config_gui.yml

# Parameters CartPole GUI starts with.
# This is useful if you need to many times restart the GUI to some particular setting,
# e.g., while testing new controller.

time_scales:  # Cannot be changed after starting the program
  dt_simulation: 0.002  # s, time step of CartPole simulation if run through GUI.
  controller_update_interval: 0.02  # s, sets how often the control input should be calculated.
  # It must be the multiple of dt_simulation.
  # The provided value is translated to the number of time simulation steps. 0 is treated the same as 1 with controller updated at every simulation time step.
  # # controller_update_interval_steps =  np.rint(controller_update_interval/dt_simulation)
  save_interval: 0.02  # s, How often to save the results of simulation.
  # We recommend to set it to the same value as controller_update_interval.
  # In such a case saving is done at the same time step as the controller is updated.
  # The same considerations as for controller_update_interval apply

gui_settings: # Can be changed in GUI
  controller_init_cpp: 'mpc'  # Load as default if loaded as part of physical-cartpole.
  controller_init_cps: 'neural-imitator'  # Load as default if loaded as cartpole simulator stand alone.

#  Possible choices for CartPole controller include:
#
#  'manual-stabilization' -> GUI slider sets directly motor power
#
#  'lqr' -> linear-quadratic regulator (LQR)
#
#  'mpc' -> model predictive control (MPC), our implementation
#
#  'do-mpc' -> mpc from do-mpc library with true equations
#  'do-mpc-discrete' -> same as above but based on discrete-time model obtained with a single step euler integration
#
#  'neural-imitator' -> neural network trained to mimic classical controller

  save_history_init: false  # Save experiment history as CSV (after experiment finished, online saving for Data Generator only).
  show_experiment_summary_init: false  # If true a window plotting the experiment history will be displayed after experiment ended.
  stop_at_90_init: false  # Block the pole if it reaches +/-90 deg (horizontal position).
  slider_on_click_init: true  # True: update slider only on click, False: update slider while hoovering over it.
  simulator_mode_init: 'Slider-Controlled Experiment'  # Effects Start/Stop button. Options are 'Slider-Controlled Experiment', 'Random Experiment', 'Replay'.
  # 'Slider-Controlled Experiment': Run experiment controlled by user (through slider)
  # 'Random Experiment': Run experiment with random target position
  # 'Replay': load and replay a recording; False: start new experiment TODO: probably something different when running with physical cartpole
  speedup_init: 1.0  # Multiplicative factor by which the simulation seen by the user differs from real time.
  # E.g. 2.0 means that you watch simulation double speed
  # WARNING: This is the target value, max speedup is limited by speed of performing CartPole simulation
  # True instantaneous speedup is displayed in CartPole GUI as "Speed-up(measured)"

random_trace_generation:  # Can be partly modified in GUI.
  # Variables for random trace generation - GUI only
  # Data Generator sets these parameters independently
  track_relative_complexity_init: 0.5  # Randomly placed target points/s, 0.5 is normal default.
  length_of_experiment_init: 1.0e3  # Seconds, length of the random length trace.
  interpolation_type_init: 'previous'  # Sets how to interpolate between turning points of random trace. Options: '0-derivative-smooth', 'linear', 'previous'.
  turning_points_period_init: 'regular'  # How turning points should be distributed. Options: 'regular', 'random'.
  # Regular means that they are equidistant from each other
  # Random means we pick randomly points at time axis at which we place turning points

  # Where the target position of the random experiment starts and end:
  start_random_target_position_at_init: 10.0
  end_random_target_position_at_init: 10.0

  # Alternatively you can provide a list of target positions.
  # e.g. turning_points_globals = [10.0, 0.0, 0.0]
  # If not None this variable has precedence -
  # track_relative_complexity, start/end_random_target_position_at_globals have no effect.
  turning_points_init: null  # Alternatively, you can provide a list of target positions.