|
|
Plně pravděpodobnostní návrh řízení pro gaussovské stochastické systémy
Belda, Květoslav ; Tesař, Ludvík
Control of stochastic systems is generally formulated as a minimization of expected value of a suitably chosen loss function of system inputs, outputs and desired behavior with respect to feedback control strategies. The standard strategies (e.g. Linear Quadratic Gaussian control) choose control actions that make the closed-loop behavior as close as possible to desired one using expected and desired output values. More general approach is to consider complex information on stochastic system behavior by complex probabilistic description. On this approach, fully probabilistic design is based. It uses probabilistic description for characterization of closed-loop of stochastic system and its desired behavior. This paper points out the basic principles of fully probabilistic design and its practical application to the control of Gaussian stochastic systems.
|
|
|
Prediktivní řízení s přibližně zadaným referenčním signálem
Belda, Květoslav
This paper deals with two specific modifications of predictive control, which cope with approximately given reference signal. The both modifications consist in different definition of requirements (reference signals) for a system behavior. The first modification takes into account only permitted ranges (limits) of required reference signal. The second modification considers only initial and final system state (position) and system outputs can occur within whole system workspace (domain). In such cases, the real desired reference signal is not strictly given, but it follows from control design. Described modifications can solve e.g. manipulation tasks or stabilization of an output signal in some permitted range. The explanation of the both modifications is documented by several examples.
|
|
|
Prediktivní řízení pro mechatronické laboratorní modely
Belda, Květoslav
The paper deals with the design of discrete adaptive model-based predictive control for simple mechatronic systems. Simple mechatronic systems are considered as Single-Input/Single-Output systems or possibly systems with low number of inputs and outputs. However, the methods of adaptation and model-based control are not generally limited to this condition. In the paper, a combination of on-line identification and generalized predictive control will be introduced. The identification is based on least squares. The predictive control arises from state-space formulation. This idea is applied to ARX models representing Input/Output formulation. The presented algorithms are derived in computationally suitable square-root form and their correctness is documented by tests on laboratory models.
|
|
|
Prediktivní řízení s časově proměnnými stavovými modely.
Belda, Květoslav ; Böhm, Josef
The papres is focused on Predictive control for time-variant state-space models both SISO and MIMO type, specially on model identification/composition, further on consideration of system nonlinaerities in equations of predictions and state-space estimation. The algorithms will by presented in coputationally-efective square-root forms and demonstrated by several simulative examples with simple linear SISO systems and with one nonlinear systems represented by MIMO model of planar redundant parallel robot.
|
guest ::
