Early human motor development has the nature of spontaneous exploration and boot-strap learning, leading to open-ended acquisition of versatile flexible motor skills. Since dexterous motor skills often exploit body-environment dynamics, we formulate the developmental principle as the spontaneous exploration of consistent dynamical patterns of the neural-body-environment system. We propose that partially ordered dynamical patterns emergent from chaotic oscillators coupled through embodiment serve as the core driving mechanism of such exploration. A model of neuro-musculo-skeletal system is constructed capturing essential features of biological systems. It consists of a skeleton, muscles, spindles, tendon organs, spinal circuits, medullar circuits, and a basic cortical model. Models of self-organizing cortical areas for primary somatosensory and motor areas are introduced. A human infant model is constructed and put through preliminary experiments. Some meaningful motor behavior emerged including rolling over and crawling-like motion. The results show the possibility that a rich variety of meaningful behavior can be discovered and acquired by the neural-body dynamics without pre-defined coordinated control circuits.
Here we will discuss the Self-organization of autonomous embodied motion . Despite being a major characteristic of living systems, Self-movement has never been viewed seriously as a central element of living systems. In fact, most current research focuses on ‘structure’ rather than ‘movement’. The theory of autopoiesis  also does not examine biological movement directly. However, Self-movement often appears as a central theme in robotics research, its self-organization has scarcely been studied. Self-organization of Self-motion is important, because We need to understand the natural intelligence of living systems, opposed to artificial intelligence, its diversity and its root in evolution. As a means for approaching these challenges in robotics research, we designed a simple chemical system that synthesizes embodied autonomous motion: a self-moving oil droplet. This chemical system provides a new example of self-movement besides biological systems and mechanical robots.
We have created a brain-machine hybrid system (BMHS) which is able to solve the chemical plume tracing (CPT) problem using a micro brain of a male silkworm moth. The purpose of the system is to investigate adaptability which is derived from interactions between the microbrain, body, and environment. In this paper, we describe a method of creating the BMHS. The BMHS is a kind of cyborg system in which the body of a living being is replaced with a mobile robot and its brain is used to control the robot. The artificial body is controlled by motor commands for the original body that are extracted from a signal recorded in the brain. For that purpose, a small measurement system to record the signals was created and a method to reconstruct the motor commands from the signals was established. We demonstrate the BMHS can behave like a silkworm moth and then confirmed the system can solve the CPT problem in an experimental environment. Our system has the potential to reveal principles of adaptability in silkworm moth behavior.
Robotics intersects with children in diverse ways; it opens new scientific research fields of understanding humans with collaborating with psychologies and other related fields. At the same time, it is expected to offer useful applications for early childhood education and therapy. But, it has to be recognized that the robotics technology could be a double-edged sword, meaning that it could be hazardous if we did not exploit it in appropriate ways. Robots are thought to have double character; sometimes they show human-like features but other times they just look like an object. We need to understand the character well and consider appropriate forms of application based on the character. In this paper, firstly we will review the past studies around robotics and children, and we will also discuss the potential risk of robotics for children. Then, we will propose some new ideas for the application of robotics for early childhood education and therapy, considering the double character of robots.
This study investigated how infants percept human body movement and humanoid robot movement using eye-tracking system. We recorded eye-movements of 5–, 9–, 12– month-old infants and adults during free-viewing of video clips demonstrating possible and impossible human arm movements and the same movements by humanoid robot. It was found that 5–month-old infants spent more time looking at the human face, regardless of the types of the movement. However, 9–, 12–month-old infants spent less time looking at the face, but more time looking at the arm when they viewed impossible human movement compared to possible human movement. Similar age-dependent changes in the visual fixation pattern were observed in humanoid robot condition. In addition, 5–month-old infants hardly looked at the robot's face-like part regardless of the types of the movement, but the preference for robot's face-like part increased with age. Taken together, these findings suggest the possibility that knowledge concerning the movement of a human body is acquired between 9 and 12 months.
Robots become to be expected to work in our daily lives, but controlling a robot in a real environment by an existing control method would be difficult because it is difficult to model the whole dynamics including the robot and the environment. Since animals can control their body flexibly and robustly in unstructured environments, many robots which reproduce a particular function of an animal have been developed. Despite the progress of these researches, a robot which operates in a real environment has not been developed. In this paper, we focus on the flexibility of biological systems and propose a simple but flexible control method for a robotic system utilizing noise. Experimental results show that our proposed method allows the control of a system without modeling the control target.
Robots are not mere applications of existing theories, but can be huristic tools to find out various new ideas. This paper proposes “huristic bio-robotics” to get new scientific findings by using robots and applys the idea for approaching to adaptive bipedal walking. We have developed a series of biped robots to realize adaptive locomotion. Experimentally realized walking cannot be extrapolated by the existing theories on bipedal walking robots and can contribute to the biomechanics to understand the physical meaning of the biological body.
Robotics has several aspects such as science, technology and art. Therefore, the purpose, method, and evaluation of robotic research are very different depending on the cases. This paper tries to clarify the difference between science and technology in robotics. Moreover, the relation between robotics science and human science is explained in this paper. As science in robotics, this paper point out the importance of synthetic science as well as analytic science. As an example of synthetic science, sensor feedback motion control based on linear transformation form vision to motion is explained and three conditions are revealed. Some simulation results demonstrate the efficacy of the feedback control. Finally, several issues concerning methodology of robot science are discussed.
In this note, we consider a control system that underlies in a biological system. We point out the existence of the Problem of Inseparability in the control system. To understand the principle of mobile adaptability embedded in the control system, we have to solve the Problem of Indivisibility. To solve the problem, we propose a concept of Implicit control law. Finally, we show the Implicit control law plays an important role for constructing the adaptive function of living thing and robot.
This article proposes a concept of soft interface for the connection between physical world where robots act and information world where robot intelligence is realized. First, this author summarizes modeling and control of object grasping and manipulation performed by a pair of soft fingertips. This author points out that a soft fingertip works as a soft interface for stable grasping and manipulation. This author also introduces two-stage control law for stable manipulation by a pair of 2-DOF fingers based on qualitative mapping between control variables and state variables. The proposed control law includes soft interface to cope with uncertainties in real world. This mapping-based approach is then applied to the control of a flexible arm based on polar coordinates.
Simple and easy human-interfaces for designing motion patterns of humanoids or human-like computer graphics have been required. One of the promising approaches is associating motion patterns with some symbol representations, and then generating new motion patterns by synthesizing those symbol representations. The proto-symbol space method based on continuous hidden Markov model (HMM) was such an approach. This method not only abstracts a symbol representation called proto-symbol but also generates new motion patterns by interpolating two proto-symbols in the proto-symbol space constructed with Kullback-Leibler divergence of the proto-symbols. Extrapolation of the proto-symbols as well as synthesis of HMMs that have different number of nodes, however, could not be achieved. In this paper, we propose an algorithm to solve these issues. We further discuss identification of synthesized proto-symbols and estimation of synthetic coefficients. To acquire accurate estimations, Bhattacharyya distance is introduced in the proto-symbol space construction. Finally, we show the feasibility of our approach with simulation experiments by a humanoid robot. This proto-symbol space could be a first synthetic step toward understanding symbolization mechanism according to Deacon's symbol developmental model.
The current paper reviews neuro-robotics studies by the author's group by which possible neuro-phenomenological accounts are explored for “symbolic” processes assumed in various aspects in human cognition. The reviews on the series of studies elucidate that many of “symbolic” computation scheme can be alternated by neuro-dynamic systems approach utilizing self-organization. It is concluded that “symbols” can be obtained just as an immanent property in the consequences of self-organization when natural interactions between the top-down intention and the bottom-up sensory-motor processes are allowed in embodied neuronal systems.
We propose a model of evolutionary communication with voice signs and motion signs between two robots. In our model, a robot recognizes other's action through reflecting its self body dynamics by a Multiple Timescale Recurrent Neural Network (MTRNN). Then the robot interprets the action as a sign by its own hierarchical Neural Network (NN). Each of them modifies their interpretation of signs by re-training the NN to adapt the other's interpretation throughout interaction between them. As a result of the experiment, we found that the communication kept evolving through repeating miscommunication and re-adaptation alternately, and induced the emergence of diverse new signs that depend on the robots' body dynamics through the generalization capability of MTRNN.