simple_pendulum.controllers.ilqr
iLQR Model Predictive Control
Submodules
simple_pendulum.controllers.ilqr.iLQR_MPC_controller
iLQR MPC Controller
- class simple_pendulum.controllers.ilqr.iLQR_MPC_controller.iLQRMPCController(mass=0.5, length=0.5, damping=0.15, coulomb_friction=0.0, gravity=9.81, inertia=0.125, torque_limit=2.0, dt=0.01, n=50, max_iter=1, break_cost_redu=1e-06, sCu=10.0, sCp=0.001, sCv=0.001, sCen=0.0, fCp=1000.0, fCv=10.0, fCen=300.0, dynamics='runge_kutta', n_x=3)
Bases:
AbstractControllerController which computes an ilqr solution at every timestep and uses the first control output.
- compute_initial_guess(N=None, verbose=True)
compute initial guess
- Parameters:
- Nint, default=None
number of timesteps to plan ahead if N==None, N defaults to the number of timesteps that is also used during the online optimization (n in the class __init__)
- verbosebool, default=True
whether to print when the initial guess calculation is finished
- get_control_output(meas_pos, meas_vel, meas_tau=0, meas_time=0)
The function to compute the control input for the pendulum actuator
- Parameters:
- meas_posfloat
the position of the pendulum [rad]
- meas_velfloat
the velocity of the pendulum [rad/s]
- meas_taufloat, default=0
the meastured torque of the pendulum [Nm] (not used)
- meas_timefloat, default=0
the collapsed time [s] (not used)
- Returns:
- des_posfloat
the desired position of the pendulum [rad] (not used, returns None)
- des_velfloat
the desired velocity of the pendulum [rad/s] (not used, returns None)
- des_taufloat
the torque supposed to be applied by the actuator [Nm]
- init(x0)
Initialize the controller. May not be necessary.
- Parameters:
- ``x0``: ``array like``
The start state of the pendulum
- load_initial_guess(filepath='Pendulum_data/trajectory.csv', verbose=True)
load initial guess trajectory from file
- Parameters:
- filepathstring, default=”Pendulum_data/trajectory.csv”
path to the csv file containing the initial guess for the trajectory
- verbosebool, default=True
whether to print from where the initial guess is loaded
- set_goal(x)
Set a goal for the controller. Initializes the cost functions.
- Parameters:
- xarray-like
goal state for the pendulum
- set_initial_guess(u_trj=None, x_trj=None)
set initial guess from array like object
- Parameters:
- u_trjarray-like, default=None
initial guess for control inputs u ignored if u_trj==None
- x_trjarray-like, default=None
initial guess for state space trajectory ignored if x_trj==None