rlawlsdn1130 / Study

Note

---

• Locomotion

• Fall Recovery

• Generate Terrain

---

TODO

• Adversarial Learning
• https://rain-bow.tistory.com/entry/%EC%A0%81%EB%8C%80%EC%A0%81-%EA%B3%B5%EA%B2%A9Adversarial-Attack-FGSMPGD

• HR robot model

• Locomotion in generated custom terrain

• Debug gathering height map information in generated custom terrain
• Add 6d commands to low-level policy
• Add spherical scans to high-level policy

Reference

• Learning to walk in confined spaces using 3D representation
• IsaacGym point-cloud example
• Nvidia / wrap
• parkour-ANYmal
• DTC-ANYmal
• extreme-parkour
• parkour

• Learning manipulation and locomotion

• Fall recovery with manipulator
• Fall recovery using exteroception observation

Reference

• Pedipulate-ANYmal
• Loco manipulation-ANYmal
• Fall recovery-ANYmal
• Deep-Whole-Body-Control
• DexterousHands

• Multi-agent DRL

• Create URDF that includes two robots and stretcher, DOF : -
• Mapping using multi robot

Reference

• openrl
• MQE
• sledge
• CMPC