Search

This book presents biologically inspired walking machines (four- and six-legged walking machines) interacting with their real environmental stimuli as agent-environment interactions. Different reactive behaviors of animals were investigated for the behavior design of the walking machine(s). On the one hand, the obstacle avoidance behavior, in analogy to the obstacle avoidance and escape behavior of scorpions and cockroaches, was implemented in the walking machines as a negative tropism. On the other hand, the sound tropism which mimics prey capture behavior of spiders is represented as a positive tropism. It was simulated on the four-legged walking machine. [ABSTRACT FROM AUTHOR]

Where Chap. 2 investigated the biologically inspired perception-action systems, this chapter focuses on applying the principles of the biological domain to create artificial perception-action systems. First, several preprocessing units of different types of sensory signals are presented. They are used to filter and recognize the corresponding sensory signals and they can be described as perception parts. Second, the neural control of the four- and six-legged walking machines, which generates and controls the locomotion of the machines, is described. Third, the combination of the neural preprocessing and the neural control is explained. It gives rise to the ability of controlling reactive behaviors such as obstacle avoidance and sound tropism. Finally, both behavior controls are merged under a so-called behavior fusion controller by applying a sensor fusion technique to give a versatile perception-action system. [ABSTRACT FROM AUTHOR]

In order to test the capabilities of the artificial perception-action systems, several experiments were carried out. First, the signal processing networks were tested with the simulated signals and the real sensor signals. Afterwards the physical sensors, the neural preprocessing and the neural control were all together implemented on the physical walking machine(s) to demonstrate different reactive behaviors. [ABSTRACT FROM AUTHOR]

This chapter describes the development of the physical components that lead to the artificial perception-action systems. It begins with the descriptions of different physical sensors which are used to sense environmental information, followed by the details of the walking machines simulated in a physical simulation environment as well as the robots we have built. [ABSTRACT FROM AUTHOR]

This chapter presents methods and tools which are to be used throughout this book. It starts with a short introduction to a biological neuron together with an artificial neuron which is followed by the comparison of network structures between feedforward and recurrent neural networks. Then the discrete-time dynamical properties of the single neuron with a recurrent connection are described. Finally, artificial evolution is presented as a tool to develop and optimize neural structures as well as the strength of synapses. [ABSTRACT FROM AUTHOR]

Most of this book is devoted to creating and demonstrating so-called artificial perception-action systems inspired by biological sensing systems (perception) and animal behavior (action). Thus this chapter attempts to provide the biological background for understanding the approach taken in this book. It begins with a short introduction to some of the necessary principles of animal behavior. Then it concentrates on the obstacle avoidance and escape behavior of a scorpion and a cockroach, and continues with the prey capture behavior of a spider. Here, attention is given to the biological sensing systems used to trigger the described behaviors. Furthermore, different morphologies of walking animals are presented as inspiration for the design of walking machine platforms. Finally, a biologically inspired locomotion control, called a "central pattern generator" (CPG), is also discussed. This concept is later employed to generate the rhythmic leg movements of the machines. [ABSTRACT FROM AUTHOR]

Research in the domain of biologically inspired walking machines has been ongoing for over 20 years [59, 166, 190, 199, 207]. Most of it has focused on the construction of such machines [34, 47, 216, 223], on a dynamic gait control [43, 117, 201] and on the generation of an advanced locomotion control [30, 56, 104, 120], for instance on rough terrain [5, 66, 102, 180, 192]. In general, these walking machines were solely designed for the purpose of motion without responding to environmental stimuli. However, from this research area, only a few works have presented physical walking machines reacting to an environmental stimulus using different approaches [6, 36, 72, 95]. On the one hand, this shows that less attention has been paid to walking machines performing reactive behaviors. On the other hand, such complex systems can serve as a methodology for the study of embodied systems consisting of sensors and actuators for explicit agent-environment interactions. [ABSTRACT FROM AUTHOR]

When an application needs a fusion mechanism, the developer has to solve an essential problem: the construction of the appropriate model. This corresponds to (i) the selection of an aggregation operator and (ii) the determination of its parameters. This process should take into account several factors. Some of them are highlighted here. [ABSTRACT FROM AUTHOR]

This chapter reviews some of the existing tools for evaluating aggregation methods and their parameters. We focus on some indices for fuzzy measures (Shapley and Banzhaf), an interaction index, and the degree of disjunction. Other methods exist. The influence function and other tools such as grosserror sensitivity and local-shift sensitivity developed in robust statistics (see Section 2.2.6) are of interest here. The tools permit us to have some knowledge on how a particular estimator might behave when embedded in a real system. In particular, we have seen that the influence function of the arithmetic mean is unbounded while that of the median is bounded. [ABSTRACT FROM AUTHOR]

In this chapter we review some aggregation operators for numerical information. While in Chapter 4 description was centered on functional equations, and operators were introduced as a natural consequence of some basic properties (unanimity, positive homogeneity, and so on), here, operators are introduced for greater modeling capabilities and generality. This progression into general aggregation operators leads to a review of operators that are particular cases of Choquet and Sugeno integrals. On the one hand, the Choquet integral generalizes not only arithmetic mean and weighted mean (the most widely used and well-known aggregation operators), but also OWA operators. On the other hand, the Sugeno integral generalizes weighted minimum, weighted maximum, and median operators. In the rest of this chapter we will use Choquet integral family to refer to aggregation operators that are generalized by the Choquet integral. In the same way, the Sugeno integral family will refer to aggregation operators that the Sugeno integral generalizes. [ABSTRACT FROM AUTHOR]

Most aggregation operators use some kind of parameterization to express additional information about the objects that take part in the aggregation process. Applying the jargon of artificial intelligence, we can say that the parameters are used to represent the background knowledge. For example, it is well known that in the case of the weighted mean, the weights — i.e., the weighting vector — play this role. In an application, we can use them to express the reliability of the information sources (sensors, experts, and so on). For example, when fusing data from sensors, we can express wich sensor is more likely to give data of better quality and which is more likely to give erroneous data. In a similar way, other aggregation functions use other parameterizations. [ABSTRACT FROM AUTHOR]

In this chapter we study some aggregation operators for numerical information. The description is focused on results based on functional equations. Therefore, not only are the operators given, but also, at least for some of them, their characterization. We refer to these results as syntheses of judgements. Although the term could be used for any aggregation operator, we restrict its use to the case of characterizations using functional equations. [ABSTRACT FROM AUTHOR]

Functional equations are equations where the unknowns are functions. A well-known example of functional equation is the following Cauchy equation: (3.1)$$ \phi (x + y) = \phi (x) + \phi (y). $$ A function φ is a solution of this equation if, for any two values x and y, the application of φ to x + y equals the addition of the application of φ to x and to y. Therefore, the equation establishes conditions that functions φ have to satisfy. Typical solutions of this Cauchy equation are the functions φ(x) = αx for an arbitrary value for α. [ABSTRACT FROM AUTHOR]

In this chapter we will review some of the concepts that are needed later in the book. In particular, we focus on measurement theory and some basic elements of probability theory and fuzzy sets theory. [ABSTRACT FROM AUTHOR]

Information fusion techniques, in general, and aggregation operators (or aggregation functions), in particular, are extensively used in several fields of human knowledge. They are used to produce the most comprehensive and specific datum about an entity from data supplied by several information sources (or the same source at different periods of time). They are used in systems to reduce some type of noise, increase accuracy, summarize information, extract information, make decisions, and so on. To illustrate this, we consider below some examples in different fields. Some of the typical applications are also included. [ABSTRACT FROM AUTHOR]

Section 1 discusses the conceptual foundations of general intelligence as a discipline, orienting it within the Integrated Causal Model of Tooby and Cosmides; Section 2 constitutes the bulk of the paper and discusses the functional decomposition of general intelligence into a complex supersystem of interdependent internally specialized processes, and structures the description using five successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. Section 3 discusses probable differences between humans and AIs and points out several fundamental advantages that minds-in-general potentially possess relative to current evolved intelligences, especially with respect to recursive self-improvement. [ABSTRACT FROM AUTHOR]

The proposal is a radical one — that human cognition is significantly weaker than we presume and AI significantly closer than we dared hope. I believe that the human mind is largely made up of tricks and sleights of hand that enamor us with much pride; but our pedestal might not be quite so high or robust as we imagine. I will pursue the argument that human cognition is based largely on 3D simulation and as such is particularly vulnerable to co-option by future advances in animation software. [ABSTRACT FROM AUTHOR]

The chapter argues that the investigations of evolutionary processes that result in human intelligence by means of mathematical/computer models can be a serious scientific basis of AI research. The "intelligent inventions" of biological evolution (unconditional reflex, habituation, conditional reflex...) to be modeled, conceptual background theories (the metasystem transition theory by V.F. Turchin and the theory of functional systems by P.K. Anokhin) and modern approaches (Artificial Life, Simulation of Adaptive Behavior) to such modeling are outlined. Two concrete computer models, "Model of Evolutionary Emergence of Purposeful Adaptive Behavior" and the "Model of Evolution of Web Agents" are described. The first model is a pure scientific investigation; the second model is a step to practical applications. Finally, a possible way from these simple models to implementation of high level intelligence is outlined. [ABSTRACT FROM AUTHOR]

It is difficult to develop an adequate mathematical definition of intelligence. Therefore we consider the general problem of searching for programs with specified properties and we argue, using the Church-Turing thesis, that it covers the informal meaning of intelligence. The program search algorithm can also be used to optimise its own structure and learn in this way. Thus, developing a practical program search algorithm is a way to create AI. To construct a working program search algorithm we show a model of programs and logic in which specifications and proofs of program properties can be understood in a natural way. We combine it with an extensive parser and show how efficient machine code can be generated for programs in this model. In this way we construct a system which communicates in precise natural language and where programming and reasoning can be effectively automated. [ABSTRACT FROM AUTHOR]

Sequential decision theory formally solves the problem of rational agents in uncertain worlds if the true environmental prior probability distribution is known. Solomonoff's theory of universal induction formally solves the problem of sequence prediction for unknown prior distribution. We combine both ideas and get a parameter-free theory of universal Artificial Intelligence. We give strong arguments that the resulting AIXI model is the most intelligent unbiased agent possible. We outline how the AIXI model can formally solve a number of problem classes, including sequence prediction, strategic games, function minimization, reinforcement and supervised learning. The major drawback of the AIXI model is that it is un-computable. To overcome this problem, we construct a modified algorithm AIXItl that is still effectively more intelligent than any other time t and length l bounded agent. The computation time of AIXItl is of the order t·2l. The discussion includes formal definitions of intelligence order relations, the horizon problem and relations of the AIXI theory to other AI approaches. [ABSTRACT FROM AUTHOR]

We present the first class of mathematically rigorous, general, fully self-referential, self-improving, optimally efficient problem solvers. Inspired by Kurt Gödel's celebrated self-referential formulas (1931), such a problem solver rewrites any part of its own code as soon as it has found a proof that the rewrite is useful, where the problem-dependent utility function and the hardware and the entire initial code are described by axioms encoded in an initial proof searcher which is also part of the initial code. The searcher systematically and efficiently tests computable proof techniques (programs whose outputs are proofs) until it finds a provably useful, computable self-rewrite. We show that such a self-rewrite is globally optimal—no local maxima!—since the code first had to prove that it is not useful to continue the proof search for alternative self-rewrites. Unlike previous non-self-referential methods based on hardwired proof searchers, ours not only boasts an optimal order of complexity but can optimally reduce any slowdowns hidden by the O()-notation, provided the utility of such speed-ups is provable at all. [ABSTRACT FROM AUTHOR]

Most traditional artificial intelligence (AI) systems of the past 50 years are either very limited, or based on heuristics, or both. The new millennium, however, has brought substantial progress in the field of theoretically optimal and practically feasible algorithms for prediction, search, inductive inference based on Occam's razor, problem solving, decision making, and reinforcement learning in environments of a very general type. Since inductive inference is at the heart of all inductive sciences, some of the results are relevant not only for AI and computer science but also for physics, provoking nontraditional predictions based on Zuse's thesis of the computer-generated universe. [ABSTRACT FROM AUTHOR]

This chapter introduces the idea of "Evolvable Hardware," which applies evolutionary algorithms to the generation of programmable hardware as a means of achieving Artificial Intelligence. Cellular Automata-based Neural Networks are evolved in different modules, which form the components of artificial brains. Results from past models and plans for future work are presented. [ABSTRACT FROM AUTHOR]

Understanding general intelligence and identifying its essential components are key to building next-generation AI systems — systems that are far less expensive, yet significantly more capable. In addition to concentrating on general learning abilities, a fast-track approach should also seek a path of least resistance — one that capitalizes on human engineering strengths and available technology. Sometimes, this involves selecting the AI road less traveled. I believe that the theoretical model, cognitive components, and framework described above, joined with my other strategic design decisions provide a solid basis for achieving practical AGI capabilities in the foreseeable future. Successful implementation will significantly address many traditional problems of AI. Potential benefits include: minimizing initial environment-specific programming (through self-adaptive configuration);substantially reducing ongoing software changes, because a large amount of additional functionality and knowledge will be acquired autonomously via self-supervised learning;greatly increasing the scope of applications, as users teach and train additional capabilities; andimproved flexibility and robustness resulting from systems' ability to adapt to changing data patterns, environments and goals. AGI promises to make an important contribution toward realizing software and robotic systems that are more usable, intelligent, and human-friendly. The time seems ripe for a major initiative down this new path of human advancement that is now open to us. [ABSTRACT FROM AUTHOR]

The Novamente AI Engine, a novel AI software system, is briefly reviewed. Novamente is an integrative artificial general intelligence design, which integrates aspects of many prior AI projects and paradigms, including symbolic, probabilistic, evolutionary programming and reinforcement learning approaches; but its overall architecture is unique, drawing on system-theoretic ideas regarding complex mental dynamics and associated emergent patterns. The chapter reviews both the conceptual models of mind and intelligence which inspired the system design, and the concrete architecture of Novamente as a software system. [ABSTRACT FROM AUTHOR]

Is there an "essence of intelligence" that distinguishes intelligent systems from non-intelligent systems? If there is, then what is it? This chapter suggests an answer to these questions by introducing the ideas behind the NARS (Nonaxiomatic Reasoning System) project. NARS is based on the opinion that the essence of intelligence is the ability to adapt with insufficient knowledge and resources. According to this belief, the author has designed a novel formal logic, and implemented it in a computer system. Such a "logic of intelligence" provides a unified explanation for many cognitive functions of the human mind, and is also concrete enough to guide the actual building of a general purpose "thinking machine". [ABSTRACT FROM AUTHOR]

In this chapter we will give a brief overview about what different methods of evaluation were applied to the SmartKom prototypes and which of them resulted in utilizable results and why others did not. Since there are no established benchmarking methods yet for multimodal HCI systems and very few debatable methods for monomodal dialogue systems our work within the SmartKom prototype evaluations was rather an exploration of new methods than the simple application of standard routines. [ABSTRACT FROM AUTHOR]

We present three types of data collections and their experimental paradigms. The resulting data were employed to conduct a number of annotation experiments, create evaluation gold standards and train statistical models. The data, experiments and their analyses highlight the importance of data-driven empirical laboratory and field work for research on intuitive multimodal human-computer interfaces. [ABSTRACT FROM AUTHOR]

There are several characteristics not optimally suited for the user state classification with Wizard-of-Oz (WOZ) data like the nonuniform distribution of emotions in the utterances and the distribution of emotional utterances in speech, facial expression, and gesture. In particular, the fact that most of the data collected in the WOZ experiments are without any emotional expression gives rise to the problem of getting enough representative data for training the classifiers. Because of this problem we collected data in our own database. These data are also relevant for several demonstration sessions, where the functionality of the SmartKom system is shown in accordance with the defined use cases. In the following we first describe the system environment for data collection and then the collected data. At the end we will discuss the tool to demonstrate user states detected in the different modalities. [ABSTRACT FROM AUTHOR]

Do users show emotions and gestures if they interact with a rather intelligent multimodal dialogue system? And if they do, what do the "emotions" and the gestures look like? Are there any features that can be exploited for their automatic detection? And finally, which language do they use when interacting with a multimodal system — does it differ from the usage of language with a monomodal dialogue system that can only understand speech? To answer these questions, data had to be collected, labeled and analyzed. This chapter deals with the second step, the transliteration and the labeling. The three main labeling steps are covered: orthographic transliteration, labeling of user states, labeling of gestures. Each step will be described with theoretical and developmental background, an overview of the label categories, and some practical advice for readers who are themselves in the process of looking for or assembling a coding system. Readers who are interested in using the presented labeling schemes should refer to the cited literature — not all details necessary for actually using the different systems are presented here for reasons of space. For information on the corpus itself, please refer to Schiel and Türk (2006). [ABSTRACT FROM AUTHOR]

This chapter gives a concise overview of the empirical Wizard-of-Oz recordings done within the SMARTKOM project. We define the abstract specifications of the intended simulated communicative situations, describe the necessary technical setup (including numerous useful practical hints), and finally outline the specifications of the resulting multimodal corpus, which may be obtained from the Bavarian Archive for Speech Signals (BAS). [ABSTRACT FROM AUTHOR]

People tend to spend an increasing amount of time in their cars and therefore desire high comfort, safety, and efficiency in that environment. A large variety of electronic devices has been made available to meet these requirements in the vehicle. These electronic devices should allow for speech interaction in order to minimize driver distraction and to maximize driver comfort. This contribution studies the user requirements for potential assistant functionalities operated by speech in the car. The architecture is of the dialogue system is defined based on the user requirements study. Speech dialogues were designed according to state-of-the-art principles of human machine interaction for the functionalities desired by the users. Results and ideas for future work conclude this contribution. [ABSTRACT FROM AUTHOR]

This paper presents SmartKom-Mobile, the mobile version of the SmartKom system. SmartKom-Mobile brings together highly advanced user interaction and mobile computing in a novel way and allows for ubiquitous access to multidomain information. SmartKom-Mobile is device-independent and realizes multimodal interaction in cars and on mobile devices such as PDAs. With its siblings, SmartKom-Home and SmartKom-Public, it provides intelligent user interfaces for an extremely broad range of scenarios and environments. [ABSTRACT FROM AUTHOR]

SmartKom-Home demonstrates the use and benefit of an intelligent multimodal interface when controlling entertainment devices like a TV, a recorder, and a jukebox, and when accessing entertainment services like an electronic program guide combining speech and a handheld display with touch input. One important point is emphasizing the functional aspect, i.e., the user's needs, conveyed to the system in a natural way by speech and gesture, are satisfied. The user does not need to know device-specific features or service idiosyncrasies. The function modeling component in SmartKom-Home has the necessary knowledge to transform the abstract user request into device commands and service queries. [ABSTRACT FROM AUTHOR]

SmartKom-Public is the result of consistent development of traditional public telephone booths for members of a modern information society in the form of a multimodal communications booth for intuitive broad-bandwidth communication. [ABSTRACT FROM AUTHOR]

This chapter describes the English-language SmartKom-Mobile system and related research. We explain the work required to support a second language in SmartKom and the design of the English speech recognizer. We then discuss research carried out on signal processing methods for robust speech recognition and on language analysis using the Embodied Construction Grammar formalism. Finally, the results of human-subject experiments using a novel Wizard and Operator model are analyzed with an eye to creating more felicitous interaction in dialogue systems. [ABSTRACT FROM AUTHOR]