ControlledPlant{ST, CT, XT, DT, PT, CPRT, CPST} <: AbstractControlProblem

Struct representing a closed-loop controlled system.

Fields

• ivp – initial-value problem
• controller – controller
• vars – dictionary storing state variables, input variables and control variables
• period – control period
• postprocessing – postprocessing of the controller output
• preprocessing – preprocessing of the controller input

Parameters

• ST: type of system
• CT: type of controller
• XT: type of initial condition
• DT: type of variables
• PT: type of period
• CPRT: type of control preprocessing
• CPST: type of control postprocessing

Notes

While typically the controller is a neural network, this struct does not prescribe the type.

simulate(cp::AbstractControlProblem, args...; kwargs...)

Simulate a controlled system for a family of random trajectories.

Input

• cp – controlled problem
• trajectories – (optional, default: 10) number of simulated trajectories

Output

An object of type EnsembleSimulationSolution.

Notes

This function uses the ensemble simulations feature from OrdinaryDiffEq.jl.

solve(prob::AbstractControlProblem, args...; kwargs...)

Solve the control problem defined by prob.

Input

• prob – controlled problem

Additional options are passed as arguments or keyword arguments; see the notes below for details. See the online documentation for examples.

Output

The solution of a reachability problem controlled by a periodic controller. The control signals are stored in the ext field with each flowpipe.

Notes

Mandatory arguments

• Use the tspan keyword argument to specify the time span (start time and time

horizon); it can be a tuple, an interval, or a vector with two components. Alternatively, use the T keyword argument to specify only the time horizon, in which case the start time is assumed to be zero.

• Use the algorithm_plant keyword argument to specify the algorithm for the

plant.

• Use the algorithm_controller keyword argument to specify the algorithm for

the controller.

Optional arguments

• Use the splitter and input_splitter keyword arguments to specify a

splitter.

Default: NoSplitter()

• Use the reconstruction_method keyword arguments to specify a reconstruction

method for the interface between plant and controller.

Default: TaylorModelReconstructor()