SIGNALS AND SYSTEMS RESEARCH CENTER (SSRC)

Location: 386 MWAH, UMD

Taek M. Kwon, PhD: 218-726-8211, tkwon@ub.d.umn.edu
Jiann-Shiou Yang, Ph.D: 218-726-6290, jyang@ub.d.umn.edu

I. INTRODUCTION

The Signals and Systems Research Center (SSRC) is part of the Department of Computer Engineering and aims to strengthn research activity and contribute to the recognition of the department. The center is open to any member of the department committed to research and scientific contribution. As members of the center, we are committed to perform hard-core research, develop innovative applications, pursue active outreach, and attract external research funding. We are also committed to train students in engineering and expose them to the state-of-art technology through active involvement in research.

II. AREAS OF EXPERTISE

III. CURRENT RESEARCH TOPICS

IV. CURRENT RESEARCH PROJECTS

This section summarizes the on-going projects of SSRC at the present time. SSRC continuously strives for developing innovative research projects and welcomes suggestions from any party at any time.

A. Development of Self-Organizing and Trainable Fuzzy Controller

Taek Mu Kwon and Michael Zervakis
Supported by Grant-in-Aid, Proposal # 15463

As the applications of fuzzy-controllers become more complicated, the attributes of self-organization and trainability become increasingly important. Indeed, the specification of fuzzy rules and membership functions for systems with a large number of state variables is extremely difficult. This project introduces a new class of self-organizing and trainable fuzzy-controllers that can be designed without specific information regarding both the membership functions and the fuzzy rules. The proposed controller derives the fuzzy rules from clusters formed in the input space, through a self-organizing process. The clustering is performed through a simple method that can adaptively allocate new clusters as more data are available to the controller. Then, the membership values of crisp inputs are determined by K-nearest-neighbor (KNN) distance measures applied to the centers of the input clusters. Finally, a KNN defuzzification processes directly estimates the crisp output of input data. An adaptation procedure for the center vector of each cluster and the corresponding output value is developed. The overall design is analyzed in terms of the existence and the uniqueness of the solution of the proposed model. This new model is under experiments with variety of applications which include: the truck upper-backer control, the S\&P 500 index prediction, and the Mackey-Glass time-series prediction.

B. Image Restoration in the Wavelet Domain

Michael Zervakis, Taek Mu Kwon, and Jiann-Shou Yang
Supported by Grant-in-Aid, Proposal # 15463

This project proposes an image restoration scheme in the wavelet domain that directly associates the multiresolution and the multichannel approaches. We present a formulation of the multiresolution image that transforms block-circulant structures into a partially-block-circulant structure. We prove that the stationarity assumption in the image domain leads to the suppression of cross-band correlation in the multiresolution domain. Moreover, the space invariance assumption leads to the loss of cross-band interference and interaction. In addition to the rigorous explanation of these effects, our formulation reveals new correlation schemes for the multiresolution signal in the wavelet domain. In essence, the proposed implementation relaxes the stationarity and space-invariance assumptions in the image domain and introduces new operator structures for the implementation of single-channel algorithms that take advantage of the correlation structure in the wavelet domain. Several image restoration examples on the Wiener-filtering approach show significant improvement achieved by the proposed method over the conventional DFT implementation.

C. Robust Estimation Approaches in Image Processing

Michael Zervakis, and Taek Mu Kwon
Supported by Grant-in-Aid, Proposal # 14835 and the Faculty Summer Research Award, Proposal # 15556

This project considers the concept of robust estimation in regularized image restoration. Robust functionals are employed for the representation of both the noise and the signal statistics. Such functionals allow the efficient suppression of a wide variety of noise processes and permit the reconstruction of sharper edges than their quadratic counterparts. A new class of robust entropic functionals is introduced, which operates only on the high-frequency content of the signal and reflects sharp deviations in the signal distribution. This class of functionals can also incorporate prior structural information regarding the original image, in a way similar to the maximum information principle. Iterative algorithms for the solution of the robust restoration problem are developed. The convergence properties of these algorithms are analyzed for continuously and non-continuously differentiable functionals.

D. Multichannel Image Processing

Michael Zervakis Supported by Grant-in-Aid, Proposal No. 14835 and the Faculty Summer Research Award, Proposal No. 14456

In this project we develop an efficient algorithm for the problem of multichannel image restoration. Existing multichannel techniques do not provide sufficient flexibility for the simultaneous suppression of the noise process and the preservation of sharp detailed structure in the estimate. The approach introduced overcomes this inefficiency by introducing the prototype Wiener structure in the smoothing process of the estimate. The corresponding algorithm is obtained from the optimization of the {\it constrained mean-square error} (CMSE) criterion, which is interpreted as a structured regularized criterion. The CMSE estimate has always a meaningful structure and lies between the minimum mean-square-error estimate and the pseudo-inverse solution. In addition, the CMSE approach enables the suppression of streak artifacts, which are often experienced due to the amplification of the noise process.

E. Adaptive Inverse Dynamics Control for a Five-Link Biped

Jiann-Shiou Yang
Supported by Grant-in-Aid, Proposal No. 15324

In this project we study a five-degree-of-freedom biped locomotion system. We generate the joint trajectories for the biped walking gait and compare them with the data collected by a TV/computer measurement of human walking. We found that the trajectories generated are close to those measured. Based on the the biped model and the computed joint trajectories, we then apply an adaptive inverse dynamics control scheme to control the biped motion. The control law has the structure of the inverse dynamics servo but uses estimates of the dynamics parameters in the computation of torques which propel the biped. The adaptation law uses the tracking error to compute the parameter estimates for the control law. To improve the convergence of the estimated parameters, we modify the timing of applying the adaptation by incorporating a dead zone operation. Our simulation results show that the adaptive control technique can be effectively used for the biped locomotion system.

F. Mixed H^2-Norm Sensitivity Minimization for Control System Design

Michael Zervakis and Jiann-Shiou Yang

In this project we deal with the problem of designing a feedback controller for linear time-invariant discrete-time systems. The approach introduced minimizes the H^2 norm of a mixed sensitivity criterion. With the standard Youla parametrization, the problem is initially converted into a problem of trading-off between two model matching problems in the $H^2$ space. Operating in the DFT (Discrete Fourier Transform) domain, we construct a minimization problem in the $l^2$-space whose dimensionality depends on the number of the inputs and outputs of the plant to be controlled and by the size of the DFT. The DFT vector-optimization problem can be efficiently solved through matrix algebraic techniques. The newly constructed problem can be made arbitrarily close to the original problem if the size of the DFT becomes large. By solving the l^2-norm minimization problem, the parameter matrix associated with the stabilizing controller can be found indirectly. An example is given to demonstrate the effectiveness of this approach.

G. Control of a Platoon of Vehicles

Jiann-Shiou Yang

We study the control of the successive vehicle spacings of a platoon of vehicles traveling up hill with an incline angle. A linear model to represent the vehicle dynamics of each vehicle within the platoon is used. The analysis on both the identical and non-identical vehicle cases was examined. Under the steady velocity changes of the lead vehicle, we found that the deviations of the vehicles from their pre-assigned positions for both cases are reasonably small and are also able to return to their steady state. %when the incline angle is not too large. The final steady state deviations of the vehicles in the platoon increase with the increase in the steepness of the hill.

V. TYPICAL PUBLICATIONS

Back to Kwon's home Page.

~