Intelligent Systems

Over the past 50 years, numerous disciplines related to intelligence have achieved results that may ultimately prove to be very powerful and beneficial to the average citizen as well as to the specialist in the area of Intelligent Control. The information and cognitive sciences are making fundamental breakthroughs that will radically alter our vision of the architecture of existing systems and design and control of the systems we are dealing with and/or create.

These breakthroughs can be summarized as follows:

A. State of the Art

1. The power and speed (per unit cost) of electronic computing has risen exponentially by more than a factor of ten per decade for over four decades.

2. The store of knowledge about how the biological brain works has also increased exponentially.

 3. Studies of complexity have emerged which demonstrated that intelligent systems have common ways of dealing with problems of complexity.

 4. A solid theoretical and mathematical foundation is being established for the scientific study of:

a) language and speech understanding

 

• strong links are found between the natural and computer languages

• inner laws are found that connect natural languages with the metasystems in which these languages have been generated

• natural languages and all symbolic systems follow a common set of laws of symbols generation and manipulation

• all symbolic systems are extensions of the morphology and evolutionary processes in the definite domains of reality

b) image understanding and perception

 

• images can be understood as an interface between reality and perception

• structures of images should be attributed both to reality and to the brain organization

• the area of image understanding (including recognition) is moving from template-based

 ad-hoc solutions toward incorporation of multi-faceted, multiresolutional methodologies

c) knowledge representation, including concept generation and meaning extraction

 

• knowledge can be understood as an interface between reality and cognition

• structures of knowledge can be attributed both to reality and to the brain organization

• processes of temporal development of knowledge allows contribute to discovery of meaning and interpretation

d) perceptual images and knowledge are organized in a multiresolutional (multigranular, multiscale) way as the model of brain and nervous system

• multiresolutional (multigranular, multiscale) representation has proven to be a way of dealing with complexity

• grouping, focusing attention and search are recognized as mechanisms of functioning of the architectures of representation and processing

e) simulation and imagination, including mechanisms of invention and insight

 

• reflection is now regarded as a component of processes of consciousness

• combinatorial development of conceptual and perceptual assemblies are related to invention and insight

f) reasoning and decision making are better understood than was provtded by introspective, behaviorist, and early cognitivist theories

• it became clear that both reasoning and decision making are linked with the phenomenon of meaning

• consciousness was recognized as a legitimate and important factor

g) planning and control

 

• the simulation nature of the planning processes is analyzed linking them both with computational search and psychological imagination

• a similarity has been discovered between the behavior generation of living systems and formation of control sequences in the "constructed" systems

5. Engineering approaches have begun to emerge for designing, constructing, testing, and exploiting practical systems with capabilities of:

 

a) intelligent production planning, resource management, and task scheduling,

b) intelligent decision making, task decomposition, goal pursuit, and reaction to unanticipated situations,

c) intelligent path planning for automated route selection, navigation, and obstacle avoidance,

 d) intelligent control for precision motion, speed, position, and force actuation,

 e) intelligent information gathering, processing, perception, sensor fusion, and situation understanding

6. Biopsychology and mathematics of intelligent systems turned out to be inseparably related to each other having their joint source in the intrinsic properties of natural intelligent systems to self-organize, self-reproduce, and self-describe (self-represent). We can hypothesize that these three "self-" capabilities can be consistently explained within a non-contradictory scientific theory.

 

a) self-organization can be considered a process of reducing the cost of functioning via development of multiresolutional architecture of representation and decision making;

b) self-reproduction can be understood as a tool of reducing the cost of subsistence as a part of temporal functioning;

c) self-description (or self-representation) can be visualized as the most efficient tool of supporting the processes of self-organization and self-reproduction by learning from experience.

Learning is an essential part of all these three self-capabilities and it requires development of a symbolic system which is easy to maintain and use. Development of the symbolic system is a part of semiotics .

 

 It is becoming possible to build intelligent control systems that can both gather and process information, as well as generate and control behavior in real time, so as to cope with situations that evolve amid the complexities of natural and man-made competition in the real world. More than 10 years of the recent activities in the area of Intelligent Systems demonstrate that a number of new directions are being successfully pursued, leading us toward development of systems emulating human intellect and its applications in a variety of areas. These directions include:

-neural networks
-fuzzy systems
-genetic algorithms
-machine learning
-evolutionary programming, etc.

 We would argue that any Intelligent System should be based upon a multiresolutional hierarchy of the "loops of functioning". Each of these loops can be treated as a control system per se. Structures of the sensory processing, knowledge representation and decision making are built in a multiresolutional way in which many contemporary pattern recognition and control methodologies are inscribed and equipped by the design.

 Neural networks and fuzzy controllers have become classical design tools, and standard component in the control engineer's standard tool-box. However, there are a number of other directions which are becoming clear and whose development we can expect in the near future. These direc tions include semiotic control, control structures for open systems, controllers with discovery of meaning, and possibly, value driven controllers will be in the focus of further development. Important issues remain of evaluating the degree of intelligence, of groups of intelligent systems with different capabilities, and their possible comparison. A number of perspectives will be explored in research and application reaching some puzzling domains such as a possibility not only of autonomous self-learning but even control systems with a degree of consciousness.

 Broad domains of application for Semiotics in the area of Intelligent Systems are analysis of large knowledge bases aimed toward supporting CAD-CAM systems, and analysis of large information flows for large systems when the problems and conflict situations should be discovered.

Semiotics

Subject of Semiotics

 Mathematical tools of semioticsinclude those used in control sciences, pattern recognition, neural networks, artificial intelligence, cybernetics. However, after unifying them within semiotics they demonstrate a consistency in being focused toward the problems of intelligent systems. Semiotics specific mathematical tools e.g. for combining signs, symbols, and extracting the meaning (performimg the semiosis) are just in the process of development. To the extent that the mathematical tools of understanding "meaning" are still being created, semiotics is in the process of emerging.

 Semiotics is a powerful theoretical tool in the area of intelligent systems especially when the large complex systems are concerned, when the multiple intelligent agents are involved, and/or when a single intelligent agent should be analyzed and/or controlled in-depth. Large and complex systems are always linked with the problem of intelligence: either because the issue of survival is critical, or because dealing with the large volumes of information cannot be done without using techniques of knowledge processing which are associated with intelligence, or both. In all cases the issue of complexity is a key one. The conventional analysis, design, control and simulation methodologies were always insufficient for analysis and design of large complex systems because the conventional methodologies do not deal with the phenomenon of complexity.

 Development of the Semiotic Modeling and Situation Analysis area (SSA), is motivated by a strong desire to make the analysis and design of Large Complex Systems, or Intelligent Systems, in general, better organized methodologically, more consistent and formally balanced. One of the features of this new methodology is extraction of knowledge from the descriptive information by its consistent analysis based upon well established algorithms. This should give an opportunity to make the descriptive information a part of the analysis of dynamic processes of control systems theory. It also requires development of new methods of dealing with large (often, multiresolutional) symbolic systems, and use of "symbol grounding" processes. All of this can be considered now a part of Semiotics.

 Several efforts to accomplish this task are known. They were pioneered by W. Haken in Germany, I. Prigogine in France, researchers from CNLS in Los Alamos National Laboratory and in Santa Fe Institute in US. In all of these effort, the opportunities of a linguistic analysis have not been explored. A. Nerode (Cornell) is moving closer to SSA in his Hybrid Control Systems. D. Pospelov and his team from Russia, made Semiotics a basis for development a variety of formal methods presently known as SSA, or Applied Semiotics.

 New directions in the area of Intelligent Systems are shaping up: semiotic modeling with situation analysis in the large complex systems (SSA), analysis of large information flows with meaning (or conflicts) discovery. These directions are in the process of emerging; we even have not yet formed an attitude toward them, and one can hear many different opinions about these directions. Some of the opinions are extremely cheerful: "this is a new approach (semiotic modeling) as applied to complex incompletely determined states of systems (situations) using linguistic analysis, non-standard logical methods, cognitive recognition, etc., etc..." while many are extremely critical: "this is a mere repetition of what we have done in the West a long time ago including knowledge bases, expert systems, linguistic structures, multi-valued logic, etc, etc..."

 Surprisingly enough, these opinions are more complementary than contradictory. It is true that each single component of the SSA, or similar approaches do not contain any revelation, and only a few innovations can be mentioned. However the synthesis of all these components within a single generalized framework, have not yet been done, and this seems to be the main contribution of the semiotics oriented researchers.

History of Semiotics

This predicament of self-referencing is characteristic of semiotics. It can be illustrated by its history following the explicit discussions that started many years ago. Ampere's theory of governing contained strong semiotic overtones, and the attention was clear to the effect that is brought by the difference among the terms we assign to the control/controlled system, and the terms that emerge within this system. Theory of Control was introduced in XIX century by J. C. Maxwell in his work "On Governors" (1868) where the fundamentals of Systems Dynamics were introduced. Semiotics was introduced in XIX century by Charles S. Peirce in his work "Logic as Semiotics: The Theory of Signs" (1893-1897) where the fundamentals of the Theory of Representation were proposed. There is no references to each other in these works although they are definitely connected by the language of a common scientific paradigm which is about to be ready for the synthesis.

As far as analysis of system's behavior is concerned, the fundamentals of the qualitative theory of representation for the behavior of dynamic systems were published by H. Poincare in 1882- 1885. Most of the forms discovered by Poincare were "rediscovered" in Nature by D'Arcy Thompson who published his "On Growth and Form" in 1917; the book can be called "the semiotics of nature". Interestingly enough, R. Thom make these forms "the 18 archetypes of all processes", gives them a mathematical status, and they end up virtually to be a part of semiotics.

Semiotics as a sub-discipline of linguistics, was already blossoming in the period of 1920-30; it was promulgated by F. De Saussure and the Geneva school; linkages with biology (communication of ants and bees) have been shown by K. Buhler in 1929. Exactly at this period of time a) the gestalt theory was developed and the powerful tools of generalization precipitated theoretically, while b) the contemporary control theory was formulated and developed both by methods of the theory of differential equations as well as the automata theory-- thus, the tools of generalization are becoming available also to the engineers.

 W. McCulloch and W. Pitts proposed the logical calculus for nervous activity in 1943. E. Scrodinger presented his physicist's analysis of "What Is Life" in 1944. N. Wiener published his "Cybernetics" in 1947, and A. M. Turing makes available his "Computing Machinery and Intelligence" in 1950. At this time, all basic knowledge for the subsequent synthesis has been prepared.

 The synthesis starts with tentative grouping. First, the structuralists emerge as a joint group with their social, cultural, psychological, and linguistic analyses (1950-70). No mathematics and control are involved explicitly yet. A subset of structuralists enhanced by researchers from different disciplines interested in modeling, constitute a direction of "Classification". G. Miller with his cross-cultural concepts, N. Geschwind with his language/brain analyses, and others demonstrate the validity of a blend of such components as linguistics, biology, mathematics, coding, and cognition. Then, in 1960-1970 a series of Symposia on Self-organization generate an interlaced body of Structuralism, Linguistics, and Classification Theory with Strong emphasis in Control Science (See, for example, "Principles of Self-Organization", Transactions of the Symposium on Self-Organization, U. of Illinois, June 8-9, 1961, Eds. H. von Foerster, G. W. Zopf, Pergamon Press, Oxford, 1962; among the participants: S. Amarel, W. Ashby, L. von Bertalanffy, W. McCulloch, A. Novikoff, G. Pask, A. Rapoport, F. Rosenblatt and others).

 Cybernetics was looking for its role in interpretation of thinking processes. In the Vol. 17 of the series "Progress in Brain Research" (Eds. N. Wiener and J. Schade, Cybernetics of the Nervous System, Elsevier Publ., Amsterdam, 1965) the first paper by M. Maron "Cybernetics, Information Processing, Thinking" raises a problem of representation which becomes dominating in the subse quent 20 years; V. Braitenberg introduces the outline of his famous "Vehicles" (in "Taxis, Kinesis and Decussation"), and J. Holland prepares the space for the future powerful theoretical moves in the paper "Universal Embedding Spaces for Automata".

 One can see in this effort with developing of the Semiotic Situational Analysis (SSA), something very fundamental, potentially more successful than other efforts in progress (such as the effort with "Chaos and Complexity" which is probably lacking the cross-cultural and linguistic compo nent, also it is not sufficiently modeling, design, and control oriented, etc.). This is why the semi otic concepts are so productive in such remote domains as the phenomena of emergence in living organisms, and structures of brain (K. Pribram, et al).

 D. Pospelov, the creator of SSA is definitely a global thinker, he is well prepared in a multiplicity of sciences- components. Unlike many prominent scientists who have specialized solely in their own domain, D. Pospelov is a broad-minded multidisciplinary scholar who has demonstrated bold and aggressive thought in constructing concepts and making associations. US scientists have seri ous and sometimes better results in each of the components of SSA. However, US never ventured to develop a scientific theoretical synthesis on a such a global scale.

 During the period of last 20 years, the architectures of large systems are being developed - natu rally related to many intelligent systems-and these architectures are developed through proce dures which are strongly related to semiotic tenets and formal methodologies (see references related to the NIST-RCS system).

 The partial joint efforts have already been initiated and partially explored. All preparatory works are already completed-for the upcoming phenomenon of Synthesis.

Topics of Semiotics

Theory of Semiotic Modeling
Theories of Cognition
Architectures of Brain
Intelligent Agents as Semiotic Entities
Extraction of Meaning from the Large Information Flows
Sensor Fusion
Representing Knowledge in Intelligent Systems
Symbol Grounding
Methods of Knowledge Aggregation and Disaggregation
Multiresolutional Representations
Multiresolutional Symbol Grounding
Integrating Knowledge from
Various Sources
Semiotics and Emergence in Living Systems
Semiotic Analysis of Intelligent Agents
Testing Semiotic Models for Adequacy
Paradoxes: Their Use in Knowledge Representation and Decision Making
Links Between Semiotics and Complexity Theory
Translation of Natural Language Texts into
Knowledge Representation
Knowledge from Databases
Meaning Extraction
Theories and Techniques of Reasoning
Alternative Logical Systems
Hybrid Multiresolutional Control Systems
Theory of Situation Analysis
Forming Decisions in Semiotic Models
Languages of Situation Analysis and Control
Phenomena of Discovery in Situation Analysis
Learning in Situation Analysis
Practical Situation Analysis
Integration of Semiotic and Cybernetic Models in Control Systems
Reflection and its models in Situation Analysis
Soft Computing in Situation Modeling
Visualization of Fuzzy Logic Representations
Semiotic Modeling Software
Large Semiotic Systems

Applied Semiotics

These workshops, together with published books and papers, suggest that the Russian school of semiotics has achieved powerful results during the period of last 20-30 years, that begin within the area of intelligent decision-making in large complex systems with incomplete knowledge, and extend far beyond this into many areas of natural and humanitarian sciences. The principal achievements of this group are in the use of Semiotic principles to develop formal schema for representation of knowledge about the world, and to successfully extend the mathematics of logic and reasoning so as to integrate many informational and cognitive disciplines into a unified whole.

This group has extended the relevant fields of mathematical logic, knowledge representation, and decision making in ways that may have a profound impact on a number of important applications, including military command and control, agile manufacturing, intelligent control of power plants and chemical processing plants, etc. There may also be economically important applications in construction automation, transportation safety and efficiency, health care, toxic and radioactive waste handling, environmental preservation and restoration, and undersea and planetary exploration and exploitation.

 Several groups in US and Europe can be considered as having similar orientation to dealing with Intelligent Systems. This conference is determined to better understand the emerging domain which can be called Applied Semiotics.

Semiotics and Intelligent Systems

1. The Problem of World Representation:

How is the world represented internal to the system?

2. The Problem of Models Verification:

How is an internal model of the world verified?

3. The Problem of Symbols Generation and Interpretation:

How is symbolic representation generated and interpreted?

4. The Problem of Duality Semiotics-Mathematics:

How are symbols and rules related to numbers and computations?

5. The Problem of the Elementary Unit of Intelligence:

What is intelligence and what are its fundamental constituencies?

6. The Problem of Symbol Manipulation:

What is the calculus of symbols

7. The Problem of Interpretation: the Discovery of Meaning:

How do we get from symbols to meaning and vice versa?

8. The Problem of Semiosis Loop:

How do we close the loop from sension through interpretation and meaning, to acting?

9. The Problem of Granularity:

How do we deal with the immensity of spatial and temporal details in the enormously large world?

10. The Problem of Cognitive Structuring:

How do organize the design of Intelligent Systems

11. Why Semiotics?

Areas of Application

• Manufacturing process control (design, engineering, process planning, production scheduling, machining, joining, finishing, assemblying, inspection, packaging, shipping, etc.)

The existing approaches lack the degree of integration they require. Local interests often overcome the large scale advantages (often the latter are even not brought to the attention of the management levels capable of evaluation the large scale advantages). Semiotic methods of situational analysis and design can drastically change the situation in this area

• Health care management(patient monitoring, laboratory testing, drug manufacturing, management of the facilities, etc.)

The efficiency of the single unit of the health care can be substantially increased by applying semiotic methods of analysis and design. The situations which include "the set of medical units vs the set of insurance units" are never analyzed seriously at present because the analysts are not equipped by the semiotic formal tools of analysis. No wonder that health care reform efforts has become a subject of political debate rather than a subject of the scientific analysis: the tools for this analysis are available to the interested parties

• Transportation Efficiency and Safety Management

(intelligent highways, vehicles functioning, air traffic control, cockpit safety, ship navigation, rail safety, etc)

The problems of this domain need a solution in which the integrated multidisciplinary approach could eliminate the disarray and confusion which entail voluntary decisions, lack of accountability and eventually, less acceptable solutions

• Power plant control (providing for efficiency, reliability, safety)

Many think that it is impossible to analyze the energy system within a unified integrated system of analysis and design. This opinion is presently dominant primarily because of the fact that most of the results in Semiotics are virtually unknown. Thus, the network dispatch is being performed in one conceptual paradigm, the energy unit control in another, and the subsystems control in the third. The customer suffers as well as the Power Systems which are foregoing substantial increase in their revenues and effectiveness

• Battlefield management(information integration, decision support, logistics, real time planning, repid response, etc.)

This area is so permeated by the diversified processes of information processing that would obtain substantial benefit from implementing Semiotic methods and tools. The multiple and deeply interrelated information flows start with linguistic and sensory inputs, require multiresolutional transformation of sensory signals and images to semantic networks, using the latter for decision making and then transformation of information in opposite direction. Solving the problems of digital battlefield by using Applied Semiotics can substantially improve the reliability and effectiveness of many processes of battlefield decision making

• Telecommunications(efficient encoding/decoding, transmitting not only a framework, but the meaning of the message in a context)

It can benefit of being treated as a part of an overall system of information processing. A multiplicity of communication problems are intrinsically semiotic ones by their nature

• Autonomous Vehicles(navigation, obstacle avoidance, landmark recognition, maneuvering in traffic, object tracking and classification, etc.)

In all of their control loops, functioning initiates with linguistic and sensory inputs, requires multiresolutional transformation of sensory signals including visual images to semantic networks. In all cases, the initial source of information are images, maps and descriptions the natural language; the system must ultimately become compatible with human beings as well as with the rest of the digital battlefield system.

• Decision making in complex systems with uncertaintydealing with incomlete, incorrect, contradictory, misleading, noisy information)..

Semiotics uses a variety of tools of reasoning and arguing, recommends new techniques of inference, e.g. falsifiers generation, etc.

Glossary

abstraction - means focusing upon some particular feature and/or property of an object, while generalization presumes unifying a set of features and/or property, object into one property, object (generalized property, object).

 agent- an operational unit for which the intelligence of the system plans and executes control sequences (possibly in coordination with other agents). Agent is expected to exist at the output of behavior generator in the form of an Actuator or a Virtual Actuator. The concept of Agent is very vague: we can call a subsystem of a system an agent, we can call a living creature an agent.

 Applied Semiotics- a new area of semiotics which focuses upon application of semiotic principles and theoretical results in a variety of scientific disciplines, arts, literature, engineering, and many other domains of human activity.

 architecture -the assignment of functions to subsystems and the specification of the interfaces between subsystems.

 behavior - is the ordered set of consecutive/concurrent changes among the states (in a simple case, "the string of changes between the consecutive states") registered at the output of the system (subsystem); which it a unified property ("regular" behavior) if a "law" of the string formation is found; if the law is not found we can call it "stochastic" or "random" behavior. Therefore, any output of the system observed during some interval of time can be considered "behavior" of this system. This means also that behavior of the system can be described as a time-tagged trajectory (motion) in the state space.

 behavior generation - is the planning and control of action designed to achieve behavioral goals.

calculus of symbol manipulation- all formal, mathematically consistent methods of logical analysis.

 classification - every observation or measurement of an object from a particular class is represented as a vector in a multidimensional space, called classification or decision space; the three main mathematical concepts of partitioning classification space among classes are: discrimination surfaces, nearest neighbors, and parametric statistical distributions.

clustering - a kind of grouping activities which put together entities based upon their similarity (closeness, resemblance, adjacency, etc); the measure of similarity presumes an existence of some inner substance of this similarity, some unity in the cluster which will allow for interpretation.

complexity - (frequently, because of the "curse of dimensionality") a factor limiting intelligent knowledge processing, including pattern recognition techniques, which is often due to the fact that general mathematical methods of the design of few efficient classification features has not been developed; utilizing many features results in a problem known as "the curse of dimensionality": learning in high-dimensional classification spaces often requires exponentially large number of training samples.

 consciousness - in J. Locke's definition "the perception of what passes in man's own mind".

decision making - the process of information processing which includes generation of alternatives and the subsequent choice of the preferable one.

 focusing attention - creation of a subset of a representation; this subset will be used subsequently as a limited set of information for the further processing; can be considered a particular case of grouping since the measure of similarity in this case is the degree of potential interest for the user; a very soft grouping.

 functional loop - a closed loop of behavior generation which runs through the following subsystems: sensors, sensory processing, knowledge storage, behavior generator, actuators, world.

 generalization- a procedure (or a set of procedures) of generating a new object (entity) from a multiplicity of its parts and attributes; the set of procedures most probably includes grouping, focusing attention, combinatorial search (GFACS). There are many methods of generalization including generalization via approximation, via averaging, via integration, via aggregation and labeling based on recognition and detection. The goal of both generalization and abstraction is to increase the efficiency of knowledge manipulation.

 grouping - creation of a subset of representation based upon their similarity (closeness, resemblance, adjacency, etc); an example of grouping is clustering; concatenation with subsequent formation of strings, formation of blobs, etc.

intelligence - This is what WWW Britannica Sampler is saying: "intelligence is a mental quality that consists of the abilities to learn from experience, adapt to new situations, understand and handle abstract concepts, and use knowledge to manipulate one's environment".

 indistinguishability zone - see tessellatum.

 intelligent control - a controller which laws provides for an appropriate way of achieving the goal when disturbances are introduced (or otherwise emerge). "Appropriately" means that the actions of the controller satisfy some measure of rationality (which minimizes costs, maximizes reliability, achieves trade-off in a situation with multiple costs, etc.)

 intelligent systems - a system with the ability to act appropriately in an uncertain environment. For interpretation of the vague term "appropriately" see intelligent control.

 interpretation - a process of explanation for the encoded set of information, when this process is performed under a particular goal, the result of interpretation discovers the meaning of the encoded situation.

introduction of symbolic systems - designing the system of signs appropriate for encoding the objects and the phenomena of the world.

inverted symbol grounding - finding the signs (within the existing system of signs) for the phenomena of the world.

language - a system of three sets including the vocabulary, the grammar, and the system of axioms. Vocabulary presumes existence of a set of signs and a set of reality (world) in which the signs of vocabulary and the phenomena of the world could be put in correspondence.

 learning - is the process of finding the regions occupied by classes or the boundary between classes in classification space; learning is based upon search, focusing attention and grouping applied consecutively as a part of the process of generalization; learning is based on a set of training data that are examples of object-vectors from known classes, otherwise the classes should be discovered by the process of meaning explication.

 levels of granularity - (are referred also as levels of a hierarchy, levels of resolution, levels of generalization, levels of abstraction)- is a representation of the system with a particular level of detail. Level of resolution, and level of granularity have the same meaning because both resolution and granularity refer to the same idea of indistinguishability zone. Level of generalization presumes that different resolution (granularity) of levels are obtained as a result of the properly performed generalization. The expression "level of abstraction" is often used instead of "level of generalization" although their meaning is not equivalent. (see Abstraction).

loop of symbol processing - it is a representation of a functional loop (see functional loop).

math models+semantics - the basis for interpreting symbolic (sign) representations.

math model - a set of differential equations (for a continuous system) or set of rules (for a discrete system) that predict what output will result from a given input.

math , mathematics - is a discipline which studies objects and systems under condition of abstraction (see). Formally, it always ascends to dealing with symbolic representation of objects. However, these are not necessarily objects of the world.

 meaning - is interpretation of a subset of reality (i.e. in its symbolic representation) which can be used in generating successful behaviors;

 meaning discovery (explication) - can be considered a pattern recognition in a non-classical sense; it refers to a process in which the set of classification features (patterns) should be found that characterize the objects and/or activities observed; the patterns (meaning) discovered will determine the space for the subsequent classification (interpretation); the subsequent descriptions will be partitioned among newly discovered classes.

model construction - is the construction, maintenance, and utilization of internal representations of the world or any subset of it.

 multiresolutional systems - systems of knowledge representation, symbol processing, image recognition, etc. which are characterized by dealing with a set of representation encoded at different scopes with different resolution; it is beneficial to narrow the scope and increase the resolution from level to level top-down; e.g. a combined need for image segmentation and multiple model matching often leads to a severe computational complexity; the direction toward resolving this difficulty that is currently being explored is multiscale (pyramidal) computational concepts for example, based on a fuzzy-logic model-based neural network

pattern recognition - in a narrow (classical) sense refers to a set of techniques in which objects are characterized by a set of classification features (patterns); this space (see classification) is partitioned among classes, and recognition consists in finding which class-region the object-vector belongs to.

 rules -symbolic representations that express physical and mathematical laws that describe the way the world works and how things relate to each other in time, space, causality, and probability. Examples include If/Then rules, formulae in predicate calculus, differential equations, control laws, geometrical theorems, and system models.

scales - a ratio between the measures of two levels of resolution (granularity, hierarchy); a ratio between the size of indistinguishability zones (tessellata) of two levels of resolution.

simultaneously - can be understood in a trivial way only for the events performed and observed at a particular level. Trivially, we mark a point at the time axis and all states corresponding to this point consider "simultaneous". Since all resolution levels have a different time scale, this trivial way cannot be applied. Events and processes belonging to different levels can be considered simultaneous if the time units of consideration overlap (fully, or partially).

search - a procedure of examining the set in expectation that an entity of interest is an element of this set; search is meaningful when a number of additional conditions is satisfied (e.g. such costs should be minimized as the time of search, the losses in subsequent using this entity, and so on). Many methods of search are known such as greedy search, Dijkstra search, A-star search, Dynamic Programming, etc.

semantics - a part of linguistics and a part of semiotics involved with the study of meaning.

semiotics - an area of research which becomes a scientific discipline dedicated to the general laws of modeling systems, in particular - intelligent systems; semiotics addresses these issues via analysis of semiosis; semiotics consists of pragmatics, semantics, and syntactics.

set theory - a part of mathematics which is interested in analysis a properties of collections of entities which do not depend on their encoding.

 successful behaviors - behavior which provides for achievement of the behavioral goals.

sign - a component of the fundamental triad: object-sign-interpretant

symbol grounding - a procedure of verifying whether no changes happened to the meaning of a symbol after formal transformations were performed with this symbol.

 symbolic entity- a data structure that represents a feature, object, or set that exists in the world, or in the world model. A symbolic entity can be a formal list, or frame, consisting of a list head with a name as an address, plus a set of attribute-value pairs, and a set of pointers that define relationships with other symbolic entities or events. These relationships can represent semantic or pragmatic meaning.

 syntactics- the part of semiotics which is responsible for using formal analysis, logical rules, etc.

 tessellatum - an indistinguishability zone, a minimum space at a particular level of resolution within which no further detail can be discovered (because all "further" details are smaller than is allowed by the measure of the particular level of resolution).

 thinking - [we do not give our definition of thinking; we give a definition taken from WWW Britannica Sampler as a food for your thinking about Semiotics: "Thought- converts symbolic responses to intrinsic or extrinsic stimuli"; "thinking - is considered to mediate between inner activity and external stimuli". Are you satisfied?]

object - an entity; it consists of the set of attributes, it can be decomposed in parts, it is a part of another entity, it can have inputs, outputs, states, input-state and state-output functions.

object theory - a theory of representing the world based upon the concept of object.

resolution - is the property of the level of hierarchy which limits the distinguishability of details.

time scale - relation between two measures of time at two levels of resolution; temporal scale and spatial scale are interrelated.

 world - a codeword for the surrounding us reality.

 world encoding - a process of substituting the objects of the world and the relationships among these objects by their symbolic representation.

 world representation - a system which integrates the results of encoding for the objects of the world and the relationships among them.

Can you add any word of importance that we have forgotten of?

 Can you challenge the definitions which we consider beneficial for the further theoretical development and instrumentals for the applications?

 

If yes, then inform the authors of these materials

James S. Albus
E-Mail: albus@cme.nist.gov
Telephone Number: (301) 975-3418

Alex Meystel
E-Mail: meystel@cme.nist.gov
Telephone Number: (301) 975-4455

National Institute of Standards and Technology
Intelligent Systems Division
Bldg. 220 Room B124
Gaithersburg, MD 20899

References

Albus, J.S,. "New Approach to Manipulator Control: The Cerebellar Model Articulation Controller (CMAC)" and "Data Storage in the Cerebellar Model Articulation Controller (CMAC)," Transactions of the ASME Journal of Dynamic Systems, Measurement, and Control", September 1975

 

Albus, J.S., Brains, Behavior, and Robotics, Byte/McGraw-Hill, 1981

 

Albus, J.S., "Outline for a Theory of Intelligence," IEEE Transactions on Systems, Man and Cybernetics, Vol. 21, No. 3, May/June 1991

 

Albus, J.S., "A Reference Model Architecture for Intelligent Systems Design," In An Introduction to Intelligent and Autonomous Control, (Antsaklis, P.J., and Passino, K.M. eds.), Kluwer Academic Publishers, 1993

 

Albus, J.S., Meystel, A., "A Reference Model Architecture for Design and Implementation of Semiotic Control in Large and Complex Systems", in Proceedings of 1995 ISIC Workshop, Monterey, 1995

 

Antsaklis, P. J., Passino, K. M., (eds.), An Introduction to Intelligent and Autonomous Control, Kluwer Academic Publishers, Boston, 1993

 

Armstrong, E., Falk, D., (eds.), Primate Brain Evolution: Methods and Concepts, Plenum Press, New York, 1982

 

Bischof, H., Pyramidal Neural Networks, LEA, Mahwah, NJ, 1995

 

Gupta M. M. , Singha N. K. , (eds.), Intelligent Control Systems, IEEE Press, 1996

Meystel, A., "Nested Hierarchical Control", In An Introduction to Intelligent and Autonomous Control, (Antsaklis, P.J., and Passino, K.M. eds.), Kluwer, 1993

 

Meystel, A., Autonomous Mobile Robots: Vehicles with Cognitive Control, World Scientific, 1991

 

Meystel, A., Semiotic Modeling and Situation Analysis: An Introduction, AdRem, Inc., 1995

 

Nadel, L., Stein, D. L., (eds.), 1993 Lectures in Complex Systems, Addison-Wesley, Reading, MA 1995

 

Newell, A., "Reflections on the knowledge level", Artificial Intelligence, v. 59, 1993, pp. 31-38

 

Newell, A., Simon, H. A., Human Problem Solving, Prentice Hall, Englewood Cliffs, NJ, 1972

 

Perlovsky, L.I. "Computational concepts in classification: neural networks, statistical pattern recognition, and model based vision". Journal of Mathematical Imaging and Vision, 4 (1), 1994.

 

Rosenbloom, P. S., Laird, J. E., A. Newell, (eds.), The Soar Papers: Research on Integrated Intelligence, vv I and II, MIT, Cambridge, MA 1993

 

Russell, S., Norvig, P., Artificial Intelligence: A Modern Approach, Prentice Hall, Upper Saddle River, 1995

 

Sebeok, T. A., Signs: An Introduction to Semiotics, University of Toronto Press, Toronto, 1994

 

Sells, P., Shieber, S. M., Wasow, T., Foundational Issues in Natural Language Processing, MIT Press, Cambridge, MA 1991

 

White, D. A., Sofge, D. A., Handbook of Intelligent Control, VNR, New York, 1992