We develop intelligent, reliable, and resource-efficient learning systems at the intersection of signal processing, artificial intelligence, and autonomous decision-making, bridging fundamental methods with real-world, industrial, and safety-critical applications. Our core research areas are:
- Signal Processing for Localization and Sensing are key enablers for modern intelligent systems, ranging from autonomous agents and robotics to extended reality and next-generation mobile communication systems. We develop data-driven methods for channel estimation, localization, tracking, motion prediction, and multimodal sensor fusion using radio, inertial, and visual data. We emphasize self-supervised and unsupervised learning to reduce reliance on annotated data. A key focus is future mobile networks (6G and beyond), where communication and sensing converge, integrating localization, perception, and communication.
- Edge AI and Resource-Efficient Learning: We design learning algorithms optimized for accuracy, memory, and energy efficiency, enabling deep learning on resource-constrained devices such as embedded systems and microcontrollers for intelligent sensing, manufacturing, and mobile applications.
- Reliable and Trustworthy Reinforcement Learning: We address safety, stability, and explainability challenges through risk-aware exploration, stabilization in uncertain environments, and interpretable reinforcement learning. Our aim is to enable trustworthy reinforcement learning in real-world, safety-critical, and industrial systems.
Selected Publications
