unilab.tasks.locomotion.go2w.manager_terms.Go2WMixedAction¶
- class unilab.tasks.locomotion.go2w.manager_terms.Go2WMixedAction[source]¶
Bases:
ActionTermConvert one community action term into Go2W motor torques per substep.
- Parameters:
cfg (
Go2WMixedActionCfg)env (
ManagerBasedRlEnv)
Methods
__init__(cfg, env)process_actions(actions)reset([env_ids])Resets the manager term.
set_motor_gain_multipliers(env_ids, ...)Attributes
Clipped policy action, matching the legacy observable action buffer.
Whether
apply_actionsneeds state produced by the previous physics substep.-
requires_substep_state_feedback:
ClassVar[bool] = True¶ Whether
apply_actionsneeds state produced by the previous physics substep.Falsedeclares that the generated control is invariant across all physics substeps in one control step, so the env may apply it once before a batched backend step. State-feedback terms must override this declaration withTrue.
-
cfg:
Go2WMixedActionCfg¶
- __init__(cfg, env)[source]¶
- Parameters:
cfg (
Go2WMixedActionCfg)env (
ManagerBasedRlEnv)