Research Focus
The objective of this project is to develop the scientific foundations of embodied intelligence and autonomous systems. We investigate how intelligent agents can construct internal representations of the physical world, reason over complex environments and execute actions that generalize across tasks and domains. Through research in world models, embodied learning, autonomous decision-making and multi-agent collaboration, we aim to develop unified frameworks that enable intelligent systems to continuously learn, adapt and operate in dynamic real-world settings.
Build a unified world model that continuously learns from multimodal sensory inputs, including vision, LiDAR, language and environmental signals.
Enable robots to understand complex environments, predict future states and reason about actions before execution, forming the cognitive foundation for physical intelligence.
Transform high-level goals into executable actions through integrated perception, planning, navigation and control.
Achieve robust autonomous operation across diverse industrial and emergency scenarios without task-specific programming or manual intervention.
Connect robots, humans and digital systems into a unified intelligent network capable of collaborative decision-making.
Enable large-scale coordination, adaptive task allocation and continuous collective learning in dynamic real-world environments.