Institute for Systems Research Technical Reports

Permanent URI for this collectionhttp://hdl.handle.net/1903/4376

This archive contains a collection of reports generated by the faculty and students of the Institute for Systems Research (ISR), a permanent, interdisciplinary research unit in the A. James Clark School of Engineering at the University of Maryland. ISR-based projects are conducted through partnerships with industry and government, bringing together faculty and students from multiple academic departments and colleges across the university.

Browse

Start date: 1980End date: 1989

Search Results

Now showing 1 - 10 of 579
  • Thumbnail Image
    Item
    Determination of the Static Friction Present in Each Finger of a Three Fingered Modular Dextrous Hand
    (1989) Uber, J.A.; Loncaric, J.; ISR
    The purpose of this experiment was to determine the static friction (stiction) present in each finger of a three fingered modular hand and determine the torque required to break away from a stuck position. The static friction is a function of both position and direction of movement of the mtor driving the finger. Upon completion of enough trial runs from each position in each direction the results obtained will be averaged together to produce a mapping of the static friction of each finger of the gripper as a function of position and direction. This stiction data base will then be incorporated as a torque look-up table into the open loop control systems architecture of the hand. This compensation has been used in other robotics control experiments to produce a significantly more accurate open loop movement, greatly easing the operational burden of the eventual closed loop control.
  • Thumbnail Image
    Item
    Real-Time Control for a Zero Gravity Robotic End Effector
    (1989) Salter, C.A.; Baras, J.; ISR
    There is no doubt that the task of gripping and handling objects in space is an important one. The ability to easily manipulate objects in a zero gravity environment will play a key role in future space activities. It is the aim of this research to develop control laws for the zero gravity robotic end effector designed by engineers at NASA Goddard. A hybrid force/position controller will be used. Sensory data available to the controller are obtained from an array of strain gauges and a linear potentiometer. Applying well known optimal control theoretic principles, the control which minimizes the transition time between positions is obtained. A robust force control scheme is developed which allows the desired holding force to be achieved smoothly without oscillation. In addition, an algorithm is found to determine contact force and contact location.
  • Thumbnail Image
    Item
    Adaptive Pattern Classification
    (1989) Teolis, A.; Shamma, S.; ISR
    Up until the recent past, the power of multi layer feed forward artificial neural networks has been untapped mainly due to the lack of algorithms to train them. With the emergence of the backpropagation algorithm; however, this deficiency has been removed. Despite this innovation, the backpropagation method is still not without its drawbacks. Among these the most prominent are the facts that i) the learning is conducted in a supervised manner and ii) that learning and operation must occur in two distinct phases. Because of these properties, the backpropagation algorithm falls short of solving a 'true' pattern classification problem. This is not to say that a network could not be trained via backpropagation to mimic a previously solved pattern classification scheme; but that the backpropagation method is not capable of autonomously generating classification schemes. A more realistic (and certainly more useful) learning scenario is that patterns would be presented without supervision to the system continuously; consequently, the system will begin to group like patterns into similar classes and continue to do so indefinitely; i.e. continuous learning. It is exactly this type of learning that is discussed here.
  • Thumbnail Image
    Item
    Design, Implementation and Testing of an 8x8 DCT Chip
    (1989) Sachidanandan, S.; JaJa, J.; ISR
    An implementation of a fully pipelined bit serial architecture to compute the 2-D Discrete Cosine Tranform of an 8x8 element matrix is presented. The algorithm used requires the minimal number of multipliers to perform the computation. The basic hardware components required by our implementation are the one bit serial adder, one bit serial subtractor, one bit pipeline multiplier and the dynamic shift register. We use two's complement arithmetic. An internal precision of 18 bits is maintained throughout the chip. We have used the two phase non-overlapping clocking scheme. The basic architecture consists of an 1-D ROW DCT followed by a TRANSPOSE operation and another 1-D COLUMN DCT. All the components were custom designed and simulated at various levels. The chips were laid out using the layout editor MAGIC and were fabricated by MOSIS. The chips were tested using the IMS Logic Master. The results of the simulation and testing are also presented here. The throughput is very high and inputs can be processed successively with no delay and just two controls. The chip is designed to operate at clock speeds of 10Mhz or more.
  • Thumbnail Image
    Item
    Learning Binaural Processing in Biological Networks
    (1989) Gopalaswamy, Preetham; Shamma, S.; ISR
    It has always intrigued man as to how the human body performs so many complicated functions with the speed and accuracy that it does. One such task is that of sound localization in space, the ability to determine the location of a sound source with considerable accuracy. A biologically realistic neural network is proposed for the binaural processing of interaural time and intensity cues that closely resembles computational schemes suggested for sterepsis (depth perception) in vision. The important feature of this network is that it does not use any neural delay lines to generate such attributes of binaural hearing such as lateralization of all audible frequencies and the detection of enhancement of signals in a noisy environment. Temporal shifts between the signals at the ears, arising from sound sources at different locations on the azimuth cause spatial disparities in the corresponding travelling waves set up on the basilar membranes in the two ears. The two dimensional network proposed uses these spatial differences between instantaneous outputs at the two ears to measure interaural differences. The network operation approximately computes the cross-correlation between the two cochlear outputs by combining the ipsilateral input at a given characteristic frequency (CF) with contralateral inputs from locally off-CF locations. Some of the results obtained from this network are presented. Having proposed a network, the next question is whether such a connection is genetically present in the body or whether it is formed over a long period of time by a gradual process of learning. Assuming that the latter solution is more plausible, two learning rules are suggested according to which the network could alter its initial random connectivities. The first learning rule is a supervised technique in which a teaching signal prespecifies the ideal response expected from the network to each input pattern presented. The error between the actual output and the desired response helps to guide the learning process in the desired direction. When the minimum of the error surface is reached, the network is said to have learned and the weights do not change any more. The teaching signal required for the supervised algorithm could be derived from the visual system. However, no physiological evidence exists that links the auditory and visual maps at the level of the olivary complex which is where early binaural processing occurs. To overcome this problem, an unsupervised learning rule is proposed which requires only the cochlear outputs from the two ears. The rule is a competitive learning strategy wherein only one neuron updates its connectivities for a particular input pattern. The neuron chosen to alter its weights is the one which responds maximally to the input. The inherent delays that exist in the neural system are used as guides to form the organized spatial map responses.
  • Thumbnail Image
    Item
    Mobile Robot Navigation Using Potential Functions
    (1989) Shahidi, R.; Shayman, M.; Krishnaprasad, P.S.; ISR
    This thesis presents a method to construct a smooth obstacle free path for a mobile robot on which to navigate. The first step is to assign potential functions to each obstacle and the goal. Then, the gradient system constructed by the gradient of the sum of the above functions generates the desired path. The construction is analytically proven to produce obstacle free paths to the goal for an environment whose obstacles can be approximated by disks. The procedure does not require complete information on the position of the obstacles beforehand, as long as they can be detected and approximated by disks. The algorithm presented shows a computationally simple way to construct paths and a systematic method to encode the geometric data about the environment into a smooth vector field, which is used for mobile robot navigation.
  • Thumbnail Image
    Item
    Neural Networks for Low Level Processing of Tactile Sensory Data
    (1989) Pati, Y.C.; Krishnaprasad, P.S.; ISR
    As the field of robotics continues to strive forward, the need for artificial tactile sensing becomes increasingly evident. Real-time, local processing of tactile sensory data also becomes a crucial issue in most applications of tactile sensing. In this thesis it is shown that analog neural networks provide an elegant solution to some of the problems of low level tactile data processing. We consider the particular problem of 'deblurring' strain data from an array of tactile sensors. It is shown that the inverse problem of deblurring strain measurements to recover the surface stress over a region of contact is ill-posed in the sense defined by Hadamard. This problem is further complicated by the corruption of sensor data by noise. We show that the techniques of 'regularization' may be used to introduce prior knowledge of the solution space into the solutions in order to transform the problem to one which is well-posed and less sensitive to noise. The particular regularizer chosen for the recovery of normal stress distributions is of the functional form of Shannon entropy. Formulation of the inverse problem so as to regularize the solutions result in a variational principles which must be solved in order to recover the surface stress. An analog neural network which provides the desired solutions to the variational principle as a course of natural time evolution of the circuit dynamics is proposed as a solution to the requirements for fast, local processing in tactile sensing. We discuss performance of the network in the presence of noise based upon computer simulations. We also demonstrate, by means of a breadboard prototype of the network, the speed of computation achievable by such a network. An integrated circuit implementation of the proposed network has been completed and the requirements of such implementations is discussed.
  • Thumbnail Image
    Item
    Development of a CAM System for the Rapid Prototyping Environment
    (1989) Hudson, Daniel C.; Kirk, J.A.; ISR
    The goal of this research was to develop a Computer-Aided Manufacturing system for rotational parts for application to small-scale manufacturing. The developed software, DPCAM, was to provide an intermediate step between the total automation of a Rapid Prototyping Center and current commerically available systems. This was achieved by applying technologies already developed for the total automation problem to the intermediate solution. Modules included in the DPCAM package include: -Design Retrieval -Part Evaluation -Estimation -N/C Coding Generation - Equipment Management The information from the modules is presented to the user through a series of "user friendly" graphics windows. While each of these functions is a "state of the art" system, incorporating techniques such as Group Technology and Manufacturability Analysis, the major contribution is their integration into a single complete CAM package. The finished system is implemented on a "PC" style computer and tested at the University of Maryland's Flexible Manufacturing Laboratory.
  • Thumbnail Image
    Item
    Combined Source-Channel Coding for Bandlimited Waveform Channels
    (1989) Vaishampayan, V.A.; Farvardin, N.; ISR
    We consider the problem of transmitting data from a continuous- ampltude, discrete-time source over a bandlimited waveform channel using a block-structured digital communication system. Our objective is to design the source encoder-decoder pair, the channel encoder-decoder pair and the modulator-demodulator pair so as to minimize a squared-error distortion measure betweent he source sequence and its replica in the receiver, subject to constrainsts on the transmitted signal power and bandwidth. We formulate the problem i a gneral sense and derive necessary conditions for optimality for the design variables, namely, the encorder map, the decoder map and the modulation signal set. We then consider two systems for which the necessary condition for optimality hold. The receiver of the first system consists of an unquantized soft-decision demodulator followed by a linear estimator-based decoder. We solve the necessary conditions for optimality using an iterative soution technique. We then study the perofrmance of this class of systems as the encoding rate increases, for a fixed bandwidth. Performance comparisons are made against a reference system as well as bounds from information theory. Significant improvements over the reference system are demonstrated and the performance is shown to coincide with an information-theoretic bound in two cases. The receiver of the second system consists of a quantized demodulator followed by an optimum decoder. For a fixed signal set, the optimal encoder and decoder conditions are solved using an iterative solution technique. Performance comparisons are made against the linear estimator-based stystem, the reference ystem and bounds from information theory. We then study the quantized demodulator- based system as the encoding rate becomes large, for a fixed bandwidth. We demonstrate that this system converges to a lienar analog modulation system in some cases to a nonlinear analog modulation system in others. Finally, we study the performance of the two systems to channel mismatch. We demonstrate that both of these systems are more robust to channel mismatch than the reference system.
  • Thumbnail Image
    Item
    Studies in Robust Stability
    (1989) Saydy, L.; Abed, E.; Tits, A.; ISR
    In this thesis, questions in the analysis and synthesis of stability robustness properties for linear and nonlinear control systems are considered. The first part of this work is devoted to linear systems., where the emphasis is on obtaining necessary and sufficient conditions for stability of parametrized families of systems. This class of robustness problems has recently received significant attention in the literature [1]. In the second part of the thesis, questions of stabilization of nonlinear systems by feedback are considered. Part I of this work addresses the generalized stability, i.e. stability with respect to a given domain in the complex plane, of parametrized families of linear time-invariant systems. The main contribution is the introduction and application of the new concepts of "guarding map" and "semiguarding map" for a given domain. Basically, these concepts allow one to replace the original parametrized system stability problem with a finite number of stability tests. Moreover, the tool is very powerful in that it allows the treatment of a large class of domains in the complex plane. The parametrized stability problem is completely solved for the case stability of a one- parameter family with respect to guarded and semiguarded domains. The primary interest in semiguarded domains arises in a process of reduction of a given multiparameter problem to one involving fewer parameters. For the two-parameter case, we consider stability of families of matrices relative to domains with a polynomial guarding map. The first step replaces the two- parameter problem by a one-parameter stability problem relative to a new domain. The second step employs a polynomial semiguarding map for the new domain to obtain necessary and sufficient conditions for stability of the new problem. The case of three or more parameters, which involves technical questions not encountered in the one- or two-parameter case, is also considered. In Part II, a class of nonlinear control systems for which the linear part satisfies special stabilizability conditions is considered. These conditions naturally give rise to certain nonstandard algebraic issues in linear systems. Sufficient conditions for the existence of a linear feedback control which stabilizes a given nonlinear control system within a prescribed ball of given radius (possibly infinite) are given. The feedback control is found to be robust in a certain sense against a class of modeling errors. A complete design methodology is obtained for planar systems and extended to a class of higher dimensional singularly perturbed nonlinear control systems. For these systems, nonlinear feedback laws achieving stabilization within prescribed cylindrical regions are presented.