Research Objectives and Significance:
The goal of our research in this area is to examine the design of relevance filtering schemes, analyze their reliability and evaluate and improve their filtering efficiency. The successful implementation of efficient relevance filtering in network gateways would help solve one of the challenges facing the design of highly scalable DIS and HLA systems. In HLA, the Data Distribution Management is the RTI component responsible for increasing the scalability of the training exercise and reducing the traffic among federates. Filtering is obtained by allowing federates to declare and use routing spaces and custom filtering schemes. Improving the effectiveness of filtering greatly improves HLA scalability.
Army Relevance:
The research on filtering and routing spaces is crucial to the scalability of HLA currently being developed under the auspices of the Defense Modeling and Simulation Office as the DoD-wide standard for simulation. The results of this research have provided good insight into the behavior of relevance filtering schemes, their real-time performance and their reliability. With this insight, the development of reliable and efficient filtering methods is greatly enhanced.
Methodology:
An HLA routing space is a multidimensional coordinate system by which federates can implement a filtering mechanism without requiring the RTI to have complex domain specific knowledge. The three components of a vector location are used as the three variables of a three-dimensional routing space. We have used a circular region of interest to determine the relevancy of transmitted messages and implemented filtering algorithms at the sending receiving gateways. Both analysis and simulation have been used to study the performance and reliability of relevance filtering algorithms.
Accomplishments:
Our two-stage filtering approach (i.e., filtering-at-transmission and filtering-at-reception) is in agreement with the hierarchical filtering concept of HLA. The distance-based extrapolation algorithms that we used in local and gateway dead-reckoning gave good results and are suitable for straightforward implementation in RTI. We designed and tested a number of reliable filtering methods for ground vehicles. Although work on the traffic reduction aspects of relevance filtering has received its due attention, the analysis of filtering errors and of the design of reliable filtering algorithms are notably lacking. Our reliability analysis provides a first step toward gaining good insight into the problem of filtering errors as well as methods to solve this problem. We have also examined the issue of improving filtering efficiency by clustering, i.e., by the careful assignment of the simulators in DIS/HLA LANs to the vehicles in the virtual terrain. Our results show that mapping the entities located in the same terrain region to hosts located in the same LAN helps to localize the set of hosts that would need to receive state update messages from each entity. Training exercises based on such initial mapping will have better intrinsic scalability because the filtering scheme will be able to cancel much of the WAN traffic.
Planned Activities:
We plan to extend our two-stage filtering scheme by adapting it to the routing space mechanism of HLA, investigating the impact of multicast algorithms on its performance and performing reliability analysis of filtering in HLA.
Research Objectives and Significance:
Data Distribution Management (DDM) is one of the six service categories defined in the HLA Interface Specification. Its purpose is to increase the scalability of large scale distributed simulations by reducing the amount of irrelevant messages exchanged among the federates that are participating in a simulation. The goal of this project is to examine the different approaches that have been taken/proposed for the current DDM and devise alternative methods that can improve their accuracy and filtering efficiency.
Army Relevance:
The results of this project can improve the performance and scalability of large distributed simulations.
Methodology:
We plan to evaluate the filtering efficiency of current DDM services. In particular, we will investigate the effects of network latency and dynamically changing subscription and update regions on the performance of DDM. Based on the evaluation study we will investigate methods for determining safety regions (for subscription/update) that provide efficient and physically correct filtering. We will also examine approaches for determining overlapping regions and clustering methods for multicast group allocation as an alternative to the gridded filtering approach.
Accomplishments:
We conducted a simulation-based performance evaluation study of the grid-based approach implemented in the STOW RTI. Our study suggested that the filtering efficiency of this approach is determined by the number of multicast group addresses available at the wide-area network level. For smaller number of multicast group addresses, a larger grid cell size is used which causes spurious connections. We formulated a two-level grid-based filtering approach that takes advantage of the properties of the underlying network architecture to reduce the number of spurious connections. We developed a simulator to analyze the grid-based filtering and the two-level grid-based approaches.
Planned Activities:
In this research, we plan to evaluate and analyze various filtering approaches. The two-level grid-based filtering approach will be evaluated and compared to the single level approach. We will also investigate alternative approaches that attempt to cluster objects based on their dynamically changing characteristics.
Research objectives and significance:
There are various applications of weighted network, since the edge weights may represent time, cost, penalty, loss or any other quantities that can be associated with. Large-scale sparse network optimization problems can be found in many application areas such as transportation, communication, computer networks, and etc. In this research, we try to develop general parallel algorithms to optimize nonstructured large-scale sparse network.
Army relevance:
The Army has many different networks such as transportation, distribution, communication, computer and other networks. The algorithm we studied can be applied to such networks and to effectively find optimal solutions to meet the requirement.
Methodology:
We define an index array to represent large-scale non-structured, sparse, weighted network G(V,E,w). We, then, studied the all-pairs shortest paths problem for large-scale sparse networks. We refine the matrix-multiplication method to solve the problem parallelly.
Accomplishments:
Different parallel algorithm are developed to find all-pairs shortest paths problem for large-scale generic sparse networks.
Planned Activities:
Other network optimization problems are going to be studied. Implementations and tests will be done with application problems.
The goal of this projects is to develop an alternate flame model using a fuzzy controller and FuzzyCLIPS to facilitate real-time simulations. The controller is being design to take its input from the computation Ising model. A fuzzy controller modeling is a useful and efficient approach for modeling control systems with complicated description of plans, or course of actions, and mathematical models that are difficult to develop. Since fuzzy models require only simple computations, they can perform very efficiently in real-time simulations.
Army Relevance:
The results of this task would provide an alternative way of simulating flames, to complement the Ising-based smoke model, in a distributed simulation environment. High speed performance and visualization of phenomenological models in virtual battlefield is important to the training and readiness of soldiers to the Army.
Methodology:
We are developing a fuzzy flame model at the particle level using physical-based models, namely the parallel version of the Ising model. A design of a fuzzy flame controller consists of determining parameters that affect temperature and particle motions in both vertical and horizontal directions. It is expected that starting with data from the Ising model, which incorporates the physical properties of particles that constitute the flame, the fuzzy controller can generate flame configurations in near-real time without intensive calculations. The development of test data using OpenGL (for the graphical visualization of flame behavior) and FuzzyCLIPS (for the fuzzification component) will involve students.
Accomplishments:
We have developed a model for flame simulation with respect to
flame particles. Graphical simulations using OpenGL have been done on
SGI computers and the model was tested with a simple parameter –
distance from a designated heat source. The ideas and concepts for
formulating the fuzzy model for flame simulation were discussed at the
investigators meeting during the fall. Continued collaborative
discussions and efforts are ongoing.
Research Objectives and Significance:
This effort extends and analyzes the mathematical modeling techniques based on the Ising model for simulating smoke and fuzzy controller modeling techniques for simulating flames in real-time simulations. The Ising model for simulating smoke has been modified to simulate flames as a collection of particles with temperature as a dominant parameter. The flames produced from this temperature-modified version of the Ising model are called Ising flames. An alternative flame model is being developed in relation to a fuzzy controller and FuzzyCLIPS that can be implemented in a dynamic terrain for real-time simulation. Flames simulated in this modeling effort are called fuzzy flames. The main objective of this research effort is to analyze the real-time performance characteristics of Ising flames and fuzzy flames and to compare the fidelity of their graphical representations of true flames.
Army Relevance:
This incorporation of phenomenological elements like fire, smoke, and flames into the virtual battlefield will help to create efficient and safer training programs for the Army. This will also allow soldiers to acquire military skills which will improve their readiness for battle. The Army will benefit by having better trained soldiers with greater readiness.
Methodology:
The mathematical analysis of the Ising model using matrix analysis has been modified to reflect temperature ranges appropriate to flame production. The effect of the temperature T is emphasized and used exclusively as a major parameter in the modeling process. The fuzzy controller has been used for modeling flames because it is a useful technique for modeling systems for which mathematical models are difficult to develop. Since fuzzy models require only simple computations, they can perform very efficiently in real-time simulations. This comparative study will compare these two approaches. The final phase of this task is to present both flames in a common graphical environment where comparisons can be effectively made.
Accomplishments:
A two-dimensional simulation of smoke based upon the Ising model
has been implemented in C code in the SGI environment. Also a more
precise mathematical formulation of the simulation problem has been
completed. A fuzzy controller model for flames has been developed. A
common graphical presentation mechanism will be developed to present
the Ising flames and the fuzzy flames for common analysis.
Objectives and Significance:
The main objective for the first and second period was to acquire knowledge in advanced distributed simulation through training and workshops, and to conduct weekly research discussions as well as a weekly seminars to give both faculty and students necessary tools for initiating basic research problems in AI reasoning, graphics, and simulation relevant to DIS. Our focuses were on developing a prototype of an intelligent autonomous control system using fuzzy logic, rough sets, neural networks, FuzzyCLIPS and OpenGL to automate entities in simulations; and on developing modeling tools for systems under uncertainty using Petri nets, fuzzy logic, and Markov chains.
We are undertaking investigations into the use of Analytical Hierarchical Processing (AHP) in the decision making process and into the use of fuzzy logic to improve the performance of accessing resources using object oriented control mechanisms. Many problems in Decision Making structure reality into constituent parts, that is, into hierarchical charts. An important step in the AHP approach is to construct a pairwise ranking. Since the latter may be often difficult to obtain, a fuzzy pairwise ranking is under study. Concurrency control is used to access resources more efficiently. However concurrency control requires information about the data base, the type of transactions and heuristics. Such information is typically only partially known and therefore decision making under uncertainty is required. We are also investigating ways of applying Data Mining to various military databases and to apply adaptive search strategies (Genetic Algorithms) for optimizing Large Join Queries which are frequent in high functionality database systems such as data warehouses etc. We are also undertaking methods of training fuzzy neural nets that are not limited to one output neuron and methods.
ADSRC laboratories were housed and setup with a network of SGI, HP and Pentium based machines for doing research and training faculty and students in advanced distributed simulation. A Special Projects Course was introduced for Research Assistants which focuses on the development of student projects undertaken with Linkage Partners. Through faculty meetings, students were provided with background in rough sets, fuzzy logic, neural networks, parallel computation, and interval computation.
Army Relevance:
Our initial development is in AI reasoning for intelligent control systems using fuzzy logic and fuzzy-neural networks. Rule based language, FuzzyCLIPS, and OpenGL are the tools used in simulation to demonstrate the positive result of a developed intelligent control prototype in route planning. It is expected that this result will be extended to DIS -HLA environments such as ModSAF to enhance simulation. This development also gave the opportunity for our students to become familiar with simulation which is one of the ADSRC goals to increase human resources in DIS community. When tactical factors are to be considered in military decisions, pairwise comparisons between such factors might be far from precisely known and therefore fuzzy comparisons form a natural setting. Hence AHP appears to be a very promising tool for tactical decisions. Typically information arrives simultaneously from different sources and method of processing such parallel inputs is necessary. Our studies of concurrency attempts to address these problems. Data Mining techniques and Genetic Algorithms allow us to explore and improve performance in huge military databases.
Accomplishments:
Several teams have been formed to study decision making under
uncertainty. Different directions were explored such as fuzzy logic,
neural networks, AHP, concurrency control, Petri nets and Markov
chains and hybrid methods combining these. An object oriented approach
was sought to implement some of the above techniques. Investigations
are being undertaken in the use of Genetic Algorithms, Data Mining and
Concurrency Control Mechanisms.
Research Objective and Significance:
The objective of this project is to develop a prototype using fuzzy logic and neural networks for autonomous control systems in navigating vehicles and route planning. A general autonomous navigation scheme was formulated which mimics how humans behave in maneuvering a vehicle. Our scheme implements fuzzy if-then rules that characterize human’s behaviors to autonomously control a vehicle. These fuzzy rules can be extended with minor modifications for any type of vehicles or planes, and any environment or terrain. A set of fuzzy-neural networks with modified backpropagation was designed to learn these fuzzy if-then rules. Finally, computer simulations were performed on a vehicle with different road conditions on a given terrain to exhibit our model. We used OpenGL and FuzzyCLIPS for our simulations.
Army Relevance:
The results of this project provide an alternative way to navigate
a vehicle in distributed simulations. This result can also be extended
to model any plan actions with route planing simulations. Since fuzzy
rules are obtained from experts in the form of linguistic information,
our model can be very flexible and reliable.
Methodology:
We developed a general autonomous navigation scheme that mimics human behaviors in maneuvering a vehicle. This scheme provides a set of fuzzy rules in the form of if-then rules. A fuzzy-neural network with modified backpropagation was then formulated and trained with this set of rules. We performed simulations using OpenGL and FuzzyCLIPS to test our results.
Accomplishment:
We have developed a prototype for autonomous
control systems using fuzzy-neural networks in navigating vehicles and
route planning. A general autonomous navigation scheme was formulated
that mimics human behaviors in maneuvering a vehicle. A set of fuzzy
rules in the form of if-then rules can be obtained form this scheme.
These fuzzy rules can be extended with minor modifications for any
type of vehicles or planes, and any environment or terrain. A set of
fuzzy-neural networks with modified backpropagation was designed to
learn these fuzzy if-then rules
Research Objectives and Significance:
Object-oriented database systems deal with complex data types hence involve long transactions. Long running transactions often imply the use of a large number of resources that are inaccessible to other incoming transactions. Concurrency Control Mechanisms may be used to access resources more efficiently. However, concurrency control may require some information about the database, type of transactions and heuristics. In many situations the information needed for concurrency control is not fully available. In such cases fuzzy logic can be used to construct useful estimates. The goal of this research is to improve the performance by applying fuzzy logic a concurrency control mechanism.
Army Relevance:
This work may have applications for some of HLA/RTI services such as Ownership Management Services. Further research in this area is needed.
Methodology:
The data model presented in this research is a semantic data model because it supports navigational processing. The idea is basically as follows: There is a threshold for when we let the transactions start execution (i.e., rare would be an example). Each transaction is associated with some information, I1, I2, ...., Ik . The idea here is to compare the fuzzy information values I1 ... Ik with the fuzzy threshold. If I1 .... Ik exceeds threshold, do something about transaction T. From history we have the fuzzy probabilities about the database files or their records. We must compare expected penalty with what price (time) we are willing to pay.
Accomplishments:
We have started identifying some of the parameters that may have
fuzzy values and could be included in the algorithm for fuzzy
concurrency control.
Research Objectives and Significance:
Many problems in Decision Making structure reality into constituent parts, that is, into hierarchical charts. The goal is at the top of the chart and the decisions are at the bottom of the chart. An important step is to perform a pair wise comparison between constituent parts at the same level and mesh these comparison to order the decisions. Often, even experts, do not precisely know the scalars to use for pairing off these parts. In the present work, we replace scalars by fuzzy expressions and develop two methods to mesh these comparisons into a decision making step.
Army Relevance:
The present work should have interesting applications to tactical decision making where it is often difficult to make precise comparisons between somewhat elusive factors that may involve training, strategic position, economic support, troop morale, etc.
Methodology:
While the classical scale for comparisons is from 1 to 10, in our work fuzzy terms such as more important , somewhat similar, much worse, etc. are used. The meshing for the decision making step is done by appending the fuzzy expected value of each decision (or rather an expression of that type). And then compare these values by using Jain´s Maximizing Sets.
Accomplishments:
A paper has been completed and has been submitted for publication.
M. Beheshti, A. Berrached, M. Barrientos, A. Guerra (UHD)
Objectives and Significance:
A project was undertaken by Drs. Moshen Beheshti and Ali Berrached in consultation with Lt. Col. George Stone of STRICOM to organize the Task descriptions of the Army Units/Missions in such a way that 1) the information is easily accessible and 2) the amount of required storage space are optimized. This project was initiated in August of 1997 and is a current project.
Accomplishments:
A survey was conducted to see where the information currently resides and it was found that information is currently stored using a Word Processor (Microsoft Word) and data is being duplicated at different places. Therefore, a large amount of disk space is needed to store this information. A database manager is being designed to take advantage of the hierarchical structure of information (each Unit has a number of missions; each mission consists of a sequence of tasks; each task is made up of a sequence of steps; etc.) and take advantage of the redundancy of information (different units may have similar missions, different missions may have similar tasks, etc.) and to eliminate the redundancy in the information. Visual Basic is used as a window-based front end to allow users to query the database. The first prototype of this software is to be completed by mid-November 1997 for testing. The prototype will include few Units, Missions and Tasks to demonstrate the effectiveness of the software. This software could also be used as a tool to assist users creating scenarios.
Simulation Information Filtering Tool (SIFT/Janus)
M. Beheshti, A. Berrached, M. Carpenter, R. Mata, D. Smith (UHD)
Task Objectives and Significance:
A project was undertaken by Drs. Moshen Beheshti and Ali Berrached in consultation with Lt. Col. George Stone and Major Wayne Stillwell of STRICOM to familiarize students with various distributed simulations and software testing procedures. This project was initiated in August of 1997 and is a current project.
Accomplishments:
Janus and Simulation Information Filtering Tool (SIFT) software packages were loaded on a HP machine at UHD. SIFT interfaces with Janus and generates simulation analysis dynamically (i.e. while the simulation is running). Both the Janus and SIFT packages were developed at STRICOM and UHD has been designated as one of the testing sites for both packages. Students and faculty learned how to create scenarios during a tutorial session on Janus and SIFT given by LTC Stone and MJR. Stillwell of STRICOM. Currently, students are experimenting with the software and generating test data using Janus and SIFT testing will begin as soon as benchmark test data are available.
R. Aló, A. Guerra, R. Ramirez, C. Trauschke, (UHD)
Objectives and Significance:
A project was undertaken by Dr. Richard Alo in consultation with Dr. Owen Deutch, SCAD Project Head, Draper Laboratories, Cambridge, Massachusetts to expose and familiarize students with various software testing procedures. This project was conducted in the summer of 1997.
Accomplishments:
Intervisibility is the ability to determine line of sight for a particular object. The importance of time complexity with intervisibility stems from being able to identify a friend or foe in real-time. In order to simulate this action, determination of time-complexity for each algorithm is necessary. The purpose of this project was to determine which intervisibility algorithm has the best time-complexity. Three algorithms were suggested, the Timer Algorithm, Sieve Algorithm, and the Draper Algorithm, but only two were used. The Sieve algorithm was pending funding, so access to the code was denied. After careful construction and testing of the Timer and Draper algorithm, it was concluded that the Timer algorithm proved to be the better of the two. The time complexity for the Timer algorithm was considerably less than the Draper algorithm.