An important medical problem in the noninvasive treatment of brain aneurysms demands special attention. The main reason is that, in many cases, it is necessary to have direct information about the size, shape and exact location of the aneurysm. To provide a medical specialist with such information, a virtual aneurysm system has been created. Problems with its realization and efficiency are discussed in this paper. A novel approach to the simulation and visualization of biomedical systems is suggested, so as to achieve higher efficiency. This approach is called "branching simulation".

The development of a grasp planner for multifingered robot hands is described. The planner is knowledge-based, selecting grasp postures by reasoning from symbolic information on target object geometry and the nature of the task. The ability of the planner to utilize task information is based on an attempt to mimic human grasping behavior. Several task attributes and a set of heuristics derived from observation of human motor skills are included in the system. The paper gives several examples of the reasoning of the system in selecting the appropriate grasp mode for spherical and cylindrical objects for different tasks.>

System identification is the term scientists and engineers use to refer to the process of building mathematical models of dynamical systems based on observed data. This paper approaches system identification as a pattern recognition problem. We use computers to simulate the system response for a variety of different mathematical models. For each distinct system model, simulated system responses tend to remain segregated in one or more amorphous regions of system response space despite (1) large variations in system parameters, (2) experimental errors, and (3) noise. The actual system response is classified with the model corresponding to the region of system response space where it is found. The classifier is an Artificial Neural Network (ANN) which implements a Generalized Vector Quantizer (GVQ). A small number of simple but powerful discriminant functions facilitate the correct classification of most of the responses in any given region. The required distribution of discriminants among the regions evolves automatically as they learn their respective functions.

An analytical performance modeling approach is presented for load sharing policies that schedule jobs based on a global system state. The main contributions of this work are: the derivation of analytical bounds to compute the average delay in large asymmetric systems under dynamic load sharing control, when job routing is based on the global system state; the experimental derivation of parameters which affect the implementation of load sharing policies at the user level; the demonstration that load sharing is effective even in the presence of high overheads; and the demonstration that in the asymmetric environment under study, carefully tuned algorithms for load sharing provide a significant improvement in performance over simpler algorithms.>

The design is discussed of the Locus family of networks rendering PC interface (PCI) service. Given are the expected user workload, the hardware costs, and the performance constraints. The workload consists of three classes of users: interactive, highly-interactive, and communication-intensive. The number of PCs is given and is equal to the number of users. It is shown that the problem can be reduced to a discrete capacity and flow assignment (CFA) problem. The backbone capacity assignment is inspired by the Lagrangian decomposition approach. It starts from a backbone capacity assignment which matches the initial flow and chooses a backbone server upgrade that gives the greatest response-time reduction per dollar. At this point, the flow deviation method is applied and iteration is continuous until the constraints are met. Several computational examples of network configuration synthesis are presented that emphasize the significance and generality of the results obtained.>

This paper presents an approach to synthesizing the control required by a dextrous multifingered robot hand to reach and safely grasp an arbitrary target object. The synthesis is based on analysis of the grasping task as performed by human beings. We divide the task into (1) a target approach phase (including target identification, hand structure and grasp mode selection, selection of approach trajectory, hand preshaping and orientation) and (2) a grasp execution phase (including shape and force adaptation). Each aspect of the required control is discussed. Particular attention is paid to the role of geometric modeling in target identification, to preshaping during the approach trajectory and to the requirements for autonomy in completion of the grasping task. The underlying philosophy is that of reflex control; each aspect of the grasping task is initiated and terminated using sensory data and rules of behavior derived from human expertise in such tasks. The contents and organization of the knowledge base which codifies this expertise are discussed. The fundamental assumptions upon which this control philosophy is based are identified throughout the paper. Preliminary results of a graphical simulation of hand preshaping are given.

A two-week short course that provides an overview of the UCLA Computer Science Department Master's Program is described.The short course program is comprised of 20 mini-courses and three plenary sessions.The audience for whom the program is intended is discussed and examples are cited indentifying that the continuing education goals of the program have been met.

Existing modeling methods in distributed parameter systems are directed to field problems for which a single equation gives an adequate representation. As a result, the techniques are severely taxed when coefficients of the governing partial differential equation (PDE) have relatively complex functional forms. The field is postulated to consist of contiguous regions that are easily modeled. During the modeling process itself, the regional boundaries are identified using a method based on pattern recognition (PR) techniques. The method was tested on computer simulations of a water quality resource system. Quantitative measures are developed and used to demonstrate the feasibility of applying the methodology to realistic distributed parameter systems (DPS's).

A network of passive electrical resistors is connected in a closed loop with a digital computer. This network functions as a parallel-operating subroutine to greatly accelerate the digital computer solution of certain classes of numerical problems. Computer systems employing the analog subroutine concept are "digital-computer-oriented" hybrid computers. Problems are programmed for conventional digital computation, and the digital computer flow chart is analyzed to determine if some particularly time-consuming blocks could be better handled by analog rather than digital techniques. A simple special-purpose analog unit is then constructed for this computational task.

The purpose of this paper is to provide a broad review of a class of unconventional computational methods for solving certain partial differential equations. The most widely used analog and digital techniques for solving initial-value problems involve the discretization of the space domain, with the aid of finite difference approximations. In the case of digital methods, there results a set of algebraic equations which must be solved for each step in the time domain. In analog techniques, finite difference approximations lead to a set of simultaneous ordinary differential equations, one for each grid point in this space domain. These equations can then be integrated using analog integrators. Although quite general in their applicability, finite differences methods often entail excessive and uneconomical computational efforts. This shortcoming, as well as the desire to explore new potential applications for hybrid computers, has led a number of investigators to experiment with alternative techniques, techniques which the author of the paper under review has chosen to classify as functional approximation methods.

When digital computers are employed to solve linear and nonlinear partial differential equations using implicit finite-difference approximations, the most time-consuming portion of the calculation is the solution of the large system of simultaneous difference equations which must be carried out for each time increment. Recognizing that an analog computer system, by virtue of its parallel mode of operation, generates solutions of simultaneous algebraic equations virtually instantaneously regardless of the number of equations involved, a combined analog and digital computer system has been developed. The system includes a small digital computer connected in a closed loop with a network of analog elements, consisting of an array of inexpensive node modules, one for each grid point in space. The purpose of this network is to solve the system of difference equations whenever this solution is required by the digital program. The analog unit, therefore, serves as a subroutine. The resulting system has a number of important advantages over pure digital computer techniques.