Intelligent Computational Systems: A Multi-Disciplinary Perspective

By Bentham Science Publishers

Intelligent Computational Systems presents current and future developments in artificial Intelligence (AI) in a multi-disciplinary context. Readers will learn about the pervasive and ubiquitous roles of artificial intelligence and gain a perspective about the need for intelligent systems to behave rationally when interacting with humans in complex and realistic domains.
This reference covers widespread applications of AI discussed in 11 chapters which cover topics such as AI and behavioral simulations, AI schools, automated negotiation, language analysis and learning, financial prediction, sensor management, Multi-agent systems, and much more.
This reference work will assist researchers, advanced-level students and practitioners in information technology and computer science fields interested in the broad applications of AI.

• Language: English
• Publisher: Bentham Science Publishers
• Release date: Aug 7, 2017
• ISBN: 9781681085029

Book preview

PART I:

SIMULATION

Simulation, Intelligence and Agents: Exploring the Synergy

Nasser Ghasem-Aghaee¹, ², *, Tuncer Ören³, Levent Yilmaz⁴

¹ Department of Computer Engineering, Sheikh Bahaei University, Baharestan, Iran

² Faculty of Computer Engineering & Information Technology, University of Isfahan, Isfahan, Iran

³ School of Electrical Engg. and Computer Science, University of Ottawa, Ottawa, ON, Canada

⁴ Computer Science and Software Engineering, Samuel Ginn College of Engineering, Auburn University, Auburn, AL, USA

Abstract

Simulation is applied to exhibit the extent of how the offered valuable functionalities of given issues are appreciated. A systematic glossary of about twenty types of intelligence provides a synoptic background for intelligent behavior that can be represented by agents. The three categories of the synergy of simulation and software agents are discussed in the following three sections: agent simulation, agent-supported simulation, and agent-monitored simulation. Extensive bibliographic analysis which is based on about 440 references supports each category of the synergy of simulation and software agents. Discussion of some desirable research directions and a conclusion section terminate the article.

Keywords: Agent-directed simulation, Agent-monitored simulation, Agent simulation, Intelligence, Software agent, Supported simulation.

---

* Corresponding author Nasser Ghasem-Aghaee: Department of Computer Engineering, Sheikh Bahaei University, Baharestan, Iran; Tel/Fax: [phone]; E-mail: [e-mail]

1. INTRODUCTION

The synergy of simulation and intelligent software agents is explored here. Many possible applications of simulation are highlighted in Sec. 2 to appreciate the research patterns it offers. A review of intelligence and intelligent entities and the agents is reviewed in Sec. 3. A systematic glossary of about twenty types of intelligence provides a synoptic background for intelligent behavior that can be represented in agents. The possibilities of the synergy of simulation and agents are reviewed in Sec. 4. Three categories of the synergy of simulation and software agents of: agent simulation, agent-supported simulation, and agent-monitored simulation, are discussed in the following three sections. Each of these three

sections is supported by a bibliographic analysis of about 440 references. Discussion on some desirable research directions and drawing a conclusion constitutes the last section.

2. SIMULATION: HIGHLIGHTS

To be able to grasp the full potential of the synergy of simulation and software agents, acquiring knowledge on different types of simulation is essential. Modeling and simulation can be perceived from the following perspectives: (1) Purpose of use, (2) Problem to be solved, (3) Connectivity of operations, (4) Types of knowledge processing, and (5) Philosophy of science. The three purposes of simulation are: (1) Perform experiments, (2) Provide experience and (3) Imitation, pretense [1]. In this article, the focus is on the experimental aspect of simulation. In this context, simulation is a goal-directed experimentation with dynamic models, (i.e., models with time-dependent behavior). One way to distinguish different types of simulations is to consider whether a simulation program runs independent from the real system, two categories of simulation became possible: stand-alone simulation and embedded simulation [2].

2.1. Stand-alone Simulation

Here, the simulation program runs independent of the system of interest; almost all types of conventional simulations are stand-alone simulation. As listed in Table 1, there exist six types of applications of stand-alone simulation: decision making, training to enhance decision skills, training to enhance motor and related decision skills, training to enhance operational skills, understanding and education, and entertainment.

Table 1 Applications of stand-alone simulation.

Simulation for decision making is run for prediction, evaluation, sensitivity analysis, planning, acquisition, design, prototyping, and proof of concept. Gaming simulation is run for training to enhance decision skills. In defense applications, this type of simulation is named constructive simulation and includes war gaming, while, it is applicable for operations other than defense like, conflict management and peace assurance. In business applications, gaming simulations include business games which can be run in zero-sum environments to enhance decision making skills subject to competition or in non-zero-sum environments to enhance decision making skills subject to cooperation [3]. Simulators are often human-in-the-loop simulations where operators use virtual equipment to develop motor skills and the associated decision skills. In defense applications they are named virtual simulation. In complex systems like scientific and social systems, simulation provides the possibility to test the given hypotheses on the nature and behavior of a system and makes them easy to understand. Simulation is an enabling technology applicable in enhancing learning/teaching many topics. In entertainment, simulation is run for simulating games and for the animation of dynamic systems.

2.2. Embedded Simulation

In embedded simulation, simulation program runs together with the system of interest. The embedded simulation is of the two purposes of: enrichment and support of real system operations (Table 2).

Table 2 Embedded simulation application.

In enrichment of real system operation, the system of interest and simulation program runs concurrently. Depending on the objectives, there exist different possibilities like: simulation-based augmented/enhanced reality and on-line diagnosis. The Simulation-based augmented/enhanced reality is an accepted form of advanced training environments where real operators can operate real equipment at augmented reality mode. The embedded simulation systems are well known especially in training associated with equipment operation [4]; and are applicable in decision systems [5]. In the on-line diagnosis, the output of the simulation program and the real system is compared on-line, where, any deviation in the real system’s behavior from the simulated behavior may indicate a malfunction.

In the second type of embedded simulation application, the objective is to support the real system operation. Here, the system of interest and the simulation program operate alternately to provide predictive displays where, real system data can be applied: 1) to calibrate the simulation model and its parameters and 2) as input to the simulation system in order to evaluate the effects of decisions regarding the real system’s behavior before feeding them to the real system.

2.3. Other Perspectives

In another context, there exists the possibility to discriminate different types of simulation, based on where the computations are carried out. The computation can be run on a single computer or on different computers in a network¹. The last possibility leads to distributed simulation [6, 7] and possibly to Web-based simulation (which could rely on agent-directed simulation and can apply mobile agents in simulation.) The advantages of distributed simulation [8] (p.261) are:

- Partitioning the simulation problem between machines with high potential of interest

- Interoperability and sharing of resources

- Exploiting graphics, ability provided by special machine

Distributed-interactive simulation is run mostly in military applications, where simulators from different types of forces are connected to form a full battle situation [9, 10]. In parallel simulation, simulation program operates on multi-processor machines something true in real world where systems naturally operate in parallel. The advantages are increased speed and/or increased size of the model [8] (p.261). Federated simulation [11] is an example of interoperability of several simulation studies; each named a federate. This simulation is based on the military requirements of the Department of Defense (DoD) of the USA and joint forces of NATO (NATO MP). Current realization relies on High Level Architecture (HLA). For HLA education, refer to Morse (2000). The HLA is applied a methodology-based simulation approach [12]. For other aspects of simulation, there exist some taxonomies available on the following topics: simulation [13], simulation languages [14, 15], simulation models [16], simulation model behavior [17], simulation model processing [18, 19], simulation quality assurance [20], and artificial intelligence (AI) in simulation [21].

3. INTELLIGENCE, INTELLIGENT ENTITIES, AND AGENTS

Intelligence is an important characteristic and is being studied since the ancient times [22]. The same researches state: Working meaningfully in the field of intelligence requires a broader background than might be the case in another field. Work in this field cross-cuts cognitive psychology, biological psychology, developmental psychology, differential psychology, educational psychology, personality psychology, cultural psychology, industrial-organizational psycho-logy, and possibly other areas of psychology. To keep up with this field and advance it, one must be able to understand and to integrate the contributions from these various aspects of the field. Intelligence exists in natural systems like in humans, animals, and in engineered systems like robots, in some AI software systems, and agents. In cyberspace, they are named info-habitants [23].

Most of the authors in AI adopted Minsky’s definition: Artificial intelligence is the science of making machines to perform things that would require intelligence if done by humans [24]. However, there exist counter examples on this, for example, taking the cubic root of a number on the human part would require, knowledge (of the algorithm) and mental ability (intelligence); while, done by a calculator, this ability is not sufficient to allow the calculator be labeled as intelligent. Some taxonomies of intelligence are given by Schmutter [25] and Sternberg [22]. In an early study where a classification of about 500 types of knowledge and knowledge processing knowledge was presented, Ören [26] wrote:

The advances in studies of the brain and cognitive sciences are very important and would have implications on artificial intelligence. However, it would be very useful to demystify intelligence and to identify its components in terms of knowledge for knowledge representation and knowledge processing (or knowledge-processing knowledge) in order to embed them in different machines having knowledge-processing abilities.

It may take long while, if not ever, before the use of the term ‘intelligence’ would mean ‘artificial intelligence’, ‘machine intelligence’, ‘computational intelligence’, or ‘synthetic intelligence’. Currently, when we use the term ‘satellite’ we no longer think of the moon, the natural satellite of the earth, whereas, in the 1960s, one had to use the term ‘artificial satellite’ to denote any man-made satellite. The objective of building man-made satellites was obviously not to replace the natural one(s) but to create new modalities of them. Similarly, the objective of AI is not to replace natural intelligence but to create new modalities of intelligence or advanced knowledge processing. Davis and Hersh who quoted from Good state that Good (1964, 1965) observes first of all that there is no point in building a machine with the intelligence of a man since it is easier to construct human brains by the usual method. One should build an ultra-intelligent machine which may be defined as a machine that can far surpass all the intellectual activities of any man however clever [27].

Once the elements of advanced knowledge representation and knowledge-processing knowledge are well defined, they can be expressed in terms of different programming paradigms such as distributed, parallel, real-time, procedural, and functional programming; and data-flow, object, rule-oriented, and agent-based paradigms.

In the Handbook of Human Intelligence, an all-embracing definition of intelligence is offered as follows: we shall try to define intelligence, as have others before us, as a goal-directed adaptive behavior. At the end of the chapter, we will argue: that this definition does indeed fit the body of ideas is agreed upon [26, 28].

3.1. Types of Intelligence

The objective here is to realize the engineered systems with advanced cognitive knowledge processing abilities. As a generic definition, the following definition is accepted: intelligence (human, animal, or machine) is an adaptive and goal-directed knowledge processing ability [29]. Intelligence has several aspects and types worthy of being discriminated. Taxonomy of intelligence types is illustrated in Fig. (1), where rectangles with continuous lines are the types of intelligence and rectangles with dashed lines being the criteria to distinguish the types of intelligence. In the later sections, a reference to intelligence may point out to any one of these types of intelligence. Intelligence is enhanceable. Any type of intelligence can best be represented in a continuum as the level or the value of the intelligence (wherever metrics and measurement processes are defined), rather than by a binary choice.

With respect to knowledge processing abilities, there exist two types of machines or systems: Systems for knowledge processing and systems with additional knowledge processing abilities. Knowledge processing machines are the computers. For a historic view of knowledge processing machines other than computers, refer to [26]. Machines or systems with additional knowledge processing abilities have knowledge processing abilities to satisfy their main purpose of existence better [26]. Such machines or systems can, perform optimization (like a tracking missile or a vehicle-sensing road). In engineering applications, systems or machines with knowledge processing abilities to support/optimize their functionality are named the smart systems (such as smart bridges); however, intelligence of these types of applications is rather low. In machine intelligence, some of the metrics applicable to humans, such as intelligence quotient (IQ) would be meaningless, since machine intelligence is independent of the machines’ age.

Fig. (1))

Types of intelligence.

A systematic glossary² of the types of intelligence is presented in Table 3. The taxonomy refers to three aspects of intelligence: the entities where intelligence is applied, the context within which intelligence applied and the components (mechanisms – structures and processes) of intelligence.

Table 3 Types of intelligence (a systematic glossary).

The objective of studying intelligence in this chapter is to distinguish its types in a sense that they would be useful in representing software agents; hence, some modifications to the definitions are made. To give full credit to the authors from whom the definitions are borrowed, their names are expressed and to clarify that they are not the original definitions, modified definitions are indicated.

3.1.1. Entities

The entities to which intelligence is applied are of three groups: abstract entities, concrete objects, and people. Abstract intelligence is intelligence applied to abstract entities such as ideas, and has the three main sub-types of: linguistic intelligence, logical-mathematical intelligence, and musical intelligence. Concrete intelligence, (named mechanistic intelligence as well) applies to concrete objects and has the three sub-types of: visual-spatial intelligence, naturalistic intelligence, and contextual intelligence.

Emotional intelligence is applied to self (intrapersonal) or to others (interpersonal). Emotional intelligence has an aspect which concerns only others, (i.e., empathy). Empathy is sensitivity to others’ feelings and concerns and taking their perspective; appreciating the differences in how people feel about things [37]. Three other aspects of emotional intelligence are: self-awareness, managing emotions, and motivation are applicable to one's self and others as well. Self-awareness of emotions is observing oneself and recognizing feelings as they happen. Emotional awareness is observing others to recognize feelings as they happen. Managing emotions is handling feelings so that they are appropriate and realizing what is behind a feeling, and finding manners to handle fears and anxieties, anger, and sadness [37].

Motivation is channeling emotions in the service of an objective; emotional control; delaying gratification and stifling impulses. Hence, emotional intelligence is the ability to monitor one’s own and others’ emotions, to discriminate among them, and to use the information to guide one’s thinking and actions. Social intelligence is emotional intelligence applied to others.

The relations between the emotional intelligence with intrapersonal and interpersonal (social) intelligence are expressed in Fig. (2). In the literature emotional intelligence is defined as to be a type of social intelligence, while, here the view is explicitly different, (Fig. 2).

Fig. (2))

Relation of emotional intelligence with intrapersonal and interpersonal (social) intelligences.

There exists a strong relation between intelligence (especially social intelligence) and personality traits and facets [38]. The basis for fuzzy agents with personality is elaborated in [39, 40].

3.1.2. Context

The context within which intelligence operates can be unfamiliar or familiar. Experiential intelligence is the ability to deal with new tasks or situations and the ability to use mental processes in an automated manner. Fluid intelligence is an intelligence type operating in unfamiliar situations. The intelligence is crystallized intelligence with practical intelligence as a subtype.

3.3. Components

The components of an intelligent knowledge processing entity consist of: performance, monitoring and control, and knowledge acquisition and improvement components. Performance components realize several types of cognitive knowledge processing such as inference, reasoning, anticipation (pro-activeness), decision making, and assessing [26]. Monitoring and control components monitor self and others during knowledge processing (introspection for self, perception for others) or after knowledge processing activities (postmortem analysis). Knowledge acquisition and improvement components are for perception, understanding [41] and learning. The last type leads to learnable intelligence, equally important for natural intelligence and for machine intelligence and software agents’ intelligence. Identification of the components or mechanisms –structures and processes of intelligence– leads to componential intelligence [35]. The two types of componential intelligence are the neural intelligence and reflexive intelligence.

3.4. Agents

Agents are autonomous software modules with perception and social ability to perform goal-directed knowledge processing, over time, on behalf of humans or other agents in software and physical environments. The last part of the definition begins with on behalf of is a truism and if omitted would be implied.

The core knowledge processing abilities of agents include: reasoning, motivation, planning, and decision making. Additional abilities of agents are needed to increase their intelligence and trustworthiness. Abilities to make agents intelligent include anticipation (pro-activeness), understanding, learning, and communication in natural and body language. Abilities to make agents trustworthy and assuring the sustainability of agent societies include being rational, responsible, and accountable, which lead to rationality, skillfulness and morality (e.g., ethical agent, moral agent).

Software agents may have most of the aspects of intelligence as outlined in Table 3. The type and level of needed intelligence depend on the knowledge processing requirement of the task. In the next section, regarding synergies of simulation and agents, it will become apparent how drastically does the intelligence represented by software agents contribute to simulation. There exists a strong relation among intelligence (especially emotional intelligence), personality traits and facets [38]. The basis for fuzzy agents with personality is elaborated in [39, 40].

3.5. Software for Agents

Java Agent DEvelopment framework (JADE) is implemented in Java language and used in developing agent applications [42, 43]. JADE is probably the most widespread agent-oriented middleware applied today. JADE is a completely distributed middleware system with a flexible infrastructure allowing easy extension with add-on modules. JADE offers a rich set of programming abstractions allowing developers to construct JADE multi-agent systems with relatively minimal expertise in agent theory. The JADEX agent framework is presented by [44], which supports reasoning by exploiting the BDI model and is considered as an extension to the widely applied JADE middleware platform.

NetLogo [45, 46] is a multi-agent programming language and modeling environment for simulating complex natural and social phenomena.

The Recursive Porous Agent Simulation Toolkit (Repast) is one of the several agent modeling available toolkits. Repast borrows many concepts from the Swarm agent-based modeling toolkit [47]. The three implementation of the Repast agent modeling toolkit is presented by North, et al. [48].

The evaluation of free Java-libraries for social scientific agent-based simulation is presented by Tobias and Hofmann [49]. A survey of various agent-based modeling platforms is presented by Nikolai and Madey [50]. Railsback et al. apply five software platforms in agent based simulation [51].

4. SYNERGIES OF SIMULATION AND AGENTS

Agent-directed Simulation refers to the synergy of software agents and simulation. As observed in Fig. (3), there are three possibilities that can be considered under two groups: 1) simulation for agents: which consists of agent simulation and, 2) agents for simulation which consists of agent-supported simulation and agent-monitored simulation [11]:

Fig. (3))

Types of agent-directed simulations.

Agent simulation is simulation of agent systems. It is also named agent-based simulation.

Agent-supported simulation is the use of agents for at least one of the following purposes:

To provide computer assistance for front-end interface functions in a computer-aided simulation study;

To provide computer assistance for back-end interface functions in a computer-aided simulation study;

To process elements of a simulation study symbolically for example, consistency checks; and

To provide cognitive abilities to the elements of a simulation study like learning, understanding and/or hypothesis formulation.

Agent-monitored simulation is the use of agents for generation and/or monitoring of agent behavior. (This is similar to the use of AI techniques –like qualitative simulation– in generating model behavior).

The simulation synergy symmetry and artificial intelligence in general [21, 29] and the simulation synergy and agents are tabulated in Table 4.

5. AGENT SIMULATION

Agents provide a natural pattern to represent intelligent entities. Agent simulation is involved in natural or engineered entities with cognitive abilities represented by agents. Here the three constituent parts of this simulation, the: applications, methodology, and software for agent simulation are presented.

Table 4 Types of simulation, based on the synergy of simulation with ai and with agents.

5.1. Applications

A synopsis of the application areas of agent simulation is presented in Tables 5a-e. Almost all references are dated after year 2000. The engineering applications of agent simulation used in electrical engineering, irrigation systems, manufacturing systems, mechatronics, network, robotics, software, and transportation / logistics are tabulated in bold in Table 5a. The economy and management applications of agent simulation, references on economy, e-commerce, and management are presented in Table 5b. The agent simulation of social systems and human behavior applications and references that cover social systems, psychology/human behavior, and physiology are presented in Table 5c. The agent simulation of environmental issues is presented in Table 5d. Table 5e is on military applications.

Table 5a Agent simulation – engineering applications.

Table 5b Agent Simulation Management/Economy Applications.

Table 5c Agent simulation social systems and human behavior applications.

Table 5d Agent simulation - environmental applications.

Table 5e Agent simulation – military applications.

Table 6a Agent simulation methodology in general.

5.2. Methodology

The list of references a synopsis of the methodology of agent simulation is presented in Tables 6a-6c. The references on agent simulation methodology, in general, are tabulated in Table 6a.

In some of the early references especially the ones by Jávor the adopted concepts are similar to that of the agents but are referred to accepted terminology of demons or daemons. The special aspects of agent simulation methodology like architecture of agent simulation systems, mobile agents; adaptive, parallel, and Web-based simulation; and chaotic, complex, and distributed systems are tabulated in Table 6b.

Table 6b Agent simulation methodology special aspects.

The methodology of emerging holonic agent simulation –for the simulation of cooperative behavior is tabulated in Table 6c.

Table 6c Agent simulation methodology: cooperation & holonic agents.

5.3. Software for Agent Simulation

The software for agent simulation is introduced in Table 7, which are clustered into three categories of: language, software and environment. In each category some recent software are presented.

6. AGENT-SUPPORTED SIMULATION

A simulation environment can have two types of, front-end and back-end interfaces. The Front- end interfaces are applied to specify, edit, or generate elements of a simulation problem; the Back-end interfaces are applied in systems to communicate with the users, the primary and auxiliary outputs of the system.

Table 7 Software for agent simulation.

Agent-supported simulation is applied in agent technology to support simulation activities in modeling and simulation environments and simulation-based problem solving environments. This support is comprised of: front-end and back-end interface activities, agent-supported symbolic processing like agent-supported validation and verification, and agent-supported cognitive abilities with respect to the elements of simulation systems like agent-supported program generation, program integration (as it would be the case in the formation of federations using HLA) and program understanding for documentation and/or maintenance purposes. For each one of the categories of knowledge and knowledge processing knowledge, an important category of knowledge processing activity is required to ensure the reliability of the associated elements. The possibilities for agent support in modeling and simulation environments are outlined in Table 8, for the following elements of simulation studies: objective of the study, parametric models, model parameters, design of experiments, experimental conditions, simulation program, and the behavior and recommendations. For each one of the elements, this support can contribute to its generation, specification/editing, and processing.

Table 8 Possibilities for agent support in modeling and simulation environments.