deluca.agents.DRC¶
-
class
deluca.agents.DRC(*args, **kwargs)[source]¶ Public Data Attributes:
Inherited from
JaxObjectnameattrsPublic Methods:
__init__(A, B[, C, K, cost_fn, m, h, …])Description: Initialize the dynamics of the model.
__call__(obs)Description: Return the action based on current state and internal parameters.
get_action(obs)Description: get action from state.
update(obs, u)update_noise(obs)update_params(obs, u)Description: update agent internal state.
Inherited from
Agent__init_subclass__(*args, **kwargs)For avoiding a decorator for each subclass
__call__(obs)Description: Return the action based on current state and internal parameters.
reset()feed(reward)Inherited from
JaxObject__new__(cls, *args, **kwargs)For avoiding super().__init__()
__init_subclass__(*args, **kwargs)For avoiding a decorator for each subclass
__str__()Return str(self).
__setattr__(key, val)Implement setattr(self, name, value).
save(path)load(path)throw(err, msg)
-
__call__(obs: jax._src.numpy.lax_numpy.ndarray) → jax._src.numpy.lax_numpy.ndarray[source]¶ Description: Return the action based on current state and internal parameters.
- Parameters
state (jnp.ndarray) – current state
- Returns
action to take
- Return type
jnp.ndarray
-
__init__(A: jax._src.numpy.lax_numpy.ndarray, B: jax._src.numpy.lax_numpy.ndarray, C: jax._src.numpy.lax_numpy.ndarray = None, K: jax._src.numpy.lax_numpy.ndarray = None, cost_fn: Callable[[jax._src.numpy.lax_numpy.ndarray, jax._src.numpy.lax_numpy.ndarray], numbers.Real] = None, m: int = 10, h: int = 50, lr_scale: numbers.Real = 0.03, decay: bool = True, RM: int = 1000, seed: int = 0) → None[source]¶ Description: Initialize the dynamics of the model.
- Parameters
A (jnp.ndarray) – system dynamics
B (jnp.ndarray) – system dynamics
C (jnp.ndarray) – system dynamics
Q (jnp.ndarray) – cost matrices (i.e. cost = x^TQx + u^TRu)
R (jnp.ndarray) – cost matrices (i.e. cost = x^TQx + u^TRu)
K (jnp.ndarray) – Starting policy (optional). Defaults to LQR gain.
start_time (int) –
cost_fn (Callable[[jnp.ndarray, jnp.ndarray], Real]) –
H (postive int) – history of the controller
HH (positive int) – history of the system
lr_scale (Real) –
decay (boolean) –
seed (int) –
-