| Teaching Solid Modelling |
5. Teaching solid modelling: the Indian scenario
Pramod Koparkar
Abstract
This paper discusses the key determinants in the teaching of solid modelling at various levels in universities and educational institutions, based mostly on the author's teaching experience over the last fifteen years or so. Much of the material relates to the situation in India, although the conclusions may well coincide with experience worldwide. The determinants considered are hardware, software, trained manpower, study material and industrial participation. The paper concludes with a snapshot of the situation today.
Hardware
The picture in respect of hardware seems now to be satisfactory as a result of advancing technology: particularly the Pentium chip, which is available at a reasonable cost. This processor is powerful enough to run solid modelling software, and is within the reach of majority of institutions. The situation was different three or four years ago, when the cost of hardware had a tremendous impact on decisions made while starting a new solid modelling course at any institution. Today, fortunately, this aspect has lost its importance.
Software
Software is a major cost in solid modelling, whether for teaching or any other application. At the lower end of the market, prices are around £UK 5,000-7,000: and this is only for a single-user licence. A typical Indian classroom has around 40-60 students taking a course; multiple-user licences are therefore needed. Typically, a class would require a ten-seat licence, so that all the students could get their hands on the system, albeit through time-sharing. This is a major expense in the Indian environment, where a typical instructor would receive a monthly salary of £ UK 100-150. Many institutions would prefer to employ ten new instructors for an existing course rather than to start a new course on solid modelling.
Another important consideration is the representation scheme used in the available modeller. Ideally, from a pedagogical point of view, the students should learn and understand all three schemes: B-rep, CSG and oct-trees. Most of the modellers available support only one of them, although they may provide one of the remaining two at the interface (i.e. a boundary modeller with a tree of Boolean operations recorded at the interface comprising a de facto CSG model). Hardly any modellers incorporate all three schemes into one package, let alone allowing interchange from one representation to another.
One solution to this problem would be to have three modellers, each with one representation. But this is only a theoretical possibility in the light of the cost analysis above.
If solid modelling software were made available in the public domain, perhaps that would help sort out these issues. Open-ended software would be welcome.
Trained manpower
The solid modelling curriculum demands skills in three different areas: computer science, mechanical engineering, and mathematics (specifically, geometry). Finding an instructor trained in all three is quite difficult. Some might show ability in two disciplines, but people comfortable in all three are very rare.
People from a strong computer science background are usually prepared to receive further training in mathematics and geometry, but are rather reluctant to look at mechanical engineering. They consider it too application-oriented.
People with a mechanical engineering background are generally conversant with the mathematics involved: but typically they are weak in aspects related to complicated data-structures (like winged-edge structures, as maintained through pointers in C or Pascal), or a hierarchy of objects (for instance, derived classes in ), or analysis of the performance of programming constructs (such as the use of recursion).
People good at mathematics are generally interested in more theoretical problems-inverting matrices, for instance-rather than the problems faced in a practical application like solid modelling. Changing the viewpoints of such people is quite difficult.
In short, due to the inter-disciplinary nature of solid modelling, getting good members of faculty is difficult: however, this problem is not restricted to India, but appears to be more or less universal.
Study material
Solid modelling suffers from a dearth of good textbooks, even though it has been an established discipline for over five years now: only four or five major texts are currently available. This shows a sharp contrast with computer graphics, for which there is a wealth of books. Many modelling books focus only on surfaces, and a distinction between solid modelling and geometric modelling is not maintained. A comprehensive treatise on solid modelling is very much needed.
Each of the existing books has an emphasis on one of the three basic representation schemes (B-reps, CSG or oct-trees). A student is forced to read more than one book in order to study the important aspects of all three schemes. Comparing different schemes is left mainly to the instructor: hardly any book gives a satisfactory treatment of this matter.
Most of the available books are aimed at the community of system developers, rather than at students in the classroom. As a result, the books adopt an unconventional style of presentation: this seems natural as most of the authors are serious workers in the field of solid modelling who have helped in advancing this technology. However, when it comes to teaching the discipline to a wider spectrum of students, a book needs to be based on different considerations. Someone who has learned about solid modelling while teaching it to students might be in a better position to appreciate the difficulties. Such a person would be a good candidate to write a new solid modelling textbook.
Industrial participation
Unfortunately, Indian industry is lagging behind the rest of the world in using solid modelling. Indeed, very few companies have adopted design automation. Consequently, participation from industry is practically insignificant in academic courses on solid modelling. There are a few honourable exceptions to this, but industry is detached, more or less, from teaching in this field.
Certain industries have used solid modelling for years now, but the outcome of that use has not been published. There is no feedback from these industries into academic decisions about what topics should be covered and what emphasis should be placed on each topic; these matters are mainly determined by academics alone. Maybe this has resulted in poor, or even useless, courses. The situation could be improved if case studies in the use of solid modelling in industry are made publically available. Recently, many industries have shown interest in using solid modelling, and the situation is likely to change very soon.
Conclusions
Solid modelling is recognized as the core topic of computerized shape handling and it is an essential course in the academic curriculum. However, in the current scenario, there are major difficulties in the supply of software, trained manpower, study material, and in industrial participation. Only the hardware position is comfortable. This is the Indian scenario today, but it probably coincides to a large extent with the world picture.
Acknowledgements
The support and facilities provided by Computervision Research and Development (India) Pvt. Ltd. are gratefully acknowledged. Special thanks are due to Dr. S.P. Mudur for a number of valuable discussions.
Disclaimer
The views and opinions presented in this paper are the author's alone. They do not reflect any official policy or view, or the work done at Computervision Research and Development (India) Pvt. Ltd. or the Computervision Corporation, US.