Kuperstein, M. (1988). Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....training time criteria. However, at run time the extra dof cannot be exploited for other purposes nor can the optimization criteria be changed 8 . Explicitly regularized inverse functions Other direct inverse kinematics approaches using neural networks can be found in [Kuperstein 87] Kuperstein 88] Kuperstein 91] Barhen, Gulati Zak 89] Martinetz, Ritter Schulten 90] Ritter, Martinetz Schulten 89] among others. A great advantage of these algorithms 7 They discuss the possible use of neural networks to implement a pseudo inverse technique, however they do not implement such a ....

Michael Kuperstein, "Neural Model of Adaptive Hand--Eye Coordination for Single Postures", Science, Vol. 239, pp. 1308--1311, 1988.

....the plant output is provided as an input to the learning controller, and the controller is required to produce as output the corresponding plant input. This approach, shown diagrammatically in Figure 18, is known as direct inverse modeling (Widrow Stearns, 1985, Atkeson Reinkensmeyer, 1988; Kuperstein, 1988; Miller, 1987) Note that we treat the plant output as being observed at time n. Because an inverse model is a relationship between the state and the plant input at one moment in time with the plant output at the following moment in time (cf. Equation 11) the plant input (u[n] and the state ....

....angles # there is a corresponding Cartesian position vector y. The mapping from # to y is a nonlinear function known as the forward kinematics of the arm. Suppose that we use the direct inverse modeling approach to learn the inverse kinematics of the arm; that is, the mapping from y to # (cf. Kuperstein, 1988). Data for the learning algorithm are obtained by trying random joint angle configurations and observing the corresponding position of the tip of the arm. A nonlinear supervised learning algorithm is used to learn the mapping from tip positions to joint angles. Figure 21 shows the results of a ....

Kuperstein, M. (1988). Neural model of adaptive hand-eye coordination for single postures. Science, 239, 1308-1311.

.... but it may also be the case that feedback from the environment can function as a form of supervision , albeit a noisy one. The idea of using vision to provide feedback has been studied in a variety of contexts, including control of eye movements [66] coordination of hand and eye movement [67, 68] and object recognition and manipulation [69] The precise form of the feedback and learning mechanisms vary, but 22 Image T u u u 1 y y World State T w 1 j j T v 1 System State Figure 12: Transformations in active vision. Changes in the world state induce changes in the ....

M. Kuperstein,"Neural Model of adaptive Hand-Eye Coordination for Single Postures ",Science, 239, pp.1308-1311, 1988.

....inverse models prove more problematic. If the correct motor command was known, which could provide an appropriate supervised error signal, then there would be no need for the inverse model. Three main approaches have been used to adapt such inverse models direct inverse modeling (Miller, 1987; Kuperstein, 1988), distal supervised learning (Jordan and Rumelhart, 1992) and feedback error learning (Kawato, 1990) The latter two models both rely on the ability to convert errors in the actual trajectory into errors in the motor command. They, unlike the direct approach, are able to acquire an accurate ....

Kuperstein M. (1988). Neural model of adaptive hand-eye coordination for single postures. Science, 239, 1308--1311.

.... et al. 1993; von Hofsten Ronnqvist, 1993) This research is complemented by a wide variety of computational models that simulate learning to reach (Berthier, 1996 1997; Berthier, Singh, Barto, Houk, 1993; Bullock Grossberg, 1988; Hinton, 1984; Kawato, 1990; Kettner, Marcario, Port, 1993; Kuperstein, 1988; Rosenbaum, Loukopoulos, Meulenbroek, Vaughan, Engelbrecht, 1995; Sporns Edelman, 1993; Vos Scheepstra, 1993) There are several important differences, though, between how these models simulate motor learning and how infants learn to reach. First, many models employ a supervised learning ....

Kuperstein, M. (1988). Neural model of adaptive hand-eye coordination for single postures.

....an important element in the development and testing of models, they must be viewed as only a stepping stone to the more interesting problems. This need of interfacing with the real world has not been completely ignored by the research community. Some examples include the work of Kawato (1990) and Kuperstein (1988). These control systems implemented for this work, however, are generally very specific to the neural control algorithm of interest, as well as to the system to be controlled. The work of Brooks (1986) is one example of a more general control architecture. This system building philosophy allows ....

Kuperstein, M., (1988) "Neural Model of Adaptive Hand-Eye Coordination for Single Postures," Science 239:1308-1311.

....needed to achieve the desired state. As the dynamics of the arm change due to growth, damage, fatigue and changes in external loading, the inverse model must be adaptable. Three main approaches for adaptation have been proposed in such inverse models direct inverse modeling (Miller 1987; Kuperstein 1988), distal supervised learning (Jordan and Rumelhart 1992) and feedback error learning (Kawato 1990) The latter two models both rely on the ability to convert errors in the actual trajectory into changes in the motor command and are able to acquire an accurate inverse model even for redundant ....

Kuperstein, M. (1988). Neural model of adaptive hand-eye coordination for single postures.

....system are discussed. I. Introduction The adaptive capabilities of motion control of biological organisms are still highly superior to the capabilities of current robot systems. Therefore, various neural network models have been developed that apply biologically inspired control mechanisms [2, 6, 7, 11, 12, 15, 16] to robot control tasks. During the last two years it has been demonstrated by means of robot simulations that the neural network model [10, 12, 17] based on Kohonen s algorithm for self organizing maps [4, 5] can be utilized for visuomotor control. In the present paper we will report on an ....

....the workspace of the robot arm. Furthermore, the system succeeded to rapidly adapt to drastically changing situations. These algorithms achieve a precision that is higher by one order of magnitude than earlier neural network implementations, like Kuperstein s system with 4 6 average deviation [6, 7]. The accuracy is currently limited by the image processing resolution and not by the control algorithm. There are several fairly simple possibilities to enhance the performance: the use of additional cameras for accuracy sensitive parts of the work space or employing cameras with better ....

M. Kuperstein. Neural model of adaptive hand-eye coordination for single postures. Science, 239:1301--1311, 1988.

....of arm control follow a learning paradigm where the first step is the calibration of the system. In theory, some external teacher provides such plant knowledge, or the system calibrates itself by performing certain training movements. Subsequently it learns to control the arm ( Kalveram, 1991] [Kuperstein, 1988]) Such a separation of calibration and control is not observed in the development of biological systems. Here calibration and control are not two distinct and sequential phases of development, but are rather intertwined, proceeding in parallel, and build upon each other. Today this view of a ....

....Today this view of a parallel development of calibration and control processes seems widely accepted by researchers working on neural modeling of adaptive eye hand coordination. Yet, most researchers model this process as a learning and not as a developmental operation ( Jordan and Flash, 1994] [Kuperstein, 1988]) Implicit to such an approach of artificial eye hand coordination is the premise that all behaviors of the system have to be learned. However, this assumption is not necessarily true for biological systems. One major difference between a biological and artificial system is that a biological ....

M. Kuperstein, "Neural model of adaptive hand-eye coordination for single postures.," Science, 239:1308--1311, 1988. Sensori-motor development 30

....controller must act in order to learn the result of its action. When training a neural network to control a robotic arm, one may explore by allowing the controller to flail for a length of time, moving the arm at random through coordinate space while it builds up data from which to build a model [Kuperstein, 1988]. This is not feasible, however, if actions are expensive and must be conserved. In these situations, we should choose a training trajectory that will get the most information out of a limited number of steps. Manually designing such trajectories is a slow process, and intuitively good ....

M. Kuperstein. (1988) Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311.

....et al. 1993) and it might be possible that such an internally generated proprioceptive signal could be used to drive recall of learned movements requiring very rapid feedback. A number of models of arm movements have been developed (Flash, 1987; Kawato et al. 1988; Bullock Grossberg, 1988; Kuperstein, 1988; Uno et al. 1989; Massone Bizzi, 1989; Lukashin, 1990; Jeannerod, 1990; Gaudiano Grossberg, 1991; Mussa Ivaldi Giszter, 1992; Burnod et al. 1992; Houk et al. 1990; Lukashin Georgopoulos, 1993; Berthier et al. 1993) Many of these are complementary to our model (although see Houk et ....

Kuperstein, M. (1988) Neural model of adaptive hand-eye coordination for single postures.

.... redundancy of of the arm allows for many solutions to the inverse, causing a form of ill posedness known as the degrees of freedom problem [1] Approaches to learning the inverse kinematic map by sampling the (x; space and directly estimating a function = f (x) have met with some success [13]. However, as Jordan and Rumelhart (1992) have pointed out, the non convexity of the map places a lower bound on the achievable error of any direct least squares algorithm. Jordan and Rumelhart propose an indirect approach to this non convexity problem based on forming an internal model of the arm ....

M. Kuperstein, "Neural model of adaptive hand-eye coordination for single postures," Science, vol. 239, pp. 1308--1311, 1988.

....A consequence of this rapid learning is that errors are not repeated and can be eliminated much more quickly than approaches that incrementally update parameters. Nonlinear parametric models can be trained by 1) exposing the model to a new data point only once (e.g. Jordan and Jacobs, 1990; Kuperstein, 1988)) or 2) by storing the data in a database and cycling through the training data repeatedly. In case 1, much more data must be collected, since the training effect of each data point is small. This leads to slower learning, since real robot movements take time, and to increased wear and tear on ....

Kuperstein, M. (1988). Neural Model of Adaptive Hand-Eye Coordination for Single Postures. Science, 239:1308--3111.

....model results in good generalization, while a choice that involves more than one type of inverse solution results in bad generalization at the boundary between the points by which the solutions have been specified. The same effect is observed when we do straightforward inverse modeling, as in [7]. We performed direct inverse modeling for the two cases (Figure 7a and Figure 8a) and the results look similar to those from Figures 7b and 8b. 5 Conclusions The method of distal supervised learning proved to be successful in learning the inverse kinematics of a six DoF anthropomorphic robot. ....

M. Kuperstein. Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311, 1988.

No context found.

Kuperstein, M. (1988). Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311.

No context found.

M. Kuperstein. Neural model of adaptive hand-eye coordination for single postures. Science, 239. 1308-1311, 1988.

No context found.

M. Kuperstein. Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311, 1988. 150

No context found.

M. Kuperstein, "Neural model of adaptive hand-eye coordination for single posture, " Science, vol. 239, 1988, pp. 1308-1311.

No context found.

M. Kuperstein. Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311, 1988.

No context found.

M. Kuperstein. (1988) Neural model of adaptive hand-eye coordination for single postures. Science, 239:1308--1311.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC