ApproximateMPCSettings

class ApproximateMPCSettings(n_hidden_layers=3, n_neurons=50, act_fn='tanh', output_act_fn='linear', device='auto', scaling=True, lbx=None, ubx=None, lbu=None, ubu=None)

Bases: object

Settings for do_mpc.approximateMPC.ApproxMPC.

This class contains the mandatory settings for Approximate MPC int do_mpc.approximateMPC.ApproxMPC. This class creates an instance of type ApproximateMPCSettings and adds it to its class attributes.

Parameters:

• n_hidden_layers (int)

• n_neurons (int)

• act_fn (str)

• output_act_fn (str)

• device (str)

• scaling (bool)

• lbx (list)

• ubx (list)

• lbu (list)

• ubu (list)

Methods

Attributes

act_fn

ApproximateMPCSettings.act_fn: str = 'tanh'

Activation function used after each layer

device

ApproximateMPCSettings.device: str = 'auto'

Type of device used. Can be auto, cuda or cpu

lbu

ApproximateMPCSettings.lbu: list = None

Lower bound for the control variables

lbx

ApproximateMPCSettings.lbx: list = None

Lower bound for the state variables

n_hidden_layers

ApproximateMPCSettings.n_hidden_layers: int = 3

Number of hidden layers

n_neurons

ApproximateMPCSettings.n_neurons: int = 50

Number of neurons per hidden layer

output_act_fn

ApproximateMPCSettings.output_act_fn: str = 'linear'

Output type

scaling

ApproximateMPCSettings.scaling: bool = True

Decides whether the state and control variables are scaled or not

ubu

ApproximateMPCSettings.ubu: list = None

Upper bound for the control variables

ubx

ApproximateMPCSettings.ubx: list = None

Upper bound for the state variables