ABSTRACT. In research on human communication and child development, the role of attention has become central. In the lecture, I will present some of this research and discuss its implications for how to develop human-robot communication that is as natural as possible. I focus on two questions: (1) How can a robot use the attention of a human to understand what the human wants to communicate? (2) How can a robot control the attention of a human in its communication? For the first question, following human gaze or pointing is required and joint attention should be achieved. The results will be improved if the robot has a model of the interests or goals of the human. For the second question, there are three main methods: speaking, looking and pointing. I will present some results from an ongoing project involving linguistic communication between an iCub and a human and show the importance of attention in the process. Finally, I will present some experiments concerning how humans interpret robot pointing, something which turns out to be quite dependent on the bodily configuration of the robot.

ABSTRACT. In response to the cognitivistic paradigm and its problems, the embodied cognition viewpoint was proposed. In robotics, this resulted in a radical move away from higher-level cognitive functions toward direct, almost "brain-less" interaction with the environment (e.g., behavior-based robotics). While some remarkable behaviors were demonstrated, the complexity of tasks the agents could master remained limited. A natural extension of this approach lies in letting the agents extract regularities in sensorimotor space and exploit them for more effective action guidance. We will use a collection of case studies featuring a quadrupedal and a humanoid robot to concretely explore this space of “minimally cognitive” phenomena, focusing in particular on the concepts of body schema, forward models and sensorimotor contingencies. The studies in the quadrupedal robot will have a more exploratory nature, trying to provide minimal but clear and quantitative examples of the concepts. Studies in the iCub humanoid robot will specifically target the development of body representations, modeling corresponding mechanisms that are believed to operate in primate brains.

ABSTRACT. AI may be emerging from decades in a sort of Middle Ages (MA) period, and now -- with some very major advances in the MA under its belt -- returning to its roots as the science of cognitive agents, with the exalted aim of a computational understanding of the mind.

If so, then it will not be an easy transition. The popular literature -- and some of the technical literature -- is rife with bold characterizations of AI in terms orthogonal (and in some cases diametrically opposite) to this one. But there are reasons to think such a transition is nevertheless underway.

At least so I will argue.

ABSTRACT. A zombie, or more properly a "philosophical zombie", is an imaginary agent that exhibits the observable phenomena of consciousness without actually being conscious. There are many possible variations of the basic concept, but underlying many or most of them is the idea that the zombie should be as capable as a conscious agent of the cognition, and also of the selection and control of action, required for maintaining an autonomous existence in a particular world. In other words, such a zombie would possess an effective cognitive architecture, and one of the issues that should concern this conference is that of the relationship between such an effective cognitive architecture and the machinery underlying the production of the observable phenomena of consciousness. To date, the zombie concept has served only as the basis of philosophical speculation and argument. However, this paper proposes and outlines an attempt to build a zombie, and in particular a physically instantiated "engineering zombie", that would have the potential to make an empirical contribution both to the BICA programme of research, and to consciousness science. Furthermore, this effort would be essentially distinct from any effort to build a conscious machine or robot, because no claims would be made about the agent’s consciousness other than for its absence, and so the usual arguments undermining work on machine consciousness, especially those affecting its funding, would be inapplicable. An intriguing possibility is that the successful construction and demonstration of a zombie might eventually constitute a failure, in that some future insight might show that the agent is in fact as conscious as we are - and is therefore not a zombie.

ABSTRACT. We have been developing BriCA (Brain-inspired Computing Architecture), the generic software platform that can combine an arbitrary number of machine learning modules to construct higher structures such as cognitive architectures inspired by the brain. We discuss requirements analysis and design principles of this cognitive computing platform, report its implementation, and describe plans for further development.

ABSTRACT. In our previous work, we attempted to unravel the mystery of human consciousness by implementing on a robot a consciousness module called a MoNAD, which we think can become the core of consciousness. In this study, we give developmental rules to a conscious system built with these MoNADs, and propose a model for a robot brain that achieves self-evolution only through interaction with the environment.

ABSTRACT. The paper shows that construction of "natural" classifications, "natural" concepts and integrated information base on the same property of the objects of the external world - the high correlation of features, describing the objects of "natural" classes. The hypothesis that the information processes of the brain and mind tuned in the course of evolution to extract highly correlated structure of features of "natural" objects by forming "natural" concepts of the objects, was set up. This hypothesis is justified by references to a number of famous works. The original mathematical model is proposed, which formalizes the "natural" classifications, "natural" concepts and the integrated information by G.Tononi, based on a mathematical representation of the system, closed upon itself by causal relationships, that form a certain "resonance" of mutual predictions of highly correlated set of attributes of objects of "natural" classes. The results of computer modeling of building "natural" classes and concepts of coded numbers, that illustrate the concepts, are introduced.

ABSTRACT. A biologically inspired robot should observe and experience the world apparently as it is, in the same way as we do. It should see and hear like we do, it should be able to touch and feel like we do. We do not convert our sensory sensations into numbers, instead we experience the different qualities of our sensory percepts as the qualities of the world. Why should a robot do it in a different way? One might argue that the robot has to do it in a different way, because computers and microprocessors cannot operate with sub-symbolic analog data; the analog/digital conversion is a must. Indeed, this is true as long as digital symbolic processing is used as the basis for robot brains. But it should also be obvious that a digital microprocessor-based programmed robot will not be able to experience the world in the same way as we do.

However, there is another approach available, namely the associative neural approach that is able to operate inherently in sub-symbolic and symbolic ways without any program code. This approach allows direct sub-symbolic perception and also the utilization of the same percepts as symbols for other things. This symbolic approach allows thus also the use of a natural language.

A biologically inspired robot should also be able to think; imagine, reason and plan ahead its actions. It should be able to learn and acquire a good/bad value framework. It should have short- and long-term memories, both explicit and procedural ones. It should have natural language inner speech and the ability to interact via spoken language. The combination of all these abilities into one system calls for a system architecture. The associative neural network based Haikonen Cognitive Architecture (HCA) is given as an example.

The HCA-based biologically inspired experimental cognitive robot XCR-1 is one of the first (maybe first) robots with inner speech that utilise sub-symbolic/symbolic hard-wired (not simulated) neural networks. XCR-1 may also be the first robot where pain appears as dynamic sub-symbolic system condition. In this talk the structure and capacities of the latest form of XCR-1 are described and the relevance of the experiments with this robot to the research of biologically inspired robots is discussed. A demo video is shown.

ABSTRACT. In this work, we present a distributed cognitive architecture used to control the traffic in an urban network. This architecture relies on a machine consciousness approach - Global Workspace Theory - in order to use competition and broadcast, allowing a group of local traffic controllers to interact, resulting in a better group performance. The main idea is that the local controllers usually perform a purely reactive behavior, defining the times of red and green lights, according just to local information. These local controllers compete in order to define which of them is experiencing the most critical traffic situation. The controller in the worse condition gains access to the global workspace, further broadcasting its condition (and its location) to all other controllers, asking for their help in dealing with its situation. This call from the controller accessing the global workspace will cause an interference in the reactive local behavior, for those local controllers with some chance in helping the controller in a critical condition, by containing traffic in its direction. This group behavior, coordinated by the global workspace strategy, turns the once reactive behavior into a kind of deliberative one. We show in the sequence that this strategy is capable of improving the overall mean travel time of vehicles flowing through the urban network.

ABSTRACT. The paper describes a novel approach to allow a robot to dance following musical rhythm. Starting from a set of given movements, the robot choices sequence of movements a suitable Hidden Markov Model, and synchronize them processing musical input. The proposed approach has the advantage that movements execution probabilities could be changed according evaluation of the dance execution in order to have an articial creative system. In the same way, a choreograph could give major importance to some movements and/or exclude others, using the system as a co-creation tool. The approach has been tested on Aldebaran NAO humanoid using different genres of music, and experimentations was conduct at presence of real human dancers to have feedback of the goodness of the robot execution.

ABSTRACT. The paper proposes a novel cognitive architecture (CA) for computational creativity based on the Psi model and on the mechanisms inspired by dual process theories of reasoning and rationality. In recent years, many cognitive models have focused on dual process theories to better describe and implement complex cognitive skills in artificial agents, but creativity has been approached only at a descriptive level. In previous works we have described various modules of the cognitive architecture that allows a robot to execute creative paintings. By means of dual process theories we refine some relevant mechanisms to obtain artworks, and in particular we explain details about resolution level of the CA dealing with different strategies of access to the Long Term Memory (LTM) and managing the interaction between S1 and S2 processes of dual process theory. An example of artificial painter is described in some experimentations by using a robotic platform.

ABSTRACT. This paper proposes an algorithm for image encoding, pooling and decoding with a FitzHugh-Nagumo model. The FitzHugh-Nagumo model was derived for simulating biological nonlinear response of nerve axon on external stimuli. When elements of the FitzHugh-Nagumo model are placed on a grid and coupled their neighboring elements, a system of the coupled elements organizes stationary pulses depending on its initial condition and coupling strength. By utilizing this nature organizing the stationary pulses on the system, we propose an algorithm for encoding a gray level image into a binary image and pooling the binary image on the system. In particular, Gaussian noise is added to the gray level image in order to obtain pulses randomly distributing according to image gray levels in image encoding. In addition, by providing the encoded binary image to the initial condition of the same system, we recover a gray level image, which means image decoding.

ABSTRACT. Pathology detection with isolated words/connected speech is still open area and needs more investigation. In this study, a novel method based on the voice intensity of the speech signal is used for automatic pathology detection with continuous speech. The proposed method determines the peaks from the speech signal to form a voice contour. Voice disorder database used in this study includes 71 voice samples of normal persons and dysphonic patients each and having five different types of the voice disorders, including vocal folds cyst, laryngopharyngeal reflux disease (LPRD), vocal folds polyp, unilateral vocal folds paralysis and sulcus vocalis. The voice features are calculated by the proposed method which measures the voice intensity of a speech signal by determining the peaks of a signal that lead to make discrimination between normal and pathological voices. The accuracy of the proposed method is 100% and area under the ROC curve is also 100.

ABSTRACT. We have developed an architecture model, MHP/RT, that is capable of sim- ulating people's daily action selection processes as an implementation of the dual-process theory [1]. MHP/RT describes people's daily behavior as a cyclic process of action selection and memory formation. In our daily life, we spend a lot of time in interacting with artifacts, and therefore it is obvious that it should affect development of individual memory systems. At the same time, any particular artifacts that exist as they are should embed in themselves their own histories including their predecessors. Occasional innovations might have caused evolution of memory structure of mankind through people's interaction with the artifacts, and might have resulted in "splicing" evolution in the socio- cultural ecology. MHP/RT suggests that the cyclic processes should dene strong constraints on sustainable innovation; as far as the cyclic processes func- tion in utilizing an artifact implemented in a technological innovation, it should survive, otherwise it should fade away. This paper contrasts technique and skill, the two distinct form of use of artifacts, and derives guidelines for design- ing sustainable artifacts that should support smooth development of people's skill necessary for using artifacts from their mere use via technique.

ABSTRACT. This article presents an artificial learning system for concept learning based on Vector Symbolic Architectures. The system is showcased through functional imitation of the concept learning by honey bees. The presented work uses the results of the real world experiment with honey bees [1] for benchmarking. It is demonstrated that the proposed pipeline features similar learning curve and the accuracy of generalization as in the living bees.

ABSTRACT. A dynamic computational cognitive model can be used to explore a selected complex cognitive phenomenon by providing some features or patterns over time. More specifically, it can be used to simulate, analyse and explain the behaviour of such a cognitive phenomenon. It generates output data in the form of time series which can only be partially compared to empirical knowledge. This leads to a challenging problem to estimate values of the parameters of the model representing characteristics of a person. A parameter estimation approach for dynamic cognitive models is presented here by combining improved Particle Swarm Optimization (PSO) and Constraint Satisfaction (CS) methods. Having collected the key features of behaviour of a phenomenon, those are translated into a set of constraints with parameters that will be solved through an improved agent based PSO technique. Through this, within PSO each agent explores the complex search space while communicating the quality of a local parameter value vector relative to their current global best solution as a swarm (through cooperation and competition). This is performed in tournaments and results of each tournament are combined to address the premature convergence issue in PSO.

ABSTRACT. Abstract This paper revolves around the concept of Affordance. It aims to develop an architectural typology based on the ecological concept of affordance. In order to achieve this aim it was based on two sources as following: 1- Gibson's definition of the concept of affordance. 2- The researches which concerned about affordance categorization, mainly Gaver Zhang and Patel contribution in this field. As a result, this paper concluded 16 typologies of affordances beside the possibilities of mixing it based on the both sources. To clarify these new typologies and provide further understanding, a wide range of architectural examples are presented and proposed, which is reported in the first part of the paper, (the analytic study). To prove this new vocabulary’s capability to diagnose and evaluate the affordance of different environments two processes have been adapted: A. Diagnostic process: the interpretation of the environments with regard to its affordance using the new vocabulary (the developed typologies). B. Evaluating process: the evaluation of the environments that have been interpreted and classified with regard to their affordances. By using the measures of emotional experience (The positive affect ‘PA’ and the negative affect ‘NA’), and the Architectural Evaluation Criteria (beauty, economy and function) which is reported in the second part of the paper, (the experimental study).The experimental study proved that the new typologies were capable reading the affordance in different environments. Also it explained how these different typologies reflected different interactions based on the previous processes. The data which was concluded from the evaluation of measures, explained how the different typologies of affordance which already reflect different environments had different evaluations. Some of them are recommended and the others are not. In other words, this paper draws a roadmap for the designers to diagnose, evaluate and afterwards analyze the affordance in the different architectural environments. Then it guided them to adapt the best interaction (affordance category) which they intended in their proposed designs.

ABSTRACT. Biological organisms effortlessly acquire a visual model for familiar spaces, enabling them to localize and find paths and do many other tasks. On the other hand, for robots, all poses must be calibrated against a canonical referenced frame, and even tasks driven by vision require state estimation onto these coordinates. In this paper, we attempt to develop a visual sense of the ambient space using only a large set of untagged images, without any reference to ground coordinates, motor parameters etc. We first introduce the {\em Visual Manifold Theorem} which states that the images captured by a camera mounted on a motor system will lie on a low-dimensional manifold homeomorphic to the motor manifold. The coordinates of the system on this manifold are in fact, generalized coordinates that describe the motion, and can be used as alternatives to the canonical coordinates traditionally used in robotics. We demonstrate this process for a simulated robot exploring a planar space. A quick exploration of the space can be used to generate a manifold based on the similarity of images captured from nearby viewpoints, without any knowledge of the motion coordinates. The work makes an useful contribution both to robotics and to computational models for acquisition of place cells in cognition.

ABSTRACT. This article presents a web-based game functionally imitating a part of the cognitive behavior of a living organism. This game is a prototype implementation of an artificial online cognitive architecture based on the usage of distributed data representations and Vector Symbolic Architectures. The game demonstrates the feasibility of creating a lightweight cognitive architecture, which is capable of performing rather complex cognitive tasks. The cognitive functionality is implemented in about 100 lines of code and requires few tens of kilobytes of memory for its operation, which make the concept suitable for implementing in low-end devices such as minirobots and wireless sensors.

ABSTRACT. We have been conducting research with the objective of enabling a robot to perform human-like autonomous behavior and communication. For this purpose, we have developed and mounted on a robot consciousness modules, termed MoNADs, which by means of groups of neurons comprise the elements that make up the functions of consciousness, emotions and feelings that are possessed by humans. In this study, we propose a consciousness model that enables a robot to evolve in accordance with rules based on definitions of pleasant and unpleasant states in a conscious system comprised of MoNADs. For the definitions of the pleasant and unpleasant states, we are using the “smoothness of the information flow” in the conscious system, where “pleasant” is the state in which the information that has come in is flowing smoothly, while the state in which information is not flowing smoothly is defined as “unpleasant.” By means of these definitions, the robot becomes able to autonomously evolve using uniform and consistent rules in both of the systems of the brain, that which governs reason and that which is responsible for emotions and feelings.

ABSTRACT. We have implemented a personal mobility performing semiautonomous control by estimating the avoidance direction of pedestrian and the avoidance judgment timing of passenger. In space of pedestrian and personal mobility coexist, it is possible to realize a safety collision avoidance by moving a mobility in the lateral before passenger's avoidance judgment. Therefore, we utilize a Microsoft Kinect which obtains the coordinates of both feet of pedestrian. And we have collision avoidance experiments between a pedestrian and a mobility with the implemented model estimating avoiding direction from feet's relative positions. We have evaluated the important body site in avoiding judgment and avoidance judgment timing. As a result, passengers gaze at pedestrian's lower body in semiautonomous control, and avoidance judgment timing is delayed about pedestrian's one step. Based on the results, we have proposed a model of passenger's motion perception and vision guidance.

ABSTRACT. This paper describes the rationale followed for the integration of Dual-PECCS, a cognitively-inspired knowledge representation and reasoning system, into two rather different cognitive architectures, such as ACT-R and CLARION. The provided integration shows how the representational and reasoning mechanisms implemented by our framework may be plausibly applied to computa- tional models of cognition based on different assumptions.

ABSTRACT. Although the thermal grill illusion has been the topic of previous research, many mysteries still remain regarding psychological determinants, neurophysiological mechanisms and so on. Also, the illusion cannot be simulated by information science and robotics. This study focuses on a very simple but interesting experiment called Hot and Cold Coils, which is known as a typical example of the thermal grill illusion. The authors aim to explain the thermal grill illusion by proposing a new and bold assumption called the conflict of concepts, and demonstrate how to construct a model by using a novel artificial consciousness module called the Module of Nerves for Advanced Dynamics (MoNAD). A simple experimental apparatus was prepared to prove the existence of the thermal grill illusion, and consists of a parallel arrangement of bars with an alternating pattern of cold and warmth at 20°C and 40°C. The authors conclude with the belief that many complex perceptions of humanity can be simulated through the use of neural networks, and that this can help us to deeply study the cognitive processes of human perception.

ABSTRACT. This paper presents a new approach to driving experimentation, based on cognitive simulation of the driver in order to predict human behaviour. The cognitive model COSMODRIVE (i.e. Cognitive Simulation MOdel of the DRIVEr) has been combined with a Vehicle-Environment-Sensors platform (named SiVIC, for Simulateur Vehicule-Infrastructure-Capteur) in order to simulate, explain and predict the driver's behaviour and mental activities. From this simulation, an experiment has been conduct at IFSTTAR - LESCOT, where hypothesis has been made with the simulation. The innovative approach is the use of a virtual simulation of a cognitive model to predict the human behaviour and then analyse collected data to validate the predicted behaviour. This article describes broadly the COSMODRIVE model and the simulation made in order to define accurate experimental hypothesis. Then, we describe the driving simulator and the experiment itself. Afterwards, data analysis provides us some results allowing us to discuss and conclude about the methodology tested with this experiment.

ABSTRACT. A new and rapid pattern recognition architecture, which employs a multilayer autonomous ratio-memory cellular nonlinear network (MARMCNN) algorithm is presented. Evolved from the celebrated idea of a Cellular Neural Network (CNN - or sometimes known as Cellular Nonlinear Network), which focuses on analogue processing, the MARMCNN is expected to provide ample processing capabilities as well as the possibility of a real hardware implementation. The proposed MARMCNN employs a rapid and modified Hebbian rule as its learning algorithm. It has been used in the classification of Harumanis mangos using MATLAB simulation as the implementation platform. Using 1000 profile odour data from 200 samples of Harumanis mangos in 5 different classification categories, a 4-layer 6�6 MARMCNN managed to produce a successful ripeness recognition rate of 98.6% while using only 5 training patterns in 10 iterations. This is comparable to a much more complex 3-layer Multi-Layer Perceptron (MLP) feed-forward Artificial Neural Network (ANN) with 4 hidden nodes using Levenberg-Marquardt learning algorithm and a Radial Basis Function (RBF) network with 4 hidden nodes. In short, the MARMCNN is capable of learning, recognizing, and classifying a complete range of gray scale odour data patterns rather convincingly. The performance, learning time, complexity and the flexibility of the proposed MARMCNN is discussed and compared to other two established classical neural network (MLP and RBF).

ABSTRACT. Validating computational models of human behavior typically involves statistically comparing human data collected during an experiment to predictions made by the model. However, these models very rarely attempt to represent sex, despite the growing indication that there are sex-based differences in neural and behavioral responses to some external stimuli. We make the case for a stronger presence of male and female models of behavior in biologically inspired cognitive architectures, an area of research that is especially susceptible to physiological differences that can cause bottom-up behavioral differences. We discuss results that highlight the importance of providing more focus on sex-based differences, including data we collected that show different performance distributions between males and females during a decision-making task. We conclude with suggestions of potential areas of application for models that take into account differences in males and females.

ABSTRACT. In this paper we describe an emotional human-machine interface as an anthropomorphic social agent which is able to comprehend emotions and react to emotional stimuli. We propose a neurobiologically inspired agent implementation that is based on mechanics of chemical and physiological processes within human brain. Implementation of model features simulation of neuromodulators such as dopamine, serotonin, and noradrenaline. Demonstration of emotions is achieved via combining aforementioned neuromodulators in different proportions. The Lovheim cube of emotions is used for this purpose. We also cover the topic of "uncanny valley" phenomenon. In conclusion of this paper we reached understanding of connection between mimics and motivation of individual. We have constructed realistic computation model which allows us to visualize agent's mimics in sync with his speech.

ABSTRACT. When we interact with other person, we need a knowledge on other’s mind and consult it to decide action of our self. We call the knowledge Model of Other, MoO. To realize a cognitive architecture to interact with human, MoO must be embedded and used in a decision making process. But to design MoO, we must know how human interact in real world in MoO view because we aim a practical human interaction task. So, in this paper, we conducted a child-robot play experiment and analyzed human behavior. From the result, we construct a model of emotion level mental status leading that may be a possible leading concept of human-agent interaction design.

ABSTRACT. Current controversy notwithstanding, biological intelligence must be, for use in the design of artificial intelligence, explicable as an entirely physical and deterministic decision machine. Such an explanation is offered here, with adaptive dynamic closed-loop control as its characterizing feature. Differing as well as confirming prior art is discussed. A conclusion is that no alleged weirdness of the human mind credibly denies us its mechanistic emulation.

ABSTRACT. The disconnect between scientific schools of thought across the world is most notable in cognitive, neural, and computer sciences. The intersection of these fields is exactly where a powerful new approach has emerged recently, known as Biologically Inspired Cognitive Architectures (BICA). Several years ago, the BICA Society started a new initiative: to build an international community of researchers unified by the BICA Challenge: the challenge to create a real-life computational replica of the human mind captured in its highest cognitive functionality, using solutions inspired by the brain. Key elements of the BICA Society mission to promote and integrate the many efforts addressing the BICA Challenge include the Annual International Conference on BICA (also known as the Annual Meeting of the BICA Society); the Elsevier journal BICA (that is indexed by Web of Science and Scopus and acquires a JCR Impact Factor early in 2016); and the upcoming public online repository developed by collective efforts, named “The MAPPED Repository of BICA” by its six components: Models, Architectures, People, Paradigms, Evaluations, and Dialogues (the latter include Videopanels). In this year, BICA Society celebrates another impressive success of its annual meeting: this time in Lyon, France. Since 2010, it has been a tradition to hold a BICA Society Panel at each BICA Society Meeting, reviewing the progress and planning for the future. This time, given only 15 minutes, we need to discuss many things, among which are (1) future BICA events: BICA 2016 in New York in July, the school “FIERCES on BICA” in Moscow in 2016 and in 2017, and the conference BICA 2017 in Moscow; (2) the MAPPED Repository of BICA, including systematic reviews of BICA among other components, and (3) our initiative to establish the James S. Albus Medal as a prestigious award issued by the BICA Society for the best recent contribution to the solution of the BICA Challenge. Details will be debated. The panel will be held as a plenary event open to all participants and will be followed by a Gala dinner on the boat Hermes in a magnificent city tour on the Rhône and Saône rivers.