Research Statement

My research focuses on how physical agents sense, reconstruct, and reason about the real world under uncertainty. I bridge communication and information theory with learning, geometry, and physics to build data-efficient, physically grounded world models for embodied intelligence.

Keywords:

embodied Intelligence & Physical Agents; World Models & Model-Based Learning; 3D Geometry, Reconstruction, and Rendering; Physical & Tactile Sensing; Uncertainty, Information Theory, and Communication; Reinforcement Learning under Physical Constraints; application domain include micro-robotic systems, legged and moving robots, and embodied sensing platforms.

My research trajectory started from communication and information theory, where I worked on UAV assisted ISAC system, grounding my training in uncertainty, representation efficiency, and signal-to-decision pipelines.

From there, my focus gradually shifted from transmission to perception: how sparse, noisy, and partial sensory signals can be reconstructed into meaningful representations of the physical world. This led me to work on tactile sensing, sparse sensor reconstruction, and learning-based methods for recovering stress fields and geometric structures.

Recently, my interests have converged on learning world models for physical scenes—combining geometry, physics, and uncertainty—spanning 3D reconstruction, physically grounded rendering, and deformable object modeling. Along this line, I have explored reinforcement learning and model-based decision-making in embodied settings, including magnetic-controlled micro-robotic systems and autonomous agents operating under physical constraints.

Rather than focusing on a specific robot morphology, I am interested in general physical agents: systems that sense, infer, and act in the real world under uncertainty. My background across communication, learning, geometry, and physical systems allows me to bridge perception, representation, and control, with the long-term goal of enabling more data-efficient and physically grounded embodied intelligence.