SIMULATION: Transactions of The Society for Modeling and Simulation International

Padgett, Mary Lou; Padgett, W. David

doi: 10.1177/003754979506500101pmid: N/A

Many real-world applications of computationalintelligence (CI) start with a good trainingsimulator. Enhancing the simulator to use many human senses and to facilitate collection andanalysis of data may produce an intelligent virtual reality (VR) system. Careful design of aVR system may encourage the use of "hooks" to allow collectionof data,conversion to meaningfulinformatian, and incorporation of intelligence into later versions and into the final, real-world product. This paper will illustrate these concepts by discussing the challenges of autonomous flight, from simulation to implementation. Fuzzy risk analysis is suggested as a performance evaluation tool for design of CI systems, here defined as systems incorporating artificial neural networks (NN), fuzzy systems, (FZ) evolutionary systems (EC) and/or virtual reality (VR). Intelligent virtual simulation is a tool of the future which promises to move from an intriguing pastime for gamers to a rigorously applied scientific discipline offering economic advantages and service to those working in applications ranging from training simulators to accessibility aids for the handicapped. To participate in interest groups or standards activities, or to obtain additional information about simulation-oriented resources for CI development, contact [email protected].

doi: 10.1177/003754979506500102pmid: N/A

Duong, Tuan; Kemeny, Sabrina; Daud, Taher; Thakoor, Anil; Saunders, Chris; Carson, John

doi: 10.1177/003754979506500103pmid: N/A

A particularly challenging neural network application requiring high-speed and intensive image processing capability is target acquisition and discrimination. It requires spatio-temporal recognition of point and resolved targets at high speeds. A reconfigur able neural architecture may discriminate targets from clutter or classify targets once resolved. By mating a 64 x 64 pixel array infrared (IR) image sensor to a 3-D stack (cube) of 64 neural-net ICs along respective edges, every pixel would directly input to a neural network, thereby processing the infor-mation with full parallelism. Being mated to the infrared sensor array, the cube would operate at 90°K temperature with <250 nanosecond signal processing speed and a low power consumption of only -2 watts. For low power and compactness in hardware, the emphasis has been on parallelism and analog signal processing. A versatile reconfigurable circuit is presented that offers a variety of neural architectures: multilayer perceptron, template matching with winner-take-all (WTA) circuitry, and a new architecture of cascade backpropagation (CBP). Special designs of analog neuron and synapse implemented in VLSI are presented which bear out high speed response both at room and low temperatures with synapse-neuron signal propagation times of ∼100 ns. The CBP learning algorithm is illustrated by solving in simulation the nonlinear 6-bit parity problem. Results show that this algorithm is robust even with synaptic resolutions limited to 5 bits. Therefore, it is particularly suitable for hardware implementation.

Rogers, George W.; Solka, Jeffrey L.; Priebe, Carey E.

doi: 10.1177/003754979506500104pmid: N/A

A parallel distributed processing approach to the computation of localized fractal dimension values in imagery is pre sented. This approach is a further development of the covering method which requires only nearest neighbor commu nication. A major benefit of our approach is the ability to readily incorporate any boundary information that may be available. Man y fractal textures or surfaces are fractal only in distribution. With this in mind, we show that compari son of the fractal dimension distribrctions via Kullback-Leibler can give an improved texture discrimination capability over comparison of computed fractal dimension. Results are presented for a set of textures.

Micheli-Tzanakou, E.; Uyeda, E.; Ray, R.; Sharma, A.; Ramanujan, R.; Dong, J.

doi: 10.1177/003754979506500105pmid: N/A

In the last couple of decades, engineers, neuroscientists and psychologists have turned their attention to face recognition by humans and computer vision systems. Images of different complexities have been tested with a variety of methods. The goals of each research vary, as vary the applications. We present a neural method of recognizing faces using features obtained from compression of these faces with different methods. The extracted fea ti tres are used as inputs to a feedforward neural network. The neural network is trained with backpropagation and ALOPEX. Different types of featicre extraction are used and the results of training and testing for recognition based on the above mentioned methods are compared. ALOPEX converges much faster than backpropagation to a global maximum. Testing in both methods is as good as the learning of the network.

Loskiewicz-Buczak, Anna; Uhrig, Robert E.

doi: 10.1177/003754979506500106pmid: N/A

This paper describes novel multisensor information fusion methods based on fuzzy logic and genetic algorithms. Unlike most fuzzy logic-based systems that perform reasoning by fuzzy IF-THEN rules, the reasoning in this work takes place by means of fuzzy aggregation connectives. These connectives are capable of combining information not only by union and intersection used in traditional set theories but also by compensatory connectives that better mimic the human reasoning process. The particular connective used in this work for the purpose of data fusion is the generalized mean aggregation connective. The distinctive feature of this information fusion method is that the optimal parameters of the aggregation connective are automatically found by a genetic algorithm. Both elitist and nonelitist strategies for genetic algorithms are investigated. Two different methods are developed. The first technique performs aggregation of evidence from two sensors in one step; if there are more sensors, information from the next sensor is fused with the data already aggregated. The second technique developed performs one step fusion from all the sensors available. The techniques devised are tested on a vibration monitoring problem and the results are described.

Cooper, Mark G.

doi: 10.1177/003754979506500107pmid: N/A

In this article we demonstrate the application of genetic algorithms (GAs) to the automatic generation of fuzzy process controllers. Since each controller is represented as an unordered list of an arbitrary number of rules, the algorithm evolves both the composition and size of the rule base from initial randomness. Evolving controllers in the form of a rule base offers unique flexibility exceeding that of prior genetic efforts. The key to this methodology is the observation that the genetic algorithm does not merely evolve bit strings, but operates over a higher-level space of control rules. Both aspects are factors in the learning algorithm. To preserve rule integrity in a reproducing pair of strings, the combined loci must match semantically. This was the obstacle that hindered prior rule-based genetic fuzzy approaches. We demonstrate our algorithm by its application to the boat rudder control problem. We believe that this methodology has great potential for scalability since string size varies with the number of rules and not the number of variables or partitions. Finally, the method's generality permits its further application to the evolution of any system that can be specified as a set of rules.

Keller, Hartmut; Stolz, Horst; Ziegler, Andreas; Braunl, Thomas

doi: 10.1177/003754979506500108pmid: N/A

AERO is an animation system using physically based modelling. It allows the creation of virtual environrnents from scenes with simple three-dimensional geometrical objects (sphere, cylinder, box, point, plane). Objects can be linked to each other by a variety of methods (rod, spring, damper, joint). Realistic object movements are achieved by a simulation procedure, where forces can be applied to the objects in addition to gravity, friction, and air resistance. Wire frame animations can be generated in real time, depending on scene complexity and computer system performance. AERO generates scene description sequences, which can be used as input for ray tracing programs to generate photorealistic animations. AERO also enables stereo graphics output, camera control, and the mounting of the camera to an object in the scene.

doi: 10.1177/003754979506500109pmid: N/A

Gardenier, John S.

doi: 10.1177/003754979506500110pmid: N/A