AMPCSampler

AMPCSampler#

class AMPCSampler(mpc)[source]#

Bases: object

Class to sample data for the ApproxMPC.

Added in version >v4.6.0..

This class randomly samples the MPC to generate data. This data can further be used to train the ApproxMPC.

Configuring and setting up the Sampler involves the following steps:

  1. Configure the Sampler controller with SamplerSettings. The Sampler instance has the attribute settings which is an instance of SamplerSettings.

  2. To finalize the class configuration setup() may be called. This method sets up the MPC and Simulator, if needed.

Usage:

The Sampler can be used in a closed loop setting by calling the default_sampling() method. This method starts generating data based on the user settings.

Example:

# pushing to class
mpc.x0 = x0
mpc.set_initial_guess()

# approximate mpc
approx_mpc = ApproxMPC(mpc)
approx_mpc.settings.n_hidden_layers = 3
approx_mpc.settings.n_neurons = 50
approx_mpc.setup()

# sampler
sampler = Sampler(mpc)
sampler.settings.closed_loop_flag = True
sampler.settings.trajectory_length = 5
sampler.settings.n_samples = 100
sampler.setup()
sampler.default_sampling()
Parameters:

mpc (do_mpc.controller.MPC) – The MPC class which is sampled.

Methods#

approx_mpc_closed_loop_sampling#

approx_mpc_closed_loop_sampling(self)#

Closed loop sampling for the ApproxMPC.

This method generates closed loop samples for the ApproxMPC. The samples are saved in the data directory.

Returns:

None

Return type:

None

approx_mpc_open_loop_sampling#

approx_mpc_open_loop_sampling(self)#

Open loop sampling for the ApproxMPC.

This method generates open loop samples for the ApproxMPC. The samples are saved in the data directory.

Returns:

None

Return type:

None

approx_mpc_sampling_plan_box#

approx_mpc_sampling_plan_box(self)#

Generate sampling plan for the ApproxMPC.

This method generates a sampling plan for the ApproxMPC. The sampling plan is saved in the data directory.

Returns:

None

Return type:

None

default_sampling#

default_sampling(self)#

Default sampling method.

This method is used to sample data for the ApproxMPC. The method generates data based on the user settings.

Returns:

None

Return type:

None

setup#

setup(self)#

Setup the Sampler class.

This method sets up the MPC, and the Simulator if needed.

Note

This method should be called before calling any other method.

Returns:

None

Return type:

None

setup_simulator#

setup_simulator(self)#

Setup the Simulator.

This function is only used when the closed_loop_flag is set to True. It sets up the Simulator for the closed loop sampling.

Warning

Ensure that mpc.settings.n_robust == 0, when closed_loop_flag is set to True. Currently the Sampler does not support robust mpc.

Returns:

None

Return type:

None

Attributes#

settings#

AMPCSampler.settings#

Sampler settings. This attribute is an instance of SamplerSettings. It is used to configure the Sampler.

Returns:

The settings for the Sampler.

Return type:

SamplerSettings

simulator_settings#

AMPCSampler.simulator_settings#

Simulator settings.

This attribute is an instance of do_mpc.simulator.SimulatorSettings. It is used to configure the Simulator.

Returns:

The settings for the Simulator.

Return type:

do_mpc.simulator.SimulatorSettings

Contents