def __init__(self):
# LQR to stabilize pole in the starting phase
self.lqr_controller = controller_lqr()
self.horizon = mpc_horizon
self.Predictor = predictor(horizon=self.horizon, dt=DT)
self.rnn_eval_time = []
self.predictor_time = []
self.nfun = []
# I do the norm and unnorm unnecessarilly!
# You need only to scale once!
"""
Get configured do-mpc modules:
"""
# State of the cart
self.s = create_cartpole_state() # s like state
self.target_position = 0.0
self.Q_hat = np.zeros(self.horizon) # MPC prediction of future control inputs
self.Q_hat0 = np.random.normal(0.0, 0.1 ,self.horizon) # initial guess for future control inputs to be predicted
self.Q_previous = 0.0
self.angle_cost = lambda angle: angle ** 2
self.angle_sin_cost = lambda angle_sin: angle_sin ** 2
self.angleD_cost = lambda angleD: angleD ** 2
self.position_cost = lambda position: (position - self.target_position) ** 2
self.positionD_cost = lambda positionD: positionD ** 2
self.Q_bounds = scipy.optimize.Bounds(lb=-1.0, ub=1.0)
self.initial_state = create_cartpole_state()
self.prediction_features_names = cartpole_state_indices_to_varnames(range(len(self.s)))
self.predictions = np.zeros((self.horizon + 1, len(self.prediction_features_names) + 1))
# Array keeping individual costs for every timestep of prediction
self.costs_names = ['angle_cost', 'angle_sin_cost', 'angleD_cost', 'position_cost', 'positionD_cost']
self.cost_array = np.zeros((self.horizon + 1, len(self.costs_names) + 1))
# Numbers of samples which should be bridged with output from a P controller
# To give RNN time to settle
self.warm_up_len = 0
self.sample_counter = 0 # Counting samples for above aim
def __init__(self):
"""
Get configured do-mpc modules:
"""
# State of the cart
self.s = create_cartpole_state() # s like state
self.target_position = 0.0
self.mpc_horizon = mpc_horizon
self.dt = dt_mpc_simulation
self.yp_hat = np.zeros(self.mpc_horizon,
dtype=object) # MPC prediction of future states
self.Q_hat = np.zeros(
self.mpc_horizon) # MPC prediction of future control inputs
self.Q_hat0 = np.zeros(
self.mpc_horizon
) # initial guess for future control inputs to be predicted
self.Q_previous = 0.0
self.E_kin_cart = lambda s: (s[POSITIOND_IDX] / v_max)**2
self.E_kin_pol = lambda s: (s[ANGLED_IDX] / (2 * np.pi))**2
self.E_pot_cost = lambda s: 1 - np.cos(s[ANGLE_IDX])
self.distance_difference = lambda s: ((
(s[POSITION_IDX] - self.target_position) / TrackHalfLength))**2
self.Q_bounds = [(-1, 1)] * self.mpc_horizon
def get_prediction_for_testing_gui_from_euler(a,
dataset,
dt_sampling,
dt_sampling_by_dt_fine=1):
# region In either case testing is done on a data collected offline
output_array = np.zeros(shape=(a.test_max_horizon + 1, a.test_len,
len(a.features) + 1))
Q = dataset['Q'].to_numpy()
Q_array = [Q[i:-a.test_max_horizon + i] for i in range(a.test_max_horizon)]
Q_array = np.vstack(Q_array)
output_array[:-1, :, -1] = Q_array
print('Calculating predictions...')
FEATURES_INDICES = cartpole_state_varnames_to_indices(a.features)
for timestep in trange(a.test_len):
state_dict = dataset.loc[dataset.index[[timestep]], :].to_dict(
'records')[0]
s = create_cartpole_state(state_dict)
output_array[0, timestep, :-1] = s[FEATURES_INDICES]
# Progress max_horison steps
# save the data for every step in a third dimension of an array
for i in range(0, a.test_max_horizon):
Q = Q_array[i, timestep]
u = Q2u(Q)
for _ in range(dt_sampling_by_dt_fine):
angleDD, positionDD = cartpole_ode_numba(s, u)
t_step = (dt_sampling / float(dt_sampling_by_dt_fine))
# Calculate next state
s[POSITION_IDX] += s[POSITIOND_IDX] * t_step
s[POSITIOND_IDX] += positionDD * t_step
s[ANGLE_IDX] += s[ANGLED_IDX] * t_step
s[ANGLED_IDX] += angleDD * t_step
s[ANGLE_IDX] = \
wrap_angle_rad(s[ANGLE_IDX])
# Append s to outputs matrix
s[ANGLE_COS_IDX] = np.cos(s[ANGLE_IDX])
s[ANGLE_SIN_IDX] = np.sin(s[ANGLE_IDX])
output_array[i + 1, timestep, :-1] = s[FEATURES_INDICES]
output_array = np.swapaxes(output_array, 0, 1)
# time_axis is a time axis for ground truth
return output_array
def __init__(self):
"""Random number generator"""
SEED = config["controller"]["mppi"]["SEED"]
if SEED == "None":
SEED = int(
(datetime.now() - datetime(1970, 1, 1)).total_seconds() *
1000.0) # Fully random
self.rng_mppi = Generator(SFC64(SEED))
self.rng_mppi_rnn = Generator(
SFC64(SEED * 2)
) # There are some random numbers used at warm up of rnn only. Separate rng prevents a shift
global dd_weight, ep_weight, ekp_weight, ekc_weight, cc_weight
dd_weight = dd_weight * (1 +
dd_noise * self.rng_mppi.uniform(-1.0, 1.0))
ep_weight = ep_weight * (1 +
ep_noise * self.rng_mppi.uniform(-1.0, 1.0))
ekp_weight = ekp_weight * (
1 + ekp_noise * self.rng_mppi.uniform(-1.0, 1.0))
ekc_weight = ekc_weight * (
1 + ekc_noise * self.rng_mppi.uniform(-1.0, 1.0))
cc_weight = cc_weight * (1 +
cc_noise * self.rng_mppi.uniform(-1.0, 1.0))
# State of the cart
self.s = create_cartpole_state()
self.target_position = 0.0
self.rho_sqrt_inv = 0.01
self.iteration = -1
self.control_enabled = True
self.s_horizon = np.zeros((), dtype=np.float32)
self.u = np.zeros((mpc_samples), dtype=np.float32)
self.u_prev = np.zeros_like(self.u, dtype=np.float32)
self.delta_u = np.zeros((num_rollouts, mpc_samples), dtype=np.float32)
self.S_tilde_k = np.zeros((num_rollouts), dtype=np.float32)
self.wash_out_len = WASH_OUT_LEN
self.warm_up_countdown = self.wash_out_len
try:
from Controllers.controller_lqr import controller_lqr
self.auxiliary_controller_available = True
self.auxiliary_controller = controller_lqr()
except ModuleNotFoundError:
self.auxiliary_controller_available = False
self.auxiliary_controller = None
self.auxiliary_controller_available = False
def cartpole_integration(s, angleDD, positionDD, dt):
"""
Simple single step integration of CartPole state by dt
Takes state as numpy array.
:param s: state of the CartPole (position, positionD, angle, angleD must be set). Array order follows global definition.
:param dt: time step by which the CartPole state should be integrated
"""
s_next = create_cartpole_state()
s_next[POSITION_IDX] = s[POSITION_IDX] + s[POSITIOND_IDX] * dt
s_next[POSITIOND_IDX] = s[POSITIOND_IDX] + positionDD * dt
s_next[ANGLE_IDX] = s[ANGLE_IDX] + s[ANGLED_IDX] * dt
s_next[ANGLED_IDX] = s[ANGLED_IDX] + angleDD * dt
return s_next
def finish_thread(self):
self.CartPoleInstance.use_pregenerated_target_position = False
self.initial_state = create_cartpole_state()
self.initial_position_slider.setValue(0)
self.initial_angle_slider.setValue(0)
self.CartPoleInstance.s = self.initial_state
# Some controllers may collect they own statistics about their usage and print it after experiment terminated
if self.simulator_mode != 'Replay':
try:
self.CartPoleInstance.controller.controller_report()
except:
pass
if self.show_experiment_summary:
self.w_summary = SummaryWindow(
summary_plots=self.CartPoleInstance.summary_plots)
# Reset variables and redraw the figures
self.reset_variables(0)
# Draw figures
self.CartPoleInstance.draw_constant_elements(self.fig, self.fig.AxCart,
self.fig.AxSlider)
self.canvas.draw()
# Enable back all elements of GUI:
self.cb_save_history.setEnabled(True)
self.cb_show_experiment_summary.setEnabled(True)
for rb in self.rbs_simulator_mode:
rb.setEnabled(True)
for rb in self.rbs_controllers:
rb.setEnabled(True)
self.start_or_stop_action = "START!" # What should happen when "START! / STOP!" is pushed NEXT time
The convention:
Pole upright position defines 0 angle
Cart movement to the right is positive
Clockwise angle rotation is defined as negative
Required angle convention for CartPole GUI: CLOCK-NEG
"""
ANGLE_CONVENTION = 'CLOCK-NEG'
"""Defines if a clockwise angle change is negative ('CLOCK-NEG') or positive ('CLOCK-POS')
The 0-angle state is always defined as pole in upright position. This currently cannot be changed
"""
# Create initial state vector
s0 = create_cartpole_state()
def _cartpole_ode(ca,
sa,
angleD,
positionD,
u,
k=k,
M=M,
m=m,
g=g,
J_fric=J_fric,
M_fric=M_fric,
L=L):
"""
def __init__(self, *args, **kwargs):
super(MainWindow, self).__init__(*args, **kwargs)
# region Create CartPole instance and load initial settings
# Create CartPole instance
self.initial_state = create_cartpole_state()
self.CartPoleInstance = CartPole(initial_state=self.initial_state)
# Set timescales
self.CartPoleInstance.dt_simulation = dt_simulation
self.CartPoleInstance.dt_controller = controller_update_interval
self.CartPoleInstance.dt_save = save_interval
# set other settings
self.CartPoleInstance.set_controller(controller_init)
self.CartPoleInstance.stop_at_90 = stop_at_90_init
self.set_random_experiment_generator_init_params()
# endregion
# region Decide whether to save the data in "CartPole memory" or not
self.save_history = save_history_init
self.show_experiment_summary = show_experiment_summary_init
if self.save_history or self.show_experiment_summary:
self.CartPoleInstance.save_data_in_cart = True
else:
self.CartPoleInstance.save_data_in_cart = False
# endregion
# region Other variables initial values as provided in gui_default_parameters.py
# Start user controlled experiment/ start random experiment/ load and replay - on start button
self.simulator_mode = simulator_mode_init
self.slider_on_click = slider_on_click_init # Update slider on click/update slider while hoovering over it
self.speedup = speedup_init # Default simulation speed-up
# endregion
# region Initialize loop-timer
# This timer allows to relate the simulation time to user time
# And (if your computer is fast enough) run simulation
# slower or faster than real-time by predefined factor (speedup)
self.looper = loop_timer(
dt_target=(self.CartPoleInstance.dt_simulation / self.speedup))
# endregion
# region Variables controlling the state of various processes (DO NOT MODIFY)
self.terminate_experiment_or_replay_thread = False # True: gives signal causing thread to terminate
self.pause_experiment_or_replay_thread = False # True: gives signal causing the thread to pause
self.run_set_labels_thread = True # True if gauges (labels) keep being repeatedly updated
# Stop threads by setting False
# Flag indicating if the "START! / STOP!" button should act as start or as stop when pressed.
# Can take values "START!" or "STOP!"
self.start_or_stop_action = "START!"
# Flag indicating whether the pause button should pause or unpause.
self.pause_or_unpause_action = "PAUSE"
# Flag indicating that saving of experiment recording to csv file has finished
self.experiment_or_replay_thread_terminated = False
self.user_time_counter = 0 # Measures the user time
# Slider instant value (which is draw in GUI) differs from value saved in CartPole instance
# if the option updating slider "on-click" is enabled.
self.slider_instant_value = self.CartPoleInstance.slider_value
self.noise = 'OFF'
self.CartPoleInstance.NoiseAdderInstance.noise_mode = self.noise
# endregion
# region Create GUI Layout
# region - Create container for top level layout
layout = QVBoxLayout()
# endregion
# region - Change geometry of the main window
self.setGeometry(300, 300, 2500, 1000)
# endregion
# region - Matplotlib figures (CartPole drawing and Slider)
# Draw Figure
self.fig = Figure(
figsize=(25, 10)
) # Regulates the size of Figure in inches, before scaling to window size.
self.canvas = FigureCanvas(self.fig)
self.fig.AxCart = self.canvas.figure.add_subplot(211)
self.fig.AxSlider = self.canvas.figure.add_subplot(212)
self.fig.AxSlider.set_ylim(0, 1)
self.CartPoleInstance.draw_constant_elements(self.fig, self.fig.AxCart,
self.fig.AxSlider)
# Attach figure to the layout
lf = QVBoxLayout()
lf.addWidget(self.canvas)
# endregion
# region - Radio buttons selecting current controller
self.rbs_controllers = []
for controller_name in self.CartPoleInstance.controller_names:
self.rbs_controllers.append(QRadioButton(controller_name))
# Ensures that radio buttons are exclusive
self.controllers_buttons_group = QButtonGroup()
for button in self.rbs_controllers:
self.controllers_buttons_group.addButton(button)
lr_c = QVBoxLayout()
lr_c.addStretch(1)
for rb in self.rbs_controllers:
rb.clicked.connect(self.RadioButtons_controller_selection)
lr_c.addWidget(rb)
lr_c.addStretch(1)
self.rbs_controllers[self.CartPoleInstance.controller_idx].setChecked(
True)
# endregion
# region - Create central part of the layout for figures and radio buttons and add it to the whole layout
lc = QHBoxLayout()
lc.addLayout(lf)
lc.addLayout(lr_c)
layout.addLayout(lc)
# endregion
# region - Gauges displaying current values of various states and parameters (time, velocity, angle,...)
# First row
ld = QHBoxLayout()
# User time
self.labTime = QLabel("User's time (s): ")
self.timer = QTimer()
self.timer.setInterval(100) # Tick every 1/10 of the second
self.timer.timeout.connect(self.set_user_time_label)
self.timer.start()
ld.addWidget(self.labTime)
# Speed, angle, motor power (Q)
self.labSpeed = QLabel('Speed (m/s):')
self.labAngle = QLabel('Angle (deg):')
self.labMotor = QLabel('')
self.labTargetPosition = QLabel('')
ld.addWidget(self.labSpeed)
ld.addWidget(self.labAngle)
ld.addWidget(self.labMotor)
ld.addWidget(self.labTargetPosition)
layout.addLayout(ld)
# Second row of labels
# Simulation time, Measured (real) speed-up, slider-value
ld2 = QHBoxLayout()
self.labTimeSim = QLabel('Simulation Time (s):')
ld2.addWidget(self.labTimeSim)
self.labSpeedUp = QLabel('Speed-up (measured):')
ld2.addWidget(self.labSpeedUp)
self.labSliderInstant = QLabel('')
ld2.addWidget(self.labSliderInstant)
layout.addLayout(ld2)
# endregion
# region - Buttons "START!" / "STOP!", "PAUSE", "QUIT"
self.bss = QPushButton("START!")
self.bss.pressed.connect(self.start_stop_button)
self.bp = QPushButton("PAUSE")
self.bp.pressed.connect(self.pause_unpause_button)
bq = QPushButton("QUIT")
bq.pressed.connect(self.quit_application)
lspb = QHBoxLayout() # Sub-Layout for Start/Stop and Pause Buttons
lspb.addWidget(self.bss)
lspb.addWidget(self.bp)
# endregion
# region - Sliders setting initial state and buttons for kicking the pole
# Sliders setting initial position and angle
lb = QVBoxLayout() # Layout for buttons
lb.addLayout(lspb)
lb.addWidget(bq)
ip = QHBoxLayout() # Layout for initial position sliders
self.initial_position_slider = QSlider(
orientation=Qt.Orientation.Horizontal)
self.initial_position_slider.setRange(
-int(float(1000 * TrackHalfLength)),
int(float(1000 * TrackHalfLength)))
self.initial_position_slider.setValue(0)
self.initial_position_slider.setSingleStep(1)
self.initial_position_slider.valueChanged.connect(
self.update_initial_position)
self.initial_angle_slider = QSlider(
orientation=Qt.Orientation.Horizontal)
self.initial_angle_slider.setRange(-int(float(100 * np.pi)),
int(float(100 * np.pi)))
self.initial_angle_slider.setValue(0)
self.initial_angle_slider.setSingleStep(1)
self.initial_angle_slider.valueChanged.connect(
self.update_initial_angle)
ip.addWidget(QLabel("Initial position:"))
ip.addWidget(self.initial_position_slider)
ip.addWidget(QLabel("Initial angle:"))
ip.addWidget(self.initial_angle_slider)
ip.addStretch(0.01)
# Slider setting latency
self.LATENCY_SLIDER_RANGE_INT = 1000
self.latency_slider = QSlider(orientation=Qt.Orientation.Horizontal)
self.latency_slider.setRange(0, self.LATENCY_SLIDER_RANGE_INT)
self.latency_slider.setValue(
int(self.CartPoleInstance.LatencyAdderInstance.latency *
self.LATENCY_SLIDER_RANGE_INT /
self.CartPoleInstance.LatencyAdderInstance.max_latency))
self.latency_slider.setSingleStep(1)
self.latency_slider.valueChanged.connect(self.update_latency)
ip.addWidget(QLabel("Latency:"))
ip.addWidget(self.latency_slider)
self.labLatency = QLabel('Latency (ms): {:.1f}'.format(
self.CartPoleInstance.LatencyAdderInstance.latency * 1000))
ip.addWidget(self.labLatency)
# Buttons activating noise
self.rbs_noise = []
for mode_name in ['ON', 'OFF']:
self.rbs_noise.append(QRadioButton(mode_name))
# Ensures that radio buttons are exclusive
self.noise_buttons_group = QButtonGroup()
for button in self.rbs_noise:
self.noise_buttons_group.addButton(button)
lr_n = QHBoxLayout()
lr_n.addWidget(QLabel('Noise:'))
for rb in self.rbs_noise:
rb.clicked.connect(self.RadioButtons_noise_on_off)
lr_n.addWidget(rb)
self.rbs_noise[1].setChecked(True)
ip.addStretch(0.01)
ip.addLayout(lr_n)
ip.addStretch(0.01)
# Buttons giving kick to the pole
kick_label = QLabel("Kick pole:")
kick_left_button = QPushButton()
kick_left_button.setText("Left")
kick_left_button.adjustSize()
kick_left_button.clicked.connect(self.kick_pole)
kick_right_button = QPushButton()
kick_right_button.setText("Right")
kick_right_button.adjustSize()
kick_right_button.clicked.connect(self.kick_pole)
ip.addWidget(kick_label)
ip.addWidget(kick_left_button)
ip.addWidget(kick_right_button)
lb.addLayout(ip)
layout.addLayout(lb)
# endregion
# region - Text boxes and Combobox to provide settings concerning generation of random experiment
l_generate_trace = QHBoxLayout()
l_generate_trace.addWidget(QLabel('Random experiment settings:'))
l_generate_trace.addWidget(QLabel('Length (s):'))
self.textbox_length = QLineEdit()
l_generate_trace.addWidget(self.textbox_length)
l_generate_trace.addWidget(QLabel('Turning Points (m):'))
self.textbox_turning_points = QLineEdit()
l_generate_trace.addWidget(self.textbox_turning_points)
l_generate_trace.addWidget(QLabel('Interpolation:'))
self.cb_interpolation = QComboBox()
self.cb_interpolation.addItems(
['0-derivative-smooth', 'linear', 'previous'])
self.cb_interpolation.currentIndexChanged.connect(
self.cb_interpolation_selectionchange)
self.cb_interpolation.setCurrentText(
self.CartPoleInstance.interpolation_type)
l_generate_trace.addWidget(self.cb_interpolation)
layout.addLayout(l_generate_trace)
# endregion
# region - Textbox to provide csv file name for saving or loading data
l_text = QHBoxLayout()
textbox_title = QLabel('CSV file name:')
self.textbox = QLineEdit()
l_text.addWidget(textbox_title)
l_text.addWidget(self.textbox)
layout.addLayout(l_text)
# endregion
# region - Make strip of layout for checkboxes
l_cb = QHBoxLayout()
# endregion
# region - Textbox to provide the target speed-up value
l_text_speedup = QHBoxLayout()
tx_speedup_title = QLabel('Speed-up (target):')
self.tx_speedup = QLineEdit()
l_text_speedup.addWidget(tx_speedup_title)
l_text_speedup.addWidget(self.tx_speedup)
self.tx_speedup.setText(str(self.speedup))
l_cb.addLayout(l_text_speedup)
self.wrong_speedup_msg = QMessageBox()
self.wrong_speedup_msg.setWindowTitle("Speed-up value problem")
self.wrong_speedup_msg.setIcon(QMessageBox.Icon.Critical)
# endregion
# region - Checkboxes
# region -- Checkbox: Save/don't save experiment recording
self.cb_save_history = QCheckBox('Save results', self)
if self.save_history:
self.cb_save_history.toggle()
self.cb_save_history.toggled.connect(self.cb_save_history_f)
l_cb.addWidget(self.cb_save_history)
# endregion
# region -- Checkbox: Display plots showing dynamic evolution of the system as soon as experiment terminates
self.cb_show_experiment_summary = QCheckBox('Show experiment summary',
self)
if self.show_experiment_summary:
self.cb_show_experiment_summary.toggle()
self.cb_show_experiment_summary.toggled.connect(
self.cb_show_experiment_summary_f)
l_cb.addWidget(self.cb_show_experiment_summary)
# endregion
# region -- Checkbox: Block pole if it reaches +/-90 deg
self.cb_stop_at_90_deg = QCheckBox('Stop-at-90-deg', self)
if self.CartPoleInstance.stop_at_90:
self.cb_stop_at_90_deg.toggle()
self.cb_stop_at_90_deg.toggled.connect(self.cb_stop_at_90_deg_f)
l_cb.addWidget(self.cb_stop_at_90_deg)
# endregion
# region -- Checkbox: Update slider on click/update slider while hoovering over it
self.cb_slider_on_click = QCheckBox('Update slider on click', self)
if self.slider_on_click:
self.cb_slider_on_click.toggle()
self.cb_slider_on_click.toggled.connect(self.cb_slider_on_click_f)
l_cb.addWidget(self.cb_slider_on_click)
# endregion
# endregion
# region - Radio buttons selecting simulator mode: user defined experiment, random experiment, replay
# List available simulator modes - constant
self.available_simulator_modes = [
'Slider-Controlled Experiment', 'Random Experiment', 'Replay'
]
self.rbs_simulator_mode = []
for mode_name in self.available_simulator_modes:
self.rbs_simulator_mode.append(QRadioButton(mode_name))
# Ensures that radio buttons are exclusive
self.simulator_mode_buttons_group = QButtonGroup()
for button in self.rbs_simulator_mode:
self.simulator_mode_buttons_group.addButton(button)
lr_sm = QHBoxLayout()
lr_sm.addStretch(1)
lr_sm.addWidget(QLabel('Simulator mode:'))
for rb in self.rbs_simulator_mode:
rb.clicked.connect(self.RadioButtons_simulator_mode)
lr_sm.addWidget(rb)
lr_sm.addStretch(1)
self.rbs_simulator_mode[self.available_simulator_modes.index(
self.simulator_mode)].setChecked(True)
l_cb.addStretch(1)
l_cb.addLayout(lr_sm)
l_cb.addStretch(1)
# endregion
# region - Add checkboxes to layout
layout.addLayout(l_cb)
# endregion
# region - Create an instance of a GUI window
w = QWidget()
w.setLayout(layout)
self.setCentralWidget(w)
self.show()
self.setWindowTitle('CartPole Simulator')
# endregion
# endregion
# region Open controller-specific popup windows
self.open_additional_controller_widget()
# endregion
# region Activate functions capturing mouse movements and clicks over the slider
# This line links function capturing the mouse position on the canvas of the Figure
self.canvas.mpl_connect("motion_notify_event", self.on_mouse_movement)
# This line links function capturing the mouse position on the canvas of the Figure click
self.canvas.mpl_connect("button_press_event", self.on_mouse_click)
# endregion
# region Introducing multithreading
# To ensure smooth functioning of the app,
# the calculations and redrawing of the figures have to be done in a different thread
# than the one capturing the mouse position and running the animation
self.threadpool = QThreadPool()
# endregion
# region Starts a thread repeatedly redrawing gauges (labels) of the GUI
# It runs till the QUIT button is pressed
worker_labels = Worker(self.set_labels_thread)
self.threadpool.start(worker_labels)
# endregion
# region Start animation repeatedly redrawing changing elements of matplotlib figures (CartPole drawing and slider)
# This animation runs ALWAYS when the GUI is open
# The buttons of GUI only decide if new parameters are calculated or not
self.anim = self.CartPoleInstance.run_animation(self.fig)
def __init__(self, dt, intermediate_steps):
self.s = create_cartpole_state()
self.intermediate_steps = intermediate_steps
self.t_step = np.float32(dt / float(self.intermediate_steps))
def __init__(self, dt, intermediate_steps):
self.s = tf.convert_to_tensor(create_cartpole_state())
self.intermediate_steps = tf.convert_to_tensor(intermediate_steps, dtype=tf.int32)
self.t_step = tf.convert_to_tensor(dt / float(self.intermediate_steps), dtype=tf.float32)