| name | Sim-to-Real Transfer Skill |
| description | Techniques for minimizing simulation-to-reality gap and validating transfer |
| slug | sim-to-real |
| category | Validation |
| allowed-tools | ["Bash","Read","Write","Edit","Glob","Grep"] |
| graph | {"domains":["domain:robotics"],"specializations":["specialization:robotics-simulation"],"skillAreas":["skill-area:motion-planning","skill-area:sensor-fusion"],"roles":["role:research-engineer"]} |
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
