Yang Lexing · Homepage
Building intelligent methods
for real engineering systems
An undergraduate in Intelligent Manufacturing, working on reinforcement learning, industrial vision measurement, and data-driven optimization for complex engineering systems.
Research Orientation
Learning should do more than fit data — it should serve real systems.
I am interested in intelligent methods that connect data, physical reasoning, and engineering decision-making, making learning models genuinely useful for modeling, optimization, and control.
About
Where engineering, learning, and physical reasoning meet.
My interests lie at the intersection of machine learning and engineering systems. I care not only about fitting models to data, but also about making them useful for modeling, optimization, and control in complex real-world systems while preserving physical meaning and engineering interpretability.
Recently, I have been working on reinforcement learning for control, industrial vision measurement, and screw compressor rotor profile optimization, with the goal of bringing learning-based methods closer to real engineering applications.
Selected Projects
What I am building now.
These projects reflect my current interest in solving real engineering problems with computational intelligence.
Optimization · Simulation · Surrogate Modeling
Screw Compressor Rotor Optimization
Physics-aware learning for high-dimensional engineering design
Built a closed-loop pipeline that integrates parameter generation, SCORG simulation, and surrogate modeling to accelerate rotor profile design and performance evaluation.
Vision · Geometry · Measurement
Industrial Vision Measurement
Dimension recovery through perspective compensation and inverse modeling
Developed a measurement pipeline combining plane reconstruction, view normalization, object detection, and neural inverse solving for industrial dimension recovery with low-cost depth sensing.
Reinforcement Learning · Robotics · Control
Reinforcement Learning for Autonomous Systems
Learning-based control with physical consistency and robustness
Exploring reinforcement learning for autonomous driving and robotic control, focusing on multimodal observations, policy robustness, and sim-to-real oriented system design.
Research Interest
Learning under physical constraints
I am especially interested in methods that improve performance while remaining meaningful for engineering analysis and decision-making.
Current Focus
RL, vision, and optimization
My recent work spans reinforcement learning control, industrial visual measurement, and simulation-driven design optimization.
Long-term Goal
Intelligent systems for industry
I hope to build systems with perception, decision-making, and continuous optimization capabilities for real industrial environments.