rcognita.systems.System

class rcognita.systems.System(sys_type, dim_state, dim_input, dim_output, dim_disturb, pars=[], ctrl_bnds=[], is_dyn_ctrl=0, is_disturb=0, pars_disturb=[])

Interface class of dynamical systems a.k.a. environments. Concrete systems should be built upon this class. To design a concrete system: inherit this class, override:

_state_dyn() :
right-hand side of system description (required)
_disturb_dyn() :
right-hand side of disturbance model (if necessary)
_ctrl_dyn() :
right-hand side of controller dynamical model (if necessary)
out() :
system out (if not overridden, output is identical to state)
sys_type: string

Type of system by description:

diff_eqn : differential equation \(\mathcal D state = f(state, action, disturb)\)
discr_fnc : difference equation \(state^+ = f(state, action, disturb)\)
discr_prob : by probability distribution \(X^+ \sim P_X(state^+| state, action, disturb)\)

where:

\(state\) : state
\(action\) : input
\(disturb\) : disturbance

The time variable t is commonly used by ODE solvers, and you shouldn’t have it explicitly referenced in the definition, unless your system is non-autonomous. For the latter case, however, you already have the input and disturbance at your disposal.

Parameters of the system are contained in pars attribute.

dim_state, dim_input, dim_output, dim_disturb: integer

System dimensions

pars: list

List of fixed parameters of the system

ctrl_bnds: array of shape [dim_input, 2]

Box control constraints. First element in each row is the lower bound, the second - the upper bound. If empty, control is unconstrained (default)

is_dyn_ctrl: 0 or 1

If 1, the controller (a.k.a. agent) is considered as a part of the full state vector

is_disturb: 0 or 1

If 0, no disturbance is fed into the system

pars_disturb: list

Parameters of the disturbance model

Each concrete system must realize System and define name attribute.

__init__(sys_type, dim_state, dim_input, dim_output, dim_disturb, pars=[], ctrl_bnds=[], is_dyn_ctrl=0, is_disturb=0, pars_disturb=[])
Parameters

  • sys_type (: string) –

    Type of system by description:

    diff_eqn : differential equation \(\mathcal D state = f(state, action, disturb)\)
    discr_fnc : difference equation \(state^+ = f(state, action, disturb)\)
    discr_prob : by probability distribution \(X^+ \sim P_X(state^+| state, action, disturb)\)

  • where –

    \(state\) : state
    \(action\) : input
    \(disturb\) : disturbance

  • time variable t is commonly used by ODE solvers (The) –

  • you shouldn't have it explicitly referenced in the definition (and) –

  • your system is non-autonomous. (unless) –

  • the latter case (For) –

  • however –

  • already have the input and disturbance at your disposal. (you) –

  • of the system are contained in pars attribute. (Parameters) –

  • dim_state (: integer) – System dimensions

  • dim_input (: integer) – System dimensions

  • dim_output (: integer) – System dimensions

  • dim_disturb (: integer) – System dimensions

  • pars (: list) – List of fixed parameters of the system

  • ctrl_bnds (: array of shape [dim_input, 2]) – Box control constraints. First element in each row is the lower bound, the second - the upper bound. If empty, control is unconstrained (default)

  • is_dyn_ctrl (: 0 or 1) – If 1, the controller (a.k.a. agent) is considered as a part of the full state vector

  • is_disturb (: 0 or 1) – If 0, no disturbance is fed into the system

  • pars_disturb (: list) – Parameters of the disturbance model

Methods

__init__(sys_type, dim_state, dim_input, …)

param sys_type

Type of system by description:

closed_loop_rhs(t, state_full)

Right-hand side of the closed-loop system description.

out(state[, action])

System output.

receive_action(action)

Receive exogeneous control action to be fed into the system.