Sung-Man Park

Experimental system identification in stochastic domain using cantilever beam

Unlikely to the conventional system identification method in deterministic domain, a stochastic approach has been attempted to figure out the system parameters of unknown system experiencing random disturbance whose stochastic information is not available. Interest is focused to identify an unknown stochastic dynamic system via stationary system output in stochastic sense. A concept to identify the system parameters in stochastic domain is proposed and implemented in terms of simulation and experiment as well. Previously developed stochastic transformation method via F-P-K (Fokker-Planck-Kolmogorov) equation is adopted to estimate the system output. Based on stationary system output only, it is attempted to identify the system parameters in inverse manner in stochastic domain successfully. Experimental study using cantilever beam is conducted to confirm quite noticeable performance of the proposed concept.

PID Tuning Based on RCGA Using Ziegler-Nichols Method

Real-coded genetic algorithm(RCGA) has better performances than conventional genetic algorithm about dealing with a large domain, the precision and the constrain problem. Also the RCGA has advantage of operation time because it doesn`t have to following about decoding operation. In this paper the ranges of PID gains are limited based on Ziegler-Nichols method to consider a long operation time problem that is the main problem of genetic algorithm. Result shows proposed method represents better performance without ignored about result of ZN tuning method and reduces the calculation time.

Stochastic system identification of unknown flexible cantilever beam under turbulent flow

Unlike the conventional system identification method in the deterministic domain, a stochastic approach has been attempted to figure out the system parameters of a dynamic system experiencing random disturbance whose stochastic information is not available. Interest is focused to identify an unknown stochastic dynamic system via stationary system output in a stochastic sense. A concept to identify the system parameters in the stochastic domain is proposed and implemented in terms of simulation and experiments. Previously developed stochastic transformation method via the Fokker-Planck-Kolmogorov (F-P-K) equation is adopted to estimate the system output. Based only on stationary system moment output, an attempt is made to identify the system parameters in inverse manner in the stochastic domain. Experimental studies using a cantilever beam under turbulent flow and a mechanical exciter are conducted to confirm the proposed concept to reveal quite noticeable performance.

Riding on moving escalator of humanoid robot using 3D-space recognition algorithm with single camera

Vision sensors are widely being used in robot system to recognize spatial condition accurately. There are two ways of implementation to recognize 3-D spatial environment. One way is using stereo camera, and the other way is using single camera by changing its position. However a new technique using single camera with geometric analysis has been developed to recognize 3-D information effectively. The algorithm using the perspective concept applied to the vision data obtained from single fixed camera is implemented for the simulation of humanoid robot system to ride on the moving escalator. An accurate recognition of 3-D spatial vision has been accomplished which enables the humanoid robot system to take a necessary action. Also experiment shows successful operation of robot to ride on moving escalator with newly developed algorithm. The newly proposed image processing algorithm using perspective geometric concept turns out to have such a big advantage as it requires much less computation time than conventional methods does. Accordingly it can take care of lots of 3-D image information in relatively very short time; hence it should be one of the powerful methods to make the robot operating in real time frame.

Model Reduction Using Stochastic Balance Technique

Recently, dynamic system has been enlarged and is normally exposed to various types of disturbance. Thus designing controller for these dynamic systems under random disturbance is not practically easy. As a result, the exact analysis for the system which is exposed to various irregular disturbance is quite important. In order to perform analysis, conventional BMR(balance model reduction) method is adopted and applied to moment equation in stochastic domain. Reliable reduced order system model has been obtained.

System Identification Using Stochastic Output Only

Most of the study on system identification has been carried out using input/output relation in physical domain. However identification concept of stochastic system has not been reported up to now. Interest is focused to identify an unknown dynamic system under random external disturbances which is not possible to measure. A concept to identify the system parameters in stochastic domain is proposed and implemented in terms of simulation. Attempt has been made to identify the system parameters in inverse manner in stochastic domain based on system output only. Simulation is conducted to reveal quite noticeable performance of the proposed concept.