News
-
Feb 2026
-
Feb 2026
Advanced Robotics & Embodied AI
Eight years managing large-scale construction projects convinced me that the industry's labor shortages and stagnant productivity share a common cause: physical production still depends on manual experience, even as digital design and planning have advanced far ahead of physical execution. My research closes this last-mile gap by advancing embodied intelligence for construction production — enabling robots to perceive physical conditions, understand design intent, and autonomously execute assembly amid variation, uncertainty, and accumulated error, rather than requiring perfectly standardized components and environments.
Using industrialized wood-frame construction as a testbed, I have built an integrated pipeline spanning component perception, digital-model updating, robotic planning, and physical assembly, validated on ABB IRB 6620 industrial robots. My work progresses through vision-driven adaptive control, learning-based assembly-level perception, and long-horizon planning with reinforcement learning — while exploring how Vision-Language-Action (VLA) models can be integrated with deterministic closed-loop control to combine semantic generalization with industrial-grade precision and safety.
Featured Projects
Randomized Single-Stud Grasping & Fixed-Target Placement Trials
High-Precision Real-Time Computer Vision 6-DoF Pose Estimation Algorithm Demo
Education
-
University of Florida · Gainesville, FL Expected May 2027
-
University of Florida · Gainesville, FL Dec 2023
-
Southwest Jiaotong University · Chengdu, China Jun 2013
Awards & Honors
Research Interests
- Embodied AI for Construction Production: Enabling robots to perceive physical conditions, understand design intent, and autonomously execute assembly amid variation, uncertainty, and error accumulation.
- 6D Pose Estimation for Robotic Assembly: Progressing from CAD-free geometric reconstruction to learning-enhanced pose refinement — combining learned coarse initialization with geometric registration for robust, high-precision estimation under occlusion and clutter.
- Perception Benchmarks & Assembly Metrics: Building high-resolution 6D pose-estimation datasets and assembly-oriented evaluation metrics, such as End-Vertex Distance, to quantify interface-level placement accuracy.
- Closed-loop Visual Servoing: Adaptive control laws and real-time feedback loops for precision-critical robotic assembly and construction-scale manipulation.
- Long-Horizon Assembly Planning: Structured assembly-graph representations with graph neural networks and reinforcement learning to sequence operations and recover from failures.
- Digital Twins & Sim-to-Real Pipelines: High-fidelity Isaac Sim/Lab environments for training, evaluating, and safely transferring robotic policies to industrial workstations.
- General-Purpose Manipulation & VLA Models: Evaluating vision-language-action models (e.g., π₀, ACT) to identify their potential and limitations for construction tasks and industrial manipulators.
Selected Publications
Advancing Robotic Automation in Wood-Framed Construction Using Vision-Driven Adaptive Control
A Framework for Automated Quality Control of Wood-Framed Panels in Robotic-Based Manufacturing Using Computer Vision and Deep Learning
Quantitative Analysis of the Role of Robotics in Promoting Industrialized Construction for Resilient Post-Disaster Recovery: Exploratory Research
Digital Twin-Enabled Adaptive Control for Perception-Driven Robotic Assembly in Industrialized Construction
Bridging the Gap: A Framework for Robotics Education in Industrialized Construction
Teaching
-
BCN4612C — Construction Estimating 2
-
BCN5905 — Advanced Construction Technology (Autodesk-Funded Course)
Contact
I welcome collaboration and invited talk opportunities.