SKRL config for LAV2 attitude control with rotation-manifold policy heads.
Control modes and their action dimensions::
cmd_ctatt_euler → 4D [thrust, roll, pitch, yaw] cmd_ctatt_rotvec → 4D [thrust, rx, ry, rz] cmd_ctatt_quat → 5D [thrust, qx, qy, qz, qw] cmd_ctatt_rotmat → 10D [thrust, r00..r22]
Projection layers ensure the rotation component of the policy output lies on a valid SO(3) manifold. Available variants per rotation type:
============== ============================================================
Euler / rotvec EulerProjection (tanh clamp)
Quaternion QuatProjection (tanh + L2-normalize)
QuatPlusProjection (tanh + sigmoid w + L2-normalize)
QuatOffsetProjection (offset by identity quat)
QuatPlusOffsetProjection (offset + sigmoid w)
Rotation RotmatProjection (tanh + SVD-orthogonalize)
matrix RotmatOffsetProjection (offset by identity matrix)
============== ============================================================
The model class accepts projection_cls for experiments that need a
non-default projection. Projection classes are re-exported here for
compatibility with existing configs.
Functions:
| Name | Description |
|---|---|
get_model_class |
Return the default attitude model class. |
get_agent_cfg |
Return the PPO agent configuration. |
get_agent_class |
Return the PPO agent class. |
get_memory_class |
Return the memory (rollout buffer) class. |
get_memory_cfg |
Return the memory configuration. |
get_trainer_class |
Return the trainer class. |
get_trainer_cfg |
Return the trainer configuration. |
get_model_class() -> Any
Return the default attitude model class.
Pass projection_cls to the returned class constructor to select a
non-default projection variant such as :class:QuatPlusProjection or
:class:RotmatOffsetProjection.
get_agent_cfg() -> Any
Return the PPO agent configuration.
get_agent_class() -> Any
Return the PPO agent class.
get_memory_class() -> Any
Return the memory (rollout buffer) class.
get_memory_cfg() -> Any
Return the memory configuration.
get_trainer_class() -> Any
Return the trainer class.
get_trainer_cfg() -> Any
Return the trainer configuration.