Search

When dealing with conflicts, game theory and decision theory can be used to model the interactions of the decision-makers. To date, game theory and decision theory have received considerable modeling focus, while hypergame theory has not. A metagame, known as a hypergame, occurs when one player does not know or fully understand all the strategies of a game. Hypergame theory extends the advantages of game theory by allowing a player to outmaneuver an opponent and obtaining a more preferred outcome with a higher utility. The ability to outmaneuver an opponent occurs in the hypergame because the different views (perception or deception) of opponents are captured in the model, through the incorporation of information unknown to other players (misperception or intentional deception). The hypergame model more accurately provides solutions for complex theoretic modeling of conflicts than those modeled by game theory and excels where perception or information differences exist between players. This paper explores the current research in hypergame theory and presents a broad overview of the historical literature on hypergame theory.

The underlying goal of a competing agent in a discrete real-time strategy (RTS) game is to defeat an adversary. Strategic agents or participants must define an a priori plan to maneuver their resources in order to destroy the adversary and the adversary's resources as well as secure physical regions of the environment. This a priori plan can be generated by leveraging collected historical knowledge about the environment. This knowledge is then employed in the generation of a classification model for real-time decision-making in the RTS domain. The best way to generate a classification model for a complex problem domain depends on the characteristics of the solution space. An experimental method to determine solution space (search landscape) characteristics is through analysis of historical algorithm performance for solving the specific problem. We select a deterministic search technique and a stochastic search method for a priori classification model generation. These approaches are designed, implemented, and tested for a specific complex RTS game, Bos Wars. Their performance allows us to draw various conclusions about applying a competing agent in complex search landscapes associated with RTS games.

Effective and efficient Cyber-security management is of central importance in military environments. Many contemporary military communication control structures are based on static, hierarchical designs, which generally lack scalability and flexibility due centralization. Thus, we propose a self organized entangled hierarchial architecture of multiple agents that decentralizes network security control and communication. In particular, we focus on the military CyberCraft container structure. The self organized multi-agent swarms are evolved based on partially observable Markov decision process formal models. Desired swarm behaviors are formalized to interact with these models. The "optimal" policy (agent rules and parameters) for a given behavior is evolved using a multi-objective evolutionary algorithm. Swarm effectiveness is compared in numerous military network security scenarios using statistical testing techniques and visualization.

Cyber-security management is of central importance in financial, business, and military environments. Thus, we propose an effective and efficient self organized entangled hierarchial architecture of multiple agents that decentralizes network security control and communication. Considering the Cyberspace container model, the self organized multi-agent swarms are evolved based on partially observable Markov decision process formal models. Desired network security swarm behaviors are defined and formalized to interact with these models. The "optimal" policy (agent rules and parameters) for a given behavior is evolved using a multi-objective evolutionary algorithm. Swarm effectiveness is compared in various network security scenarios using statistical testing techniques and visualization.

This paper explores the cognitive domains of information, knowledge and intelligence for signal processing. Since the initial cognitive efforts into the field of multi-agent cooperation, research has begun to polarize in several different types of cooperation. The nature of image registration implies information processing that is cognitive in classifying sensory image data. Key functionality presented in this image registration flight control concept development technology roadmap include: image applications of neighborhood and formation algorithms; adaptive wavelet techniques for optimizing imagery transmission and processing; and application of emerging wavelet transformations for adaptive multiscale and multiresolution image applications.

Computer network Cyber-security is a very serious concern in many commercial, industrial, and military environments. This paper proposes a new computer network security approach defined by self organized agent swarms (SOMAS) which provides a novel computer network security management framework based upon desired overall system behaviors. The SOMAS structure evolves based upon the partially observable Markov decision process (POMDP) formal model and the more complex Interactive-POMDP and Decentralized-POMDP models. Example swarm specific and network based behaviors are formalized and simulated. This paper illustrates through various statistical testing techniques, the significance of this proposed SOMAS architecture.

A wide variety of signal and image processing applications, including the US Federal Bureau of Investigation's fingerprint compression standard [3] and the JPEG-2000 image compression standard [26], utilize wavelets. This paper describes new research that demonstrates how a genetic algorithm (GA) may be used to evolve transforms that outperform wavelets for satellite image compression and reconstruction under conditions subject to quantization error. The new approach builds upon prior work by simultaneously evolving real-valued coefficients representing matched forward and inverse transform pairs at each of three levels of a multi-resolution analysis (MRA) transform. The training data for this investigation consists of actual satellite photographs of strategic urban areas. Test results show that a dramatic reduction in the error present in reconstructed satellite images may be achieved without sacrificing the compression capabilities of the forward transform. The transforms evolved during this research outperform previous start-of-the-art solutions, which optimized coefficients for the reconstruction transform only. These transforms also outperform wavelets, reducing error by more than 0.76 dB at a quantization level of 64. In addition, transforms trained using representative satellite images do not perform quite as well when subsequently tested against images from other classes (such as fingerprints or portraits). This result suggests that the GA developed for this research is automatically learning to exploit specific attributes common to the class of images represented in the training population.

An agent system that develops and evolves its own structure can facilitate more accurate responses to complex environments. The purpose of the paper then is to explore this idea built upon our unmanned aerial vehicle (UAV) swarm model and simulation that uses autonomous self-organized concepts. The specific objective is to re-engineer this UAV foundation based upon a formal design model with focus on bio-inspired agent attack through emergent control structures. The overall design approach should give UAVs or generic agents the ability to not only react to dynamic environments but develop the controls in order to change behaviors spontaneously. To allow these behaviors to properly evolve, a multi-objective evolutionary algorithm generates a self-organized rule-based agent swarm. Heterogeneous UAV swarms are tested against difficult targeting scenarios that evolve specific attack behaviorial techniques. Statistical observations indicate that bio-inspired techniques integrated with the emerging entangled (non-hierarchical) framework provide desired complex UAV swarming behaviors in dynamic environments.

Many sectors of the military are interested in Self-Organized (SO) systems because of their flexibility, versatility and economics. The military is researching and employing autonomous and swarming ground robots, Unmanned Aerial Vehicles (UAVs) and Water Vehicles, medical agents, and 'Cyber-craft' security agents. The processes for effectively developing these systems are still in their infancy. Currently, little effort is focused on building simple agent rules with low-level SO systems communication in order to facilitate emergent behaviors. Note that only with the use of effective control structures can the full potential of these systems realized. Presented is an innovative new paradigm for developing SO-based autonomous vehicles. Using a formal design model, the Interactive Partially Observable Markov Decision Process, a full understanding of this SO domain is possible. With this design model and a focused effort on the minimization of computational and informational complexity, emergent entangled control hierarchies allow the SO rules to operate efficiently and effectively. This work extends the formal model decomposition technique, and in doing so ties in the information theoretic optimization to develop emergent structures. Preliminary computational results reflect limited success.

This paper summarizes the results of a continuing investigation into the evolution of transforms that minimize the error present in satellite images compressed and subsequently reconstructed under conditions subject to quantization error. Using coefficients describing the Daubechies-4 (D4) discrete wavelet transform (DWT) as a starting point, our genetic algorithm (GA) evolves real-valued coefficients describing matched forward and inverse transform pairs that reduce mean squared error (MSE) by 17.9% (0.86 dB) on satellite images used for training, and by an average of more than 11.0% (0.5 dB) on a large test set of satellite images. This result improves upon previous work on satellite images, which evolved only the reconstruction transform, and establishes evolutionary computation as a viable methodology for identifying state-of-the-art solutions to this difficult class of problems.

Interest in Self-Organized (SO) unmanned aerial vehicles (UAVs) systems is increasing because of their flexibility, versatility and economics. Many countries and industries are developing autonomous and swarming UAVs for reconnaissance, surveillance, intelligence gathering, and target engagement and neutralization. The processes for effectively developing these systems are still in their infancy. Currently, little effort is focused on building simple agent rules with low-level SO systems communication in order to facilitate emergent behaviors. Note that only with the use of effective control structures can the full potential of these systems realized. Presented is an innovative new paradigm for developing SO-based autonomous vehicles. Using a formal design model, the Interactive Partially Observable Markov Decision Process, a full understanding of this SO domain is possible. With this design model and a focused effort on the minimization of computational and informational complexity, emergent entangled control hierarchies allow the SO rules to operate efficiently and effectively. This work extends the formal model decomposition technique, and in doing so ties in the information theoretic optimization to develop emergent structures. Preliminary computational results reflect limited success.

In recent years, there has been increased interest in the use of evolutionary algorithms (EAs) in the design of robust image transforms for use in defense and security applications. An EA replaces the defining filter coefficients of a discrete wavelet transform (DWT) to provide improved image quality within bandwidth-limited image processing applications, such as the transmission of surveillance data by swarms of unmanned aerial vehicles (UAVs) over shared communication channels. The evolvability of image transform filters depends upon the properties of the underlying fitness landscape traversed by the evolutionary algorithm. The landscape topography determines the ease with which an optimization algorithm may identify highly-fit filters. The properties of a fitness landscape depend upon a chosen evaluation function defined over the space of possible solutions. Evaluation functions appropriate for image filter evolution include mean squared error (MSE), the universal image quality index (UQI), peak signal-to-noise ratio (PSNR), and average absolute pixel error (AAPE). We conduct a theoretical comparison of these image quality measures using random walks through fitness landscapes defined over each evaluation function. This analysis allows us to compare the relative evolvability of the various potential image quality measures by examining fitness topology for each measure in terms of ruggedness and deceptiveness. A theoretical understanding of the topology of fitness landscapes aids in the design of evolutionary algorithms capable of identifying near-optimal image transforms suitable for deployment in defense and security applications of image processing.

Government, commercial, scientific, and defense applications in image processing often require transmission of large amounts of data across bandwidth-limited channels. Applications require robust transforms simultaneously minimizing bandwidth requirements and image resolution loss. Image processing algorithms take advantage of quantization to provide substantial lossy compression ratios at the expense of resolution. Recent research demonstrates that genetic algorithms evolve filters outperforming standard discrete wavelet transforms in conditions subject to high quantization error. While evolved filters improve overall image quality, wavelet filters typically provide a superior high frequency response, demonstrating improved reconstruction near the edges of objects within an image. This paper presents an algorithm to generate transform filters that optimize edge reconstruction, improving object edge resolution by an average of 17%. Edges within satellite images are isolated, and image transforms are evolved to optimize both the edge and non-edge portions of reconnaissance photographs. Such filters provide an increased object resolution over standard wavelets and traditionally evolved filters for varied applications of image processing.

Modern image processing applications often require robust performance in noisy or bandwidth-limited situations. In this research, we employ genetic algorithms (GAs) to evolve image transforms that reduce quantization error in reconstructed signals and images. The resulting transforms produce higher quality images than current wavelet-based transforms at a given compression ratio and thus allow transmission of compressed data at a lower bandwidth. We evaluate state-of-the-art variation operators for evolving reconstruction filters. Our results indicate that the careful selection of these operators has a strong positive effect upon the evolutionary search for superior image transforms.

Government, commercial, scientific, and defense applications inimage processing often require transmission of large amounts of data across bandwidth-limited channels. Applications require robust transforms simultaneously minimizing bandwidth requirements and image resolution loss. Image processing algorithms take advantage of quantization to provide substantial lossy compression ratios at the expense of resolution. Recent research demonstrates that genetic algorithms evolve filters outperforming standard discrete wavelet transforms in conditions subject to high quantization error. While evolved filters improve overall image quality, wavelet filters typically provide a superior high frequency response, demonstrating improved reconstruction near the edges of objects within an image. This paper presents an algorithm to generate transform filters that optimize edge reconstruction, improving object edge resolution by up to 24%. Such filters provide an increased object resolution over standard wavelets and traditionally evolved filters for varied applications of image processing.

In recent years, wavelets have been widely applied instate-of-the-art image processing algorithms, providing efficient compression while maintaining superior image quality. However, wavelet performance may not be sufficient when extreme compression ratios are required. Defense applications often require robust transforms simultaneously minimizing bandwidth requirements and image resolution loss. Image processing algorithms take advantage of quantization to provide substantial lossy compression ratios at the expense of resolution. Recent research demonstrates that genetic algorithms (GAs) evolve filters out performing standard discrete wavelet transforms in conditions subject to high quantization error. Evolved filters must be trained using images appropriate to their intended application. We present a set offifty satellite images used to evolve image transforms appropriate for satellite and unmanned aerial vehicle (UAV) reconnaissance applications. We identify the best training and test images. Image transforms evolved using appropriate training images reduce the mean squared error (MSE) by an average of greater than 15% across the entire image set under conditions subject to high quantization error.

Military imaging systems often require the transmission of copious amounts of data in noisy or bandwidth-limited situations. High rates of lossy image compression may be achieved through the use of quantization at the expense of resulting image quality. We employ genetic algorithms (GAs) to evolve military-grade transforms capable of improving reconstruction of satellite reconnaissance images under conditions subject to high quantization error. The resulting transforms outperform existing wavelet transforms at a given compression ratio allowing transmission of data at a lower bandwidth. Because GAs are notoriously difficult to tune, the selection of appropriate variation operators is critical when designing GAs for military-grade algorithm development. We test several state-of-the-art real-coded crossover and mutation operators to develop an evolutionary system capable of producing transforms providing robust performance over a set of fifty satellite images of military interest. With appropriate operators, evolved filters consistently provide an average mean squared error (MSE) reduction greater than 17% over the original wavelet transform. By improving image quality, evolved transforms increase the amount of intelligence that may be obtained reconstructed images.

Contemporary signature-based intrusion detection systems are reactive in nature and are storage-limited. Their operation depends upon identifying an instance of an intrusion or virus and encoding it into a signature that is stored in its anomaly database, providing a window of vulnerability to computer systems during this time. Further, the maximum size of an Internet Protocol-based message requires a huge database in order to maintain possible signature combinations. To tighten this response cycle within storage constraints, this paper presents an innovative artificial immune system (AIS) integrated with a multiobjective evolutionary algorithm (MOEA). This new distributed intrusion detection system (IDS) design is intended to measure the vector of tradeoff solutions among detectors with regard to two independent objectives: best classification fitness and multiobjective hypervolume size. AIS antibody detectors promiscuously monitor network traffic for exact and variant abnormal system events based on only the detector's own data structure and the application domain truth set. Applied to the MIT-DARPA 1999 insider intrusion detection data set, this new software engineered AIS-MOEA IDS called jREMISA correctly classifies normal and abnormal events at a relative high statistical level which is directly attributed to finding the proper detector affinity threshold.

Lossy image compression algorithms sacrifice perfect image reconstruction in favor of decreased storage requirements. Previous research demonstrates that a genetic algorithm can improve image reconstruction in the presence of quantization error by replacing the wavelet reconstruction coefficients with a set of evolved coefficients. This paper expands previous research efforts by using an improved fitness function, exploring standard versus local genetic search operators, and evolving coefficient sets that perform quite well for multi-resolution analysis (MRA). Test results indicate that our improved evolutionary system consistently outperforms the standard discrete wavelet transform (DWT) for image reconstruction under compression conditions which are subject to quantization error.

In the search for a robust and efficient algorithm to be used for computer virus detection, we have developed an artificial immune system genetic algorithm (REALGO) based on the human immune system's use of reverse transcription ribonucleic acid (RNA). The REALGO algorithm provides memory such that during a complex search the algorithm can revert back to and attempt to mutate in a different "direction" in order to escape local minima. In lieu of non-existing virus generic templates, validation is addressed by using an appropriate variety of function optimizations with landscapes believed to be similar to that of virus detection. It is empirically shown that the REALGO algorithm finds "better" solutions than other evolutionary strategies in four out of eight test functions and finds equally "good" solutions in the remaining four optimization problems.

Deception problems are among the hardest problems to solve using ordinary genetic algorithms. Recent studies show that Bayesian optimization can help in solving these problems. This work compares the results acquired from the multiobjective fast messy genetic algorithm (MOMGA-II), multiobjective Bayesian optimization algorithm (mBOA), and the nondominated sorting genetic algorithm-II (NSGA-II) when applied to three different deception problems. The three deceptive problems studies are: interleaved minimal deceptive problem, interleaved 5-bit trap function, and the interleaved 6-bit bipolar function. The unmodified MOMGA-II, by design, explicitly learns building block linkages which is required if an algorithm is to solve these hard deception problems. Preliminary results using the MOMGA-II are favorable.

Deception problems are among the hardest problems to solve using ordinary genetic algorithms. Designed to simulate a high degree of epistasis, these deception problems imitate extremely difficult real world problems. [1]. Studies show that Bayesian optimization and explicit building block manipulation algorithms, like the fast messy genetic algorithm (fmGA), can help in solving these problems. This paper compares the results acquired from an extended multiobjective fast messy genetic algorithm (MOMGA-IIa), ordinary multiobjective fast messy genetic algorithm (MOMGA-II), multiobjective Bayesian optimization algorithm (mBOA), and the non-dominated sorting genetic algorithm-II (NSGA-II) when applied to three different deception problems. The extended MOMGA-II is enhanced with a new technique exploiting the fmGA's basis function to improve partitioned searching in both the genotype and phenotype domain. The three deceptive problems studied are: interleaved minimal deceptive problem, interleaved 5-bit trap function, and interleaved 6-bit bipolar function. The unmodified MOMGA-II, by design, explicitly learns building block linkages, a requirement if an algorithm is to solve these hard deception problems. Results using the MOMGA-IIa are excellent when compared to the non-explicit building block algorithm results of both the mBOA and NSGA-II.

Interest in protein structure prediction is widespread, and has been previously addressed using evolutionary algorithms, such as the simple genetic algorithm (GA), messy GA (mga), fast messy GA (fmGA), and linkage learning GA (LLGA). However, past research used off the shelf software such as GENOCOP, GENESIS, and mGA. In this study we report results of a modified fmGA, which is found to be "good" at finding semi-optimal solutions in a reasonable time. Our study focuses on tuning this fmGA in an attempt to improve the effectiveness and efficiency of the algorithm in solving a protein structure and in finding better ways to identify secondary structures. Problem definition, protein model representation, mapping to algorithm domain, tool selection modifications and conducted experiments are discussed.

Publisher Summary The protein folding problem (PFP) involves the determination of the 3D structures of proteins given only their amino acid sequences. One reason for the significant interest in general and efficient techniques to solve this type of problem is that such knowledge would facilitate understanding of the tremendous amount of genetic information produced by the Human Genome Project. This chapter refers to efforts aimed at identifying in vivo structures of naturally occurring proteins or structures of arbitrary polypeptides in arbitrary environments. In particular, it looks at the general problem of identifying three-dimensional structures of arbitrary polypeptides in arbitrary environments, and focuses on a particular kind of evolutionary algorithm (the messy genetic algorithm) for use on this task. It concentrates on improved effectiveness through the exploitation of domain constraints (i.e., dihedral-angle constraints inspired by the Ramachandran plot) and the exploitation of prior secondary structure analysis. The work described is performed on high-grade parallel and heterogeneous computing resources.

Solving optimization problems with multiple (often conflicting) objectives is generally a quite difficult goal. Evolutionary algorithms (EAs) were initially extended and applied during the mid-eighties in an attempt to stochastically solve problems of this generic class. During the past decade a multiplicity of multiobjective EA (MOEA) techniques have been proposed and applied to many scientific and engineering applications. Our discussion's intent is to rigorously define and execute a quantitative MOEA performance comparison methodology. Almost all comparisons cited in the current literature visually compare algorithmic results, resulting in only relative conclusions. Our methodology gives a basis for absolute conclusions regarding MOEA performance. Selected results from its execution with four MOEAs are presented and described.

Swarm behavior as demonstrated by flocks of birds, schools of fish, and swarms of insects provide a useful method for implementing a distributed network of mobile sensor platforms. Such mobile sensor swarm systems are useful for various search or surveillance activities. Swarm behavior ensures safe separation between swarm members while enforcing a level of cohesion. These two properties, when considered in the context of sensors and wireless communications, provide for low redundancy coverage and a robust and reliable communications system. This paper examines particle swarm behavior through simulation with respect to such a sensor network. Analysis of swarm behavior for various parameter settings indicate a classification methodology. This provides a foundation for a proposed taxonomy.

The concepts of efficiency and effectiveness must be addressed in conducting research into using a Evolutionary Algorithm (EA) for optimization problems. The increased use of evolutionary approaches for real-world applications, containing multiple objectives and high dimensionality, has led to the design and generation of a number of Multiobjective Evolutionary Algorithms (MOEA). When analyzing these algorithms, the issues of effectiveness and efficiency are extremely important and typically drive the urge to parallelize these algorithms. The parallelization of MOEAs is a relatively new concept, with few researchers contributing work in this area. This parallelization process is not a simple task and involves the analysis of various parallel models and the parameters associated with these models. This paper presents a thorough analysis of the various parallel MOEA models, the issues associated with these models and recommendations for using these models in MOEAs. In particular, these parallelization concepts are applied to the Multiobjective Messy Genetic Algorithm II.

Why test MOEAs? To evaluate, compare, classify, and improve algorithm performance (effectiveness and efficiency)! What is an MOEA test? An MOP test function, an MOP test suite, pedagogical functions, or a real-world problem? How to find an appropriate MOEA test? MOEA literature, historical use, test generators, or well known real-world applications. When to test? Incremental algorithm and test development starting early or wait until end of development. How to design an MOEA test? Assumptions, computational platform selection, statistical tools, metric selection, experimental plan, and an on-going process. Therefore, considerable effort must be expended not only to define proper MOP tests and generate the proper design of MOEA experiments, but also to employ the honest selection of appropriate testing metrics and associated statistical evaluation and comparison. In this chapter, the development of various MOP test suites is addressed, and in the next chapter, their use in appropriate MOEA evaluations is discussed.

Regarding the scientific method of experimentation, it is desirable to construct an accurate, reliable, consistent and non-arbitrary representation of MOEA architectures and performance over sets of MOPs. In particular, through the use of standard procedures and criteria, one should attempt to minimize the influence of bias or prejudice of the experimenter when testing an MOEA hypothesis. The design of each experiment must conform then to an accepted “standard” approach as reflected in any generic scientific method. When employing this approach, the detailed design of MOEA experiments can draw heavily from outlines presented by Barr et al. (1995), and Jackson et al. (1991). These generic articles discuss computational experiment design for heuristic methods, providing guidelines for reporting results and ensuring their reproducibility. Specifically, they suggest that a well-designed experiment follow these steps: (1) Define experimental goals; (2) Choose measures of performance (metrics); (3) Design and execute the experiment; (4) Analyze data and draw conclusions; and (5) Report experimental results. This chapter applies these concepts in developing experimental MOEA testing procedures using appropriate MOP test suites from Chapter 3.

Solving optimization problems with multiple (often conflicting) objectives is, generally, a very difficult goal. Evolutionary algorithms (EAs) were initially extended and applied during the mid-eighties in an attempt to stochastically solve problems of this generic class. During the past decade, a variety of multiobjective EA (MOEA) techniques have been proposed and applied to many scientific and engineering applications. Our discussion's intent is to rigorously define multiobjective optimization problems and certain related concepts, present an MOEA classification scheme, and evaluate the variety of contemporary MOEAs. Current MOEA theoretical developments are evaluated; specific topics addressed include fitness functions, Pareto ranking, niching, fitness sharing, mating restriction, and secondary populations. Since the development and application of MOEAs is a dynamic and rapidly growing activity, we focus on key analytical insights based upon critical MOEA evaluation of current research and applications. Recommended MOEA designs are presented, along with conclusions and recommendations for future work.

The induction of decision rules from data is important to many disciplines, including artificial intelligence and pattern recognition. To improve the state of the art in this area, we introduced the genetic rule and classifier construction environment (GRaCCE). It was previously shown that GRaCCE consistently evolved decision rule sets from data, which were significantly more compact than those produced by other methods (such as decision tree algorithms). The primary disadvantage of GRaCCe, however, is its relatively poor run-time execution performance. In this paper, a concurrent version of the GRaCCE architecture is introduced, which improves the efficiency of the original algorithm. A prototype of the algorithm is tested on an in- house parallel processor configuration and the results are discussed.

This paper provides insight into combining stochastic and deterministic search methods using evolutionary algorithms (EAs) such as evolutionary programming, evolutionary strategies, and genetic algorithms integrated with depth-first search with backtracking, branch and bound, and best-first search algorithms such as A*. An important view of such an integration focuses on both theoretical analysis and experimental evaluation. Included in the discussion is the constraining impact of the “No Free Lunch Theorem” on combined search performance. Also, a variety of combinations of EAs and deterministic combined search methods are proposed along with expert system components. Discussion of anticipated results of such architectures speculates about existence of high performance integrated search environments. A particular successful specific NPC problem environment is presented that employs search metrics for short-term performance evaluation and combined search algorithm selection.

Contemporary computer-aided design (CAD) programs incorporate excellent functional tools but can be enhanced through Artificial Intelligence (AI), particularly expert systems. Expert systems(ES) for design, however, must incorporate "design process models" explicitly. To study the various phases of such a development, a specific expert system has been generated to aid the control engineer in some aspects of control system design. Conventional and AI programming concepts are combined to generate a generic "design associate". Through this development insight evolved regarding concepts, modeling and implementation of general expert systems for design. This paper discusses these aspects as they relate to control system design as well as to the general area of AI expert systems for CAD. The various components of expert system organization including human-machine interfacing, search processing and knowledge representation selections are analyzed with a general ES architecture for CAD control engineering applications provided.