Electricity: motive power systems – Positional servo systems – Program- or pattern-controlled systems
Reexamination Certificate
2000-10-24
2003-03-18
Nappi, Robert E. (Department: 2837)
Electricity: motive power systems
Positional servo systems
Program- or pattern-controlled systems
C318S568240, C318S568100, C318S568110, C318S568200, C700S245000
Reexamination Certificate
active
06534943
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a robot device and a learning method of a robot device, and more particularly, is suitably applied to a walking-type robot, for instance.
2. Description of the Related Art
Heretofore, a walking pattern of a walking-type robot has been generated in such a manner that the designer selects the respective parameter values of various parameters for walking control by the use of a dedicated program, or selects the respective parameter values of the various parameters by the use of the result that has been led mathematically in accordance with a certain controlling rule.
However, in the case where a lot of walking patterns are required, like the case where such a function that the walking pattern is gradually changed as the robot grows is installed on, for instance, a walking-type amusement robot, it is required to respectively select parameters that are suited for the various parameters for walking control, for each walking pattern, out of a redundant parameter slace. And, there have been such problems that, to perform such a work by hands, the designer has been required to have a high order of expertise of controlling theory, and otherwise, significant time has been required.
As an approach for solving such problems, there is such a method of modeling the walking pattern of a walking-type robot, performing simulations on a computer at the same time of changing sequentially various parameters for walking control, selecting the adequate parameters on the basis of the result of the simulations, and applying those parameters to the actual robot.
However, according to such a method, the result of the simulation is not ensured in an actual robot since it is difficult to correctly describe the model of the robot in a simulation or the dynamic characteristics, physical laws of the robot and the like, or because of the minute difference of the dynamic characteristics in each robot even of the same model. This is counted as a large problem at the site of development of the walking-type robot.
Therefore, it is inferred that, if the walking pattern of a walking-type robot is optimized and can be generated for each robot without requirement of a high order of expertise and/or a complicated work, it will be possible to make the walking pattern generating work easy, and to easily obtain wide variety of walking patterns that are suited to every use.
SUMMARY OF THE INVENTION
In view of the foregoing, an object of this invention is to provide a robot device and a learning method of robot device which can make the walking pattern generating work easy.
The foregoing object and other objects of the invention have been achieved by the provision of a walking-type robot device which comprises: a controlling means for controlling so that walking is performed in response to parameters for controlling the walking; an evaluating means for evaluating the walking; and a parameter updating means for updating the parameters so that the evaluation of the walking by the evaluating means is enhanced.
As a result of this, in this robot device, a walking pattern is gradually improved as the parameters are updated. Hence, according to this robot device, it is possible to obtain a walking pattern having a high evaluation, without requirement of a high order of expertise and/or a complicated work.
Besides, the present invention provides a walking-type robot device which comprises: a controlling means for causing the robot to perform walking that accords with parameters which prescribe the driving phase of each leg of the time of walking; an evaluating means for evaluating the velocity of the walking; and a parameter updating means for updating the parameters so that the evaluation of the walking by the evaluating means is enhanced.
As a result of this, in this robot device, a walking pattern gradually changes from gentle walking into dynamic walking as the parameters are updated. Hence, according to this robot device, it is possible to easily detect a gait that is most suited for the hardware of the very robot and/or the conditions such as the movement environment, as the gait of the time of dynamic walking.
Besides, the present invention provides a learning method of a walking-type robot device that comprises: a first step of causing the robot device to perform walking on the basis of the parameters; a second step of evaluating walking; and a third step of updating the parameters so that the evaluation of walking is enhanced.
As a result of this, in this robot device learning method, a walking pattern is gradually improved as the parameters are updated, and so it is possible to obtain a walking pattern having a high evaluation, without requirement of a high order of expertise and/or a complicated work.
The nature, principle and utility of the invention will become more apparent from the following detailed description when read in conjunction with the accompanying drawings in which like parts are designated by like reference numerals or characters.
REFERENCES:
patent: 5063492 (1991-11-01), Yoda et al.
patent: 5212632 (1993-05-01), Ito et al.
patent: 5343397 (1994-08-01), Yoshino et al.
patent: 5350912 (1994-09-01), Ishida
patent: 5426586 (1995-06-01), Ozawa
patent: 5551525 (1996-09-01), Pack et al.
patent: 5758734 (1998-06-01), Hong et al.
patent: 5841258 (1998-11-01), Takenaka
patent: 5929585 (1999-07-01), Fujita
patent: 6038493 (2000-03-01), Tow
patent: 6068073 (2000-05-01), Roston et al.
patent: 6231140 (2001-11-01), Fujita et al.
patent: 6317652 (2001-11-01), Osada
patent: 6381515 (2002-04-01), Inoue et al.
patent: 6411055 (2002-06-01), Fujita et al.
patent: 6438454 (2002-08-01), Kuroki
patent: 11-188678 (1999-07-01), None
patent: 2000-153476 (2000-06-01), None
Nehmzow, Ulrich “Physically Embedded Genetic Algorithm Learning in Multi-Robot Scenarios: The PEGA Algorithm” University of Essex, Department of Computer Science 1999.
Fujita Masahiro
Hanagata Osamu
Hornby Gregory
Takamura Seiichi
Yamamoto Takashi
Frommer Willliam S.
Frommer & Lawrence & Haug LLP
Nappi Robert E.
Smith Tyrone
Sony Corporation
LandOfFree
Robot device and learning method of robot device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Robot device and learning method of robot device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Robot device and learning method of robot device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3079902