DISTANCE SKILL LEARNING
Non-verbal motional skill plays a very important role in manufacturing. The transfer of skill is important in two aspects: One is for passing the skill on to others, either from generation to generation or from one place to another. The other is for integrating this information into design. For both of these purposes, the symbolization of motional skill is necessary in order effectively to represent it and to use it.
Therefore, in this work, first we digitized motion data and allotted attributes to each data chunk. This research work aims at achieving a very interactive skill teletraining system which permits a student to learn very interactively with his teacher so that he or she can understand what are the points and where they should pay their attention in learning the skill motion.
Our system digitizes the model motion of a teacher using Fastrak and a student follows this motion which is sent by way of a video and the student's motion is corrected by a computer based upon the comparison of the data between a teacher and a student. What characterizes our system is that the teacher can interrupt at any moment and give necessary instructions to what he or she should do during his or her motion.
This is not only useful for transferring skill over the network across a long distance, but it also serves a great deal in transferring skill from generation to generation. Presently, we do not have an adequate tool for that purpose. Another advantage might be that we could use it for describing the knowledge and experience which have been difficult to put into manuals because they are non-verbal in nature. Our system may be able to put such non-verbal knowledge or experience in data base in the broad sense so that such manufacturing skill may be taken into account at the stage of design, too.
Then, we developed a client/server system on the internet for facilitating scaling up. The teacher and the student can teach and learn on their own client machines. Thus, our system permits the teacher and the student to share the processes of learning in addition to the results.
INTRODUCTION
Products are getting more and more diversified and they are being produced over a wide region. Thus, it becomes more important to transfer or share manufacturing skill between distant locations. Especially in Japan, as can be understood from the fact that OJT (On the Job Training) plays an important role in transferring manufacturing skill, non-verbal skill in the form of motion plays an important role in manufacturing.
But the conventional product development has been one way from design to manufacturing. Although the introduction of concurrent engineering played a very important role in integrating design and manufacturing, still some of the manufacturing knowledge and experience are hard to bring them toward the earlier stages. Motional skill is one of them.
However, there are some means of transferring motional skill. For example, there are distance learning system which may be used for this purpose. But most of these systems have been relying heavily on such verbal media as text, figures, tables, etc and even if they may deal with motion, most of them are by way of TV or video so that the information flow is only one way from a teacher to students, and the feedback from a student to a teacher was very poor. Besides, they just send the motion without attaching any semantic meaning to it, so that the student find it very difficult to know the points of focus and on the other hand the teacher also finds it very difficult to understand how the student is progressing.
Further, if we wish to integrate motional skills into design, the mere transmission of motion would not be enough. We need to discretize the motion and turn them into symbolic forms.
The conventional distance learning systems have been developed only with conveying the final motion to the receiver on mind. But to transfer motion, we have to transmit not only its result but also its processes. For example, the teacher does not understand if the student is doing well or not unless he receives how the student is following the model motion step by step. In other words, the points of focus should be shared between the sender and the receiver.
The discretization and symbolization of motion would enable a sender and a receiver to share the common perspective and to establish true interactiveness between them. Of course, if motion is digitized, it would greatly reduce the transmission burden and by symbolizing motion, we could attach semantic meaning to each motion.
It must be pointed out that if we use a network for distance learning, the amount of team learning among students decreases and it becomes difficult for a teacher to grasp the whole class. Thus the burden increases immensely on the part of a teacher, compared with the conventional class room teaching. The same kind of problem arises if we transfer motional skills over the network. Thus, more flexible and interactive environment is needed for skill transfer over the network.
In this work, a prototype system is developed, base upon internet client/server technology to realize the skill transfer with the aim of solving the above mentioned problems.
DEVELOPMENT OF SKILL TRAINING SYSTEM USING FASTRAK
System Outline Using Fastrak
Most distance learning systema up to now send knowledge and experience one way from a teacher to a student. There are few, if any, that permit feedback from a student to a teacher. This kind of interactiveness is essential for skill training. Therefore, we take up calligraphy as a sample of skill training. In our preliminary system, the position of a hand is measured in 3-D using a magnetic sensor, and guide the student in real time to a right motion by utilizing a musical tone.
The system is composed of
(a) 3-D measuring part,
(b) Data base for storing and retrieving the right motion,
(c) Judgment part,
(d) Musical guidance part,
as shown in Fig.1
Fig. l System Outline
The system flow is
Fig.2 3-D measuring part up
Fig.3 Retrieval of right motion up
(c) Judgment part
Differences in elevation, pitch and roll between right and a student's motions are evaluated.
The motion of a pentip is evaluated based on which sector it is located at among 8 divided
sectors of a circle (Fig.4).
(d) Musical guidance part
If a difference is detected, a student is guided to a right motion by a musical tone.
Fig.4 Guidance Part up
DEVELOPMENT OF CLIENT / SERVER SYSTEM
Securing Interactiveness
In learning motion, it is very important that not only the motion but also the points of focus are shared between the sender and the receiver. In the conventional classroom, the teacher and the students share the room physically so that this kind of problem was much smaller.
Most of the conventional distance learning systems divide the system into two: One for teaching and the other for learning and the interactiveness or the reciprocity between the sender and the receiver was not considered at all or if considered, it is very poor . In fact, most of these systems are of the presentation type.
Thus, in this work, we considered the following three interfaces:
(1) teacher-student interface (man-man interface)
This is the fundamental interface. The teacher sends to the student instructions, motion, its
points of focus and receives from the student his or her results of learning, and sends back
the corrected motion with necessary guidance. The student learns from the model motion
and sends his or her results of learning and learns what are inadequate by studying the
corrected motion (Fig.5(a)).
(2) teacher-computer and student-computer interface (man-machine interface)
This is the interface for the teacher to make corrections and prepare guidance, and for the
student to learn where there are and what are the inadequacies.
In this system, motion is digitized and attributes are allotted to the chunk of motion data.
The teacher inputs the model motion and make corrections for the student's result based upon
these digitized data and attributes. The student inputs his or her motion.
In our system, the computer can compare the model and learning motions while the student
is in the process of learning. Therefore, almost the same environment can be realized as if
the teacher is directly teaching the student on the spot. Thus, the student can learn very
effectively since he or she can get very quick response from the "virtual" teacher, and can
repeat the process any number of times as he or she desires. This interface takes care of the
synchronous aspects of teaching and learning (Fig.5(b)).
(3) Computer-network interface
This interface takes care of the asynchronous aspects and uses the client/server technology.
Motion data are digitized in terms of positions in 2 dimension and are stored in and
controlled by the server with the attributes. Thus, the teacher and the student can retrieve
data whenever they need to so that it reduces the burden of the teacher considerably and the
student find it easier to share the points of focus with the teacher. And the client systems are
divided in role into two: One for teaching and the other for learning. Thus, our system is
smaller in scale compared with the other systems (Fig.5 (c)).
Fig.5 Concept of Interfaces for teletraining process up
Outline of the Client/Server System
Our system is composed of teaching client, learning client and file server systems. And the teaching client system is composed of motion extraction, motion display and motion correction subsystems. And the learning client system is composed of motion display, motion training and corrected motion display subsystems (Fig.6).
Fig.6 System outline up
In this work, we took up the sample of penmanship for simplicity.
The flow of the system is as follows:
(1) Input the model motion data of the teacher to the server.
(2) Download the model motion data to the learning client system.
(3) Display the model motion on the learning client system.
(4) Repeat learning using the motion training subsystem.
(5) Save the results of learning and transfer data to the teaching client system.
(6) Correct the student's motion using the motion correction subsystem and prepare
guidance for correction and save these data.
(7) Download the corrected motion data to the learning client system and the student
repeats the learning processes.
The system was developed using Sun Microsystems JDK (JavaDevelopers Kit) ver. 1.0.2 and NetScape Navigator ver.3.0.1 was used for WWW browser. And the system was implemented on Apple Macintosh PowerPC7200/90 and Quadra950.
Teaching Client System
[ Motion extraction subsystem] This subsystem extracts the model motion of the teacher as 2 dimensional position data and saves them on the server. In this system, a window is created on a WWW browser to extract motional data and the teacher writes a character using a pen tool. To extract a model motion, not only the geometry of a character and the sequence of the pen motion, but also the motion of a pen between characters where no characters are being written are very important. This subsystem allots attributes to the pen motion during the character writing and between writing and extracts 2 positional data in real time, and save them on the server (Fig.7(a)).
[Motion display subsystem] This subsystem retrieves the student's data from the server and
reproduce his or her motion using the animation function. And it should also be emphasized
here that not only the motion of character writing, I. e. the geometry of a character and the
sequence of writing a character, but also the motion between writing are also reproduced.
In most of the conventional systems, the teacher can only receive the final geometry of the
character or if he or she can receive the sequential data, he or she cannot easily make
corrections on these data to feed them back to students (Fig.7 (b)).
[ Motion correction subsystem] The teacher confirms the pen motion of the student using motion display subsystem and gives corrections on the motion using a pen tool and the degree of inadequacy is shown by the color of a pen. And he or she gives necessary guidance together with these corrections. These data of the teacher are stored in a file separate from that of the student.
In our system, the repeated processes of learning can be stored. So that the teacher can observe each process and can find very easily what points are difficult for the student to learn or where the student makes mistakes very often. Thus, he or she can improve his way of teaching and may develop newer methods for improved teaching (Fig.7 (c)).
Fig.7 Teaching client system up
Learning Client System
[Motion display subsystem] This subsystem reproduces the teacher's model motion from the stored digitized data using the animation function and shows whether the motion is during writing or during pen movement without writing by different colors (Fig.8 (a)).
[Motion training subsystem] The student observes the model motion by using the motion display subsystem, learns it and save the learning result and send it to the server (Fig.8(b)).
[Corrected motion display subsystem] This subsystem shows before correction (Fig.8 (c)).
Fig.8 Training client system up
File Server System
The file server system stores the model motion, student's results and his or her corrected motion. Whenever it received a demand call either from a teaching client system or from a learning client system, the file is transmitted. The network communication between the file server and the clients is based on TCP/IP protocol (Fig.9).
Fig.9 File server system up
Characteristics of The System
(1) The teacher can observe the process of learning and can give guidance based upon these observations. The student can learn effectively by comparing his or her motion with the corrected one and by identifying where there are inadequacies. Therefore, the teacher and the student can easily share the points of focus so that effective learning can be possible.
(2) The synchronous and asynchronous aspects in teaching and learning are made clear. So that the following features can be given to the system (Fig. 10).
(3) By controlling all motion data uniquely on the server, the teacher can understand how the student is progressing and can make and the learning process are separately taken care of by the client systems as the teaching client and the learning client so that the whole system is much smaller compared with the conventional ones.
(4) Motion is digitized in terms of 2 dimensional positions and attributes are allotted to each data chunk. Thus, all data are uniquely represented so that the burden of transmission decreased considerably.
(5) Java language is used. This facilitates the easy use of the client system on the part of the teacher and the student by using GUI on the WWW navigator, and it permits multi-platform usage.
Fig.10 System features up
SUMMARY
Non-verbal motional skill plays a very important role in manufacturing. The transfer of the skill is important in two aspects: One is for passing the skill on to others, either from generation to generation or from one place to another. The other is for integrating this information into design. For both of these purposes, the symbolization of motional skill is necessary in order effectively to represent it and to use it.
Most of the systems developed so far to transfer motion skill is of the presentation type and
information is transmitted one way from a sender to a receiver. Even if they feed back the
receiver's reaction,
the function is very poor because the sender and the receiver cannot share the points of
focus
and cannot interact with each other. Therefore, seamless interaction between the two cannot
be secured
in the present available systems.
In learning or in transferring the motional skill, not only the result but also the process of learning and the points of focus should be shared between the teacher (sender) and the student (receiver). But the frameworks in the present systems are not capable either fundamentally or practically due to very heavy transmission burden.
To solve these problems, motion data have to be intelligently digitized, coded and turned into symbols. Therefore, in this work, we digitized motion data and allotted attributes to each data chunk and developed a client/server system on the internet. The teacher and the student can teach and learn on their own client machines.
Our system permits the teacher and the student to share the processes of learning in addition to the results. So that it is easy to share the points of focus or to tell what or where is important. Our system reduces the burden of a teacher considerably. In the conventional distance learning systems, the burden of a teacher increases considerably with the increase of students. So the present systems can accommodate only the limited number of participants. We distinguished the synchronous and asynchronous aspects in teaching and learning so that the teacher and the student can teach and learn at any time and at any place each desires. The transmission burden decreased, therefore, very much in our system and the use of Java made our system multi-platform accessible.
Although there are still many problems left for solution, we firmly believe this is the first step in the right direction toward developing such a system for transferring the motional skills which are very important in manufacturing.
This work is financially supported in part by the Grant-in-aid (International Academic Studies) 08044164 from the Ministry of Education, Japan. One of the authors, Shuichi Fukuda, would like to thank Japan Society for the Promotion of Science to let him spend 8 months in the US to study on the situations in the area of distance learning. This opportunity provides us the motivation for this work.
(1) S.Fukuda, Y.Matsuura and T.Takahashi, 1966, 2nd Image Sensing Symposium, 113-116 (in Japanese)
(2) S. Fukuda, Y. Matsuura and P. Wongchuphan, 1996, Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, 786
(3) S. Fukuda, Y. Matsuura, 1996, Japan-USA Symposium on Flexible Automation, ASME
(4) T. J. Olscheske, 1996, 12th Annual Conference on Distance Teaching and Learning, Post conference workshop
Back to Table of
Contents