About Me
I am a robotics researcher working at the intersection of world models, cognitive robotics, and autonomous planning. I am interested in how robots can act through sparse, task-relevant internal representations rather than relying on dense reconstruction and reactive control.
My current work centers on TIWM (Tokenized Intent World Model), a framework for intent-driven planning through sparse internal structure. In the long term, I aim to connect interpretable cognitive structure with long-horizon autoregressive planning for more generalizable embodied intelligence.
- Cognitive Robotics
- World Representation
- Deep Reinforcement Learning
- Robotic Middleware (ROS/ROS2)
- LLM/VLM for Robotics
MEng Computer Technology
Central China Normal University
MSc Computer Science
University of Wollongong
BSc Computer Science and Technology
Huaiyin Institute of Technology
- Sparse Semantic Grounding & World Representation
I study how task-relevant and semantically grounded structure can be distilled from high-dimensional perceptual inputs. Rather than relying on dense reconstruction alone, this line of work explores sparse internal representations that preserve behavioral relevance while reducing computational burden.
- Cognitive Skill Templates
I explore structured action representations that combine symbolic abstraction with learned policies. The goal is to support contextual reasoning, rapid adaptation, and robust recovery in complex and uncertain human–robot interaction scenarios.
- Intent Autoregression & Spatiotemporal Co-evolution
I investigate how autoregressive intent representations can support future-oriented reasoning in sparse latent spaces. This formulation aims to enable long-horizon spatiotemporal planning while remaining more efficient and interpretable than dense future rollout approaches.
