers emerging solutions and convergence with SDV and robotics.

This presentation explores the evolution from connected mobility towards cooperative and intelligent mobility from a European research perspective. It highlights how 5G and emerging 6G technologies extend connectivity and information exchange towards cooperative perception and distributed intelligence. Drawing on European research initiatives and real-world experimentation, including TU Chemnitz examples, it discusses the integration of communication, sensing, AI, and computing, and opportunities for 6G-enabled mobility and Europe–Korea research collaboration.

This presentation reviews AI and data-driven methods for spacecraft thermal control, including surrogate modeling, automated thermal-model correlation using thermal-vacuum test data, on-orbit telemetry anomaly detection, and design optimization. Drawing on KARI’s experience with the Korea Pathfinder Lunar Orbiter (KPLO, Danuri) and the Korean Lunar Lander under development, it outlines where AI adds value, where physics-based analysis remains indispensable, and a practical roadmap for integrating both approaches.

Anonymizing sensitive information such as names or income may not fully protect privacy when released datasets contain rich information. This presentation introduces Differential Privacy (DP) as a mechanism for protecting individual privacy in large datasets and reviews research progress, including DP approaches for Large Language Models (LLMs).

As artificial intelligence transitions from central cloud infrastructures to real-time physical systems, the AI Control Unit has emerged as an essential hardware platform for mission-critical physical AI. While conventional automotive control relied on fragmented ECUs, modern intelligent mobility and autonomous agents demand unified, high-efficiency heterogeneous compute, deterministic low latency, and robust functional safety to execute complex multimodal AI at the edge. Drawing upon extensive industrial engineering experience—from flight control systems (FC/GCS) and intelligent edge IoT to NVIDIA Jetson-based embedded AI architectures—this presentation explores the technological evolution and strategic advantages of next-generation ACUs. It addresses key engineering bottlenecks including thermal dissipation, real-time deterministic scheduling, and edge privacy protection. Furthermore, the session highlights the strategic expansion of ACU platforms across diverse physical AI verticals, including autonomous mobile robots (AMRs), humanoid robotics driven by Vision-Language-Action (VLA) models, and intelligent healthcare devices, demonstrating how modular ACU platforms bridge academic algorithmic innovation with reliable industrial commercialization.

Next-generation advanced mobility demands reliable, efficient, and robust drive and control technologies capable of operating in harsh and unpredictable environments. This presentation explores how JUSTEK integrates AI with its core hardware technologies—vibration reduction, air-bearing-based dust-free systems, and extreme-environment motors for space applications—to address these challenges. In motion control, physics-based models are combined with neural networks and reinforcement learning to compensate for nonlinear friction, thrust ripple, and structural vibration in real time. In motor design for planetary rover drives, TD-VAE surrogate modeling, hybrid Physics-Informed Neural Networks (PINNs), and DQN with Tree Search reinforcement learning are used to accelerate electromagnetic analysis and optimize designs beyond the computational constraints of conventional finite element methods (FEM). The presentation concludes with a vision for an autonomous, self-optimizing motion and drive ecosystem enabled by edge-AI real-time control, digital twin diagnostics, and AI-driven motor design automation, highlighting opportunities for open innovation and collaborative research.