Andrzej Turnau

Real-time controller design based on NI Compact-RIO

The paper is focused on NI Compact-RIO configured as a controller for the active magnetic levitation used here as a benchmark for time-critical systems. Three real-time configurations: soft, soft with IRQ and hard FPGA are considered. The quality of the real-time control has been tested for each configuration.

FPGA as a part of MS WINDOWS control environment

The attention is focused on the Windows operating system (OS) used as a control and measurementenvironment. Windows OS due to extensions becomes a real-time OS (RTOS).Benefits and drawbacks of typical software extensions are compared. As far as hardwaresolutions are concerned the field programmable gate arrays FPGA technology is proposed toensure fast time-critical operations. FPGA-based parallel execution and hardware implementationof the data processing algorithms significantly outperform the classical microprocessoroperating modes. Suitability of the RTOS for a particular application and FPGA hardwaremaintenance is studied.

Model Identification Dedicated to the Time-Optimal Control

A model identification procedure is applied to the well known benchmark problem of the pendulum hinged to a cart. There is a dynamical model of the entire system. The PWM control signal and DC motor impact introduced electrically by EMF are included. A concatenation of trajectories collected during several control experiments is used to fit the parameters of the pendulum-cart mathematical model. The identification of model parameters is dedicated to the control goal. Several collected points of trajectories are neglected. The model matching corresponds to intervals.

Time-Optimal Control Supported by PD in Real-Time* *

Abstract The Magnetic Levitation System (MLS) is ideal for modelling because it is frictionless. With a model well-fitted to the system one can generate adjoint equations and solve both the state and adjoint differential sets numerically to obtain the time-optimal sequences corresponding to the desired initial and final positions of the levitating sphere. This action is repeated for a broad set of the sphere position differences. The resulting bang-bang controls are associated with the sphere levels and stored in the computer memory. Each sphere level is equipped with an appropriately tuned PD controller to maintain levitation. This means that the damping and elasticity coefficients are sustained at the values close to each other despite changes of the distance between the sphere and electromagnet. In this way the time-optimal, open-loop control applied for the position change aided by the PD controller at steady states becomes a variable control structure and appears to be the correct and rapid strategy to experiment with MLS in the real-time.

Neural adapted controller learned on-line in real-time

Abstract An adapted controller consisting of a PD structure aided by a neural network is considered. It is dedicated to an unstable magnetic levitation system (MLS) to maintain a levitated sphere in the real-time. The main motivation is to create a self-tuning intelligent device that can easily change the controller parameter estimates based on adaptive update rules. How to formulate these rules in order to not disturb or ruin the system stability is the most serious question. This issue is discussed and verified by real-time experiments carried out at our apparatus that enables to vary its parameters. The neural network learned on-line guarantees the robustness of the control despite the system parameter are changed and disturbances are introduced. One can expect a better stabilizing performance as well.

Semi-Active Suspension Laboratory System

The paper describes the semi-active suspension laboratory system (SAS), built to demonstrate and test a number of control algorithms. The heart of the system is the automotive engineering damper with the restoring force controlled by magnetic field. We use in our apparatus the Lord 1097 magnetorheological (MR) damper manufactured by the American company. MR devices benefit from the ability of MR fluids to rapidly change rheological properties upon exposure to a magnetic field. The main advantage of the SAS is its portability as a demonstrative experimental rig to test control damping algorithms. The damped less oscillations are compared to the On-off damped oscillations of the apparatus. The MR damper provides an effective solution to SAS control in a variety of applications.

A ROBUST REPETITIVE CONTROL SCHEME WITH RELAXED MINIMUM TIME CRITERION

Abstract A repetitive control scheme is proposed for constrained nonlinear optimal control problems. The lower level algorithm adjusts switching times for bang control arcs and parameters of interval polynomial approximations for interior control arcs. It is based on a linearization of optimal controller and performs reduced optimization with changes of control structure. The upper level finds the optimal control and recalculates the linearization each time the deviation from the optimal solution becomes too large. The linearized controller is analytically derived. The upper level uses the MSE method to determine the reference optimal control structure. Simulation and experimental tests show that the proposed approach yields an optimizing nonlinear controller, able both to ensure close to minimum-time point-to-point transition as well as to stabilize the state.

A new current based slip controller for ABS

A laboratory Anti-Lock Brake System (ABS) is examined. The architecture of the ABS system is shown. The real-time experiments related to a control action to stabilize the slip at a certain level are taken into consideration. A new slip control algorithm differs from the previously used approaches. It is based on the measured current of the braking DC motor. The experimental results collected in the real-time for an old (relay) and new (current based) slip control are compared.

Desensitization of the time-optimal trajectories

The time-optimal trajectories are very sensitive to disturbances. The use of optimal strategy in real-time inevitably results in a departure from the optimal path. A return to the trajectory is no longer possible. When the time-optimal is replaced to the fix horizon problem then the real system moves from the optimal to nearing trajectories. Moreover, a further desensitization is possible by a proper shifting. The nearing desensitized trajectories are robust to disturbances at the cost of a slight increase in the control time.