|
|
|
| HOME | PROGRAM OVERVIEW | ORGANIZATION | SITE MAP |
|
| Tetsuo Sawaragi |
| Complex System Control and Design Group Leader |
| Department of Mechanical Engineering and Science |
| Graduate School of Engineering |
The focus of research on a future mechanical system needs to be changed from the machine itself as an entity that is highly accurate and highly efficient to a whole integrated system, which involves the environment that surrounds the machine and a human who operates it. In general, there are limited numbers of systems in which entire functions are accomplished by machines alone. In most cases, the interactions between human and the external environment accomplish the original functional purposes of mechanical systems; however, the theory for its system design and control has yet to be established.
Essentially, autonomous and proactive processes, typically seen in living systems, are not steered only by external forces. Instead, they can autonomously change the relationships among the internal elements that constitute themselves, while taking in external disturbances and adapting themselves to them. In this way, the internal dynamics of each element and the interactions among these elements form a mutual feedback system. Our research group will perform mathematical and experimental analysis of these adaptation processes of internal dynamic systems, and develop a system design theory using those process models. To this end, we have focused on the following three key subjects:
We will cope with the problems inherent in a design of autonomous mobile robots, a design of man-machine systems and a systemic functional emergence arising from the interactions among organic cells and non-linear material elements. We will carry out our research through active and progressive collaboration among all the group members, focusing on the following three subjects (Figure 1) and seeking the possibility of fusion and integration between these disparate research topics:
Complex mechanical systems can be defined as mechanical systems that consist of multiple elements with their complex interactions and that form a variety of structures and behaviors being affected by an external environment. Each element has its own internal dynamics, and these internal states encounter the competition between two contradicting trends: "stability", which is associated with the extent of autonomy maintained inside, and "adaptability", which represents plasticity for adaptation to the environment and surrounding elements. Furthermore, interactions between the elements underlie a further level of dynamics that allows the evolution of complex behavior, and at the same time, a rational functional design is realized by selecting a nominal option, while other versatile options are suppressed.
Our group aims to clarify the principles with which systems dynamically and autonomously form orders and emerge new functions, and apply our findings to the design of mechanical systems in which functional elements constituting the system transform their nature in response to their environment. This innovative approach to mechanical system design is rooted in the evolutionary and adaptive principles found in life systems that are characterized by its nature of plasticity and loose-couplingness present among components and their interactions. Since complex systems are defined not by any fixed relationships, but by the evolving interactions between its constituent elements, conventional analytical methods are insufficient. We augment them by using constructive approaches, in which we develop a simulation model, and compare its dynamics with experimental observations (Figure 2). In other words, we build up our understanding of a phenomenon by combining several basic processes using an elementary model, aiming to acquire a constructive understanding of nature. Although quantitative forecasts will be difficult to make using such analytical methods, this approach will play a significant role in the qualitative prediction and comprehension of phenomena of universal behavior classes. In this way, we hope to make complex subjects comprehensible and applicable for practical development.
In this section, we briefly explain our group's research programs.
The stability of motion in multi-dimensional dynamical systems represents a mechanism in which their behaviors are compressed into low dimensional dynamics through interaction among their constituent elements. In autonomous mobile mechanical system design, the degree of freedom is compressed in the manner that even involves a degree of freedom of the environment in addition to that intrinsic to the robot, which creates qualitative stability (isomorphism) of motion patterns. At the same time, by generating versatility within, adaptation to the environment is created. Alternate repetition of the order-structure-formation phase and collapse phase underlies the adaptation of an autonomous mobile mechanical system, which we consider to represent a complex system.
The group of Tsujita and Aoi (the Tsuchiya laboratory) carries out research on the control and design theory of biped robots, taking self-organization and phase transition by control parameters as the principle of motion control. In this research, the nervous system, which is capable of generating voluntary activity patterns related to locomotion, draws in and strongly couples the actions while interacting with a body that is in physical contact with the environment. This results in a mechanism that can generate versatile and adaptive walking motions. The group designs quadruped and biped robots that incorporate this motion control system, which autonomously forms and affects four types of locomotion patterns (walk, trot, bounce, and pace) according to changes in its environment, such as floor inclination, walking speeds and loads. The group develops this mechanism further into one that can create voluntary movements. As an example, the group has shown that by demonstrating the robot's motion and posture control when turning to an intended direction over various turning radii.
Nakanishi studies the design of systems that would dynamically control adaptation in response to complex changes in dynamic characteristics, in order to deal with unanticipated changes in an environment or in a model. In this research, he uses an unmanned helicopter to explore control system design using a neural network as an adaptive component. By combining off-line learning on a simulator and on-line learning in the actual environment, he demonstrated robust adaptation to the problems encountered by actual machines such as model learning errors, changes in environment such as ground effects and gusts, and changes in dynamic characteristics. As a control method with versatility, he adds multiple modules that can adapt to environmental changes and pursue control system design in which individual modules selectively learn adaptations to complex environmental changes and function together as a coherent system.
For social robots that perform social interactions with people using body motions, motion learning is an important technical issue for a robot to enhance its autonomy by adaptively organizing its pre-existing internal structures and to elicit human responses. However, true social behavior in robots is probably not possible, given the limitations on abilities to construct and use an objective external environment model to forecast accurately the behavior of other people. Learning should be focused on the process of transforming the robot itself, rather than model the environment. Through its interaction with others and its internalization, robots define a new reality, then constantly change and optimize their behavior. Taniguchi in the Sawaragi laboratory studies the ability of face robots to trace moving objects. He has proposed an adaptive organization mechanism that allows the robot to organize tracing motions intrinsically, without external instruction signals in learning in the sensorimotor system. In this way, the robot can trace and keep an object in its line of vision by way of adaptive body movement, and learn new strategies for doing so through experience.
Complex phenomena generally occur at interfaces where antagonistic heterogeneous effects coexist. At the interface of man-machine systems, multiple peripheral influences interact and interfere with intrinsic properties to generate such complex phenomena.
Such behaviors often exceed design specifications. For example, on a footbridge, the rhythm of a human and that of the rolling oscillations of the bridge interact with each other, and human rhythm unconsciously synchronizes with the movement of the bridge and thus the rolling becomes amplified. The research group of Utsuno in the Matsuhisa laboratory carries out research on the interaction of such rolling oscillations of bridges with body motion governed by the nervous system that works as a rhythm generator. They analyze these complex behaviors observed at the time of human locomotion on a light and flexible structure such as a pedestrian bridge. They have found that this phenomenon of entrainment of a human's walking pace can be experimentally reproduced by the use of a trapezoid pendulum model, and have also performed mechanical model analysis of the dynamics of the interactions involved in the synchronization.
Yokokoji and Saida in the Yoshikawa laboratory describe the importance of coherence of vision and tactile sense at vision-tactile sense interface to the virtual environment. Specifically, they are investigating a bi-directional motion transfer based on the concept of mechano-media, where a mechanical system takes charge for mapping of human action beyond spatial and temporal aspects. This knowledge, they hope, will enable the design of robots that perform these motions with human-like flexibility, with all the multiple degrees of freedom they entail. Furthermore, through the analysis on the velocity profile of hand and finger movement in grasping motion on a virtual platform, they aim to elucidate the general principles on the selection and combination of degrees of freedom depending on the object type, as well as the segmentation of motions.
Horiguchi in the Sawaragi laboratory conducts studies based on an assumption that the interactions occurring in a human-robot collaboration and the strength of their global association are determined by a specification of an interface design connecting these autonomies. The group has found that these properties manifest themselves in the dynamics of both autonomous and collaborative behavior of human and robot. Often, they have observed that mutually adaptive behaviors become coordinated, thus optimizing the work output of a human-machine combination. Finally, the group has been investigating interface design for tele-operation robot, which is intended to promote the bi-directional exchange of intentions by equivalent and semi-independent parallel loops between a human and a robot, and to share "isomorphism of tasks" through mutual adjustment of their individual behavior.
To understand the behavior of a living organism, it is essential to elucidate the organic behavior of aggregated as well as single cells. In a life system, there is inherent diversity at the individual level, but at the group level, there is also an inherent mechanism that becomes increasingly more stable and deterministic. It is an extremely important challenge to elucidate the dynamics governing the process in which a macroscopic order emerges and disintegrates controlling the degrees of freedom in such an organization or group consisting of multiple elements. Such knowledge may be utilized for the design of a dynamic and functional mechanical structure system.
Yamamoto, in the Tomita laboratory, studies medical engineering focusing dynamics of organisms and their environments in terms of hierarchy in a biological system in the context of interactions among cells and between cells and tissues. In an initial experiment to observe the structure and function of cartilage cells and tissues without imposing any dynamic environment conditions, ES cells did not differentiate into cartilage cells; however, the group aims to confirm that, when a dynamic state of the environment is altered, morphology and function are enhanced to organize and adapt to the environmental changes. To investigate the dynamics of this system further, specifically the functional and structural adaptations that constituent units undergo, the group has developed an ES cell-cartilage regeneration simulation model using cellular automata.
In order to create microstructures that simulate living organisms, the emergence of hierarchies in the interaction fields, and the pattern formation mechanism in view of topological changes of such functions ought to be elucidated. Compliant mechanisms that actively exploit the structural flexibility of a mechanical structure can realize the mechanical function as the structure itself by adding required flexibility to an appropriate position within the structure. It is, therefore, suitable for structures that cannot be composed entirely of rigid structures. Nishiwaki, in the Yoshimura laboratory, investigates new topological design optimization in structure design of compliant mechanism. To date, this type of optimization has only been conducted empirically; this group intends on studying it analytically, and is focusing especially on vibrating structures. They also have developed this theory into multi-stage design optimization with the intention of designing a method of the functional mechanical structure system for multi-physics phenomenon that deals with the physical coupling phenomena.
| HOME | PROGRAM OVERVIEW | ORGANIZATION | SITE MAP |
|