Laboratory of Artificial Intelligence and Robotics

At the Laboratory of AI and Robotics (LAIR), we aim to develop generalizable robotic intelligence that enables autonomous robots to perceive, understand, and interact with complex real-world environments. Unlike conventional AI systems operating primarily on pre-collected data, autonomous robots must continuously perceive and reason about dynamic and partially observable environments while making decisions under physical and computational constraints. Our research addresses these challenges by integrating spatial perception, representation learning, reasoning, and embodied action within a unified robotic intelligence framework. A central direction of our research is the development of persistent 3D spatial representations that integrate heterogeneous sensory observations and support geometric, semantic, and physical understanding of the environment. Based on these representations, we investigate sensor-agnostic simultaneous localization and mapping (SLAM), spatial grounding and reasoning, autonomous navigation, and action-centric planning and control. We further explore how robots can actively acquire information through movement and interaction, continuously update their understanding of the environment, and adapt their behaviors to previously unseen conditions. Ultimately, our research seeks to establish a generalizable framework in which perception, spatial memory, reasoning, and action are closely integrated, enabling autonomous robots to operate and adapt across diverse environments, sensing modalities, and robotic platforms. The ultimate goal is to develop adaptive robots that can automatically explore and enroll unknown environments and learn for their own purposes. To achieve this goal, we pursue the fusion of robotics and AI, which can be mutually beneficial, a