import warnings
import numpy as np
from scipy.stats import truncnorm
from ...random_component import RandomComponent
[docs]class MechanicalLoad(RandomComponent):
"""The MechanicalLoad is the base class for all the mechanical systems attached
to the electrical motors rotor.
It contains an mechanical ode system as well as the state names, limits and
nominal values of the mechanical quantities. The only required state is
'omega' as the rotational speed of the motor shaft in rad/s.
ConstantSpeedLoad can be initialized with the initializer as an
class parameter by instantiation. ExternalSpeedLoad takes the first value
of the SpeedProfile as initial value.
Initialization is given by initializer(dict). Can be a constant state value
or random value in given interval.
dict should be like:
{ 'states'(dict): with state names and initial values
'interval'(array like): boundaries for each state
(only for random init), shape(num states, 2)
'random_init'(str): 'uniform' or 'normal'
'random_params(tuple): mue(float), sigma(int)
Example initializer(dict) for constant initialization:
{ 'states': {'omega': 16.0}}
Example initializer(dict) for random initialization:
{ 'random_init': 'normal'}
"""
@property
def j_total(self):
"""
Returns:
float: Total moment of inertia affecting the motor shaft.
"""
return self._j_total
@property
def state_names(self):
"""
Returns:
list(str): Names of the states in the mechanical-ODE.
"""
return self._state_names
@property
def limits(self):
"""
Returns:
dict(float): Mapping of the motor states to their limit values.
"""
return self._limits
@property
def nominal_values(self):
"""
Returns:
dict(float): Mapping of the motor states to their nominal values
"""
return self._nominal_values
@property
def initializer(self):
"""
Returns:
dict: The motors initial state and additional initializer parameters
"""
return self._initializer
OMEGA_IDX = 0
#: Parameter indicating if the class is implementing the optional jacobian function
HAS_JACOBIAN = False
#: _default_initial_state(dict): Default initial motor-state values
_default_initializer = {}
def __init__(self, state_names=None, j_load=0.0, load_initializer=None):
"""
Args:
state_names(list(str)): List of the names of the states in the mechanical-ODE.
j_load(float): Moment of inertia of the load affecting the motor shaft.
"""
RandomComponent.__init__(self)
self._j_total = self._j_load = j_load
self._state_names = list(state_names or ["omega"])
self._limits = {}
self._nominal_values = {}
load_initializer = load_initializer or {}
self._initializer = self._default_initializer.copy()
self._initializer.update(load_initializer)
self._initial_states = self._initializer.get("states", {state: 0.0 for state in self._state_names})
[docs] def initialize(self, state_space, state_positions, nominal_state, **__):
"""Initializes the state of the load on an episode start.
Values can be given as a constant or sampled random out of a statistical distribution. Initial value is in
range of the nominal values or a given interval.
Args:
nominal_state(list): nominal values for each state given from physical system
state_space(gymnasium.spaces.Box): normalized state space boundaries
state_positions(dict): indexes of system states
"""
# for order and organization purposes
interval = self._initializer["interval"]
random_dist = self._initializer["random_init"]
random_params = self._initializer["random_params"]
if isinstance(nominal_state, (list, np.ndarray)):
nominal_state = np.asarray(nominal_state, dtype=float)
elif isinstance(self._nominal_values, dict):
nominal_state = [nominal_state[state] for state in self._initial_states.keys()]
nominal_state = np.asarray(nominal_state)
# setting nominal values as interval limits
state_idx = [state_positions[state] for state in self._initial_states.keys()]
upper_bound = nominal_state[state_idx]
lower_bound = upper_bound * np.asarray(state_space.low, dtype=float)[state_idx]
# clip nominal boundaries to user defined
if interval is not None:
lower_bound = np.clip(lower_bound, a_min=np.asarray(interval, dtype=float).T[0], a_max=None)
upper_bound = np.clip(upper_bound, a_min=None, a_max=np.asarray(interval, dtype=float).T[1])
else:
pass
# random initialization for each load state (omega)
if random_dist is not None:
if random_dist == "uniform":
initial_value = (upper_bound - lower_bound) * self.random_generator.uniform(
size=len(self._initial_states.keys())
) + lower_bound
random_states = {state: initial_value[idx] for idx, state in enumerate(self._initial_states.keys())}
self._initial_states.update(random_states)
elif random_dist in ["normal", "gaussian"]:
# specific input or middle of interval
mue = random_params[0] or (upper_bound - lower_bound) / 2 + lower_bound
sigma = random_params[1] or 1
a = (lower_bound - mue) / sigma
b = (upper_bound - mue) / sigma
initial_value = truncnorm.rvs(
a,
b,
loc=mue,
scale=sigma,
size=(len(self._initial_states.keys())),
random_state=self.seed_sequence.pool[0],
)
random_states = {state: initial_value[idx] for idx, state in enumerate(self._initial_states.keys())}
self._initial_states.update(random_states)
else:
raise NotImplementedError
# constant initialization for each motor state (current, epsilon)
elif self._initial_states is not None:
initial_value = np.atleast_1d(list(self._initial_states.values()))
# check init_value meets interval boundaries
if (lower_bound <= initial_value).all() and (initial_value <= upper_bound).all():
initial_states_ = {state: initial_value[idx] for idx, state in enumerate(self._initial_states.keys())}
self._initial_states.update(initial_states_)
else:
raise Exception("Initialization Value have to be in nominal boundaries")
else:
raise Exception("No matching Initialization Case")
[docs] def reset(self, state_space, state_positions, nominal_state, **__):
"""
Reset the motors state to a new initial state. (Default 0)
Args:
nominal_state(list): nominal values for each state given from
physical system
state_space(gymnasium.Box): normalized state space boundaries
state_positions(dict): indexes of system states
Returns:
numpy.ndarray(float): The initial motor states.
"""
self.next_generator()
if self._initializer:
self.initialize(state_space, state_positions, nominal_state)
return np.asarray(list(self._initial_states.values()))
else:
return np.zeros_like(self._state_names, dtype=float)
[docs] def set_j_rotor(self, j_rotor):
"""
Args:
j_rotor(float): The moment of inertia of the rotor shaft of the motor.
"""
self._j_total += j_rotor
[docs] def mechanical_ode(self, t, mechanical_state, torque):
"""
Calculation of the derivatives of the mechanical-ODE for each of the mechanical states.
Args:
t(float): Current time of the system.
mechanical_state(ndarray(float)): Current state of the mechanical system.
torque(float): Generated input torque by the electrical motor.
Returns:
ndarray(float): Derivatives of the mechanical state for the given input torque.
"""
raise NotImplementedError
[docs] def mechanical_jacobian(self, t, mechanical_state, torque):
"""
Calculation of the jacobians of the mechanical-ODE for each of the mechanical state.
Overriding this method is optional for each subclass. If it is overridden, the parameter HAS_JACOBIAN must also
be set to True. Otherwise, the jacobian will not be called.
Args:
t(float): Current time of the system.
mechanical_state(ndarray(float)): Current state of the mechanical system.
torque(float): Generated input torque by the electrical motor.
Returns:
Tuple(ndarray, ndarray):
[0]: Derivatives of the mechanical_state-odes over the mechanical_states shape:(states x states)
[1]: Derivatives of the mechanical_state-odes over the torque shape:(states,)
"""
pass
[docs] def get_state_space(self, omega_range):
"""
Args:
omega_range(Tuple(int,int)): Lower and upper values the motor can generate for omega normalized to (-1, 1)
Returns:
Tuple(dict,dict): Lowest and highest possible values for all states normalized to (-1, 1)
"""
return {"omega": omega_range[0]}, {"omega": omega_range[1]}