Sim to real

Skill a5c-ai/babysitter/library/specializations/robotics-simulation/skills/sim-to-real

Techniques for minimizing simulation-to-reality gap and validating transferFrom its SKILL.md

Install

npx -y skills add a5c-ai/babysitter --skill sim-to-real

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

1.7 KB, 236 tokens by cl100k_base, as published. Nobody here has run it

Sim-to-Real Transfer Skill

Overview

Expert skill for bridging the simulation-to-reality gap through domain randomization, system identification, and transfer validation techniques.

Capabilities

Implement domain randomization (physics, appearance, dynamics)
Configure system identification for simulation parameters
Set up adaptive domain randomization
Implement domain adaptation techniques
Configure noise injection for robust policies
Set up reality gap metrics and monitoring
Implement progressive network transfer
Configure latency simulation
Set up sensor noise modeling
Implement hardware-in-the-loop validation

Target Processes

sim-to-real-validation.js
digital-twin-development.js
rl-robot-control.js
field-testing-validation.js

Dependencies

Simulation environments (Gazebo, Isaac Sim)
Physical robot access
System identification tools

Usage Context

This skill is invoked when processes require transferring simulation-trained models or behaviors to real robot hardware with minimal performance degradation.

Output Artifacts

Domain randomization configurations
System identification results
Reality gap analysis reports
Transfer validation metrics
Sensor noise models
Calibrated simulation parameters

What ships with it: 1 file

648 B alongside SKILL.md

README.md648 B