About PSL Tech
PSL Tech is the engineering blog of PHYSICS SiM Lab.
PHYSICS SiM Lab builds training infrastructure for Physical AI using physics-based simulation and synthetic data. We connect data generation, environment construction, and validation so robots can learn the experiences they need in the real world through accurate and varied simulation environments.
What we build
- Physically validated synthetic data — We generate vision, joint trajectories, and tactile and contact forces as episodes with shared timestamps. Results that pass physical validation are rendered and provided as training data.
- Simulation infrastructure — We build digital twin environments, simulators supporting ROS and VR teleoperation, and Sim-Ready assets with physical properties and collision geometry.
- General-purpose robot control — Through the UDM (Unified Dextrous Manipulator) framework, we develop grasping and motion-planning technology for different robots and hands.
- Validation and evaluation — We research frameworks and benchmarks for evaluating the physical consistency of data, precise robot motion, and world-model predictions and spatial representations.
What we document here
We share implementation work on simulation solvers and robot control, reproducible experiments, and validation of performance and limitations. Alongside results, we document operating conditions and the reasons behind our choices, aiming to make these engineering notes useful in other environments.
Get in touch
This introduction is based on the official website. Subscribe to new posts through the English RSS feed.