Computer Aided Concept Design for Vehicle Stylists

Dr Samantha Porter

VIDE
Coventry School of Art and Design Coventry University

This piece of research was carried out on grant number GR/H5171 2 from the Control, Design and Production Group of the EPSRC.

It is commonly accepted in the manufacturing industries that the earlier concurrent design and engineering are able to occur the greater the reduction in both design lead time and the cost of the design product cycle is reduced. A basic requirement of concurrent design and engineering is a simple 3 dimensional (3D) computer model. This piece of research was concerned with the production of 3D models from the traditionally 2D concept design stage of the design cycle with a focus on vehicle styling/design.

The main aims of the project were:

• To establish procedures for computer-aided concept generation and design evaluation for vehicle stylists.
• To devise and perform experiments for evaluating various combinations of concept generation and evaluation techniques.

To allow these aims to be achieved d was first necessary to examine the relationship between concept creation and concept evaluation. During each state of concept design, the designer will evaluate what has been created and iteratively develop it. At the end of each stage, there are larger-scale evaluations when the design team or management assess the proposals. For this reason in considering concept creation it was necessary to also consider its evaluation.

Previous research (Tovey,1991) showed that a concept design method should allow a designer to view a design proposal as a three-dimensional entity as early in the process as possible, and that the traditional lateness of the move into 3D is based on cost and difficulty. New computer-assisted methods may overcome these constraints. The research has concentrated on methods which give fast, early, development to 3D, and the validity of this approach has been demonstrated through collaboration with vehicle designers.

The context of sketching in concept creation

The industrial designer has two particular functions:

• To visualise the product concept;
• To represent alternative design solutions.

The purpose of sketching a major and continuing part of the design process is to help designers to arrive quickly and easily at a stage where a satisfactory design can be specified for development for manufacture. Representing a proposal visually is a fundamental part of industrial design. The representations are conventionally 2D drawings or 3D physical models, depending on when in the design process they are to be used. We characterise the presentation requirements by considering the management intervention points. Automotive styling design can be considered as two consecutive processes:

• concept design, which conventionally uses theme sketches and dimensionally-constrained sketches;
• design development, which conventionally uses tape drawings, scale models, and full-sized models.

The theme sketch is the initial expression of how a proposed design is intended to look For such sketches to be useful in the design process there must be enough visual information in them for an informed observer to:

• understand what they imply about the design proposal
• understand the visual impact of the implied design as a whole
• understand its specific visual characteristics.

The package representation of a proposed design is a collection of visual cues sufficient to communicate the fixed dimensional parameters of a design. We similarly define a package-constrained sketch as one using a package representation to constrain and guide the designer's sketching. It allows an observer to:

• understand the dimensional constraints and proportions of the proposal
• understand the fit or interference of the theme proposal wnh the package.

For a developed system to be truly useful in the design process it was necessary that the designers' sketches could be used directly as a source of evaluable and developable models. Those models, produced almost as rapidly as the sketches themselves, would allow for 3D evaluation of a concept much earlier in the process than is tradnbnally possible. The availability of a 3D model would also allow much earlier access to other disciplines and allow the possibilRy of very early concurrent engineering.

Specification of design requirements

Any concept design system must accommodate both the needs of stylists and those of management.

Stylists' needs

The requirements of designers using compuners have been analysed by Robert Aish (1990) and this list has been modified to apply more particularly to concept sketching.

These modified criteria are:

rapidity: the speed at which the concept can be represented;
versatility: the range of tools and techniques available;
modifiability: the ability to modify a representation once created;
visibility: the degree to which the representation can be assessed, both in itself and in relation to other representations.

Form creation for concept design must have the following characteristics (Tovey, 1992):

• Inputting geometric information must be quick
  
• Geometry must be easy to specify
  
• Representations must not be overly precise
  
• Geometry must be easy to modify
  
• Reviewing alternatives must be readily possible
  
• Visual display quality must be adequate for evaluation
  
• The overall interface must be user friendly.
  

Management's needs

The management needs have been defined previously (Birtley, 1990), and are as follows:

• that the presentation technique facilitate the design process
• that it allow others to participate in the process (such as design managers, other members of a
• design team or the client); and
• that it communicate the completed design proposal to others; this will be to the client,

Once both stylists and management needs had been defined it was possible to investigate the possibilities for Computer-Aided Concept Design System.

Identification of Practical Methods for Concept Design Experiment.

The initial approach was an identification of practical methods for concept design.

The framework shown below was developed to enable a formal exhaustive analysis of designers working methods.

The matrix reduces the concept design process to a series of five abstract categories with a number of alternative parameters. The previously gathered working knowledge of design practice, summarised below, allowed the identification from the total possible combinations from the matrix a number of abstract categorised methods which accommodated the requirements of designers in concept design.

In a series of concept design sketches, the designers will:

• Usually work in both Orthographic and Perspective categories of view.
• Usually work in both Freehand and Infommed categories of context
• Within the technique aspect. Area on its own is unlikely to support anything other than
• Simple intelligences.
• In a developmental series of representations, be working in an environment of 2D, 2.5D or 2D
• on 3D at the start, and would want to progress into, or be working in some semblance of 3D at the finish.
• Intelligence will always include Simple, and may include Information-bearing.

The nature of the styling process means that sketching will always be a primary activity and a number of practical investigations were carried out in three areas using existing hardware and software packages.

• 2D sketch related techniques
• 3D sketch related tecniques
• matching and combining sketches

Practical Investigations

Equipment

The materials used in all the experiments were: Silicon Graphics Indigo and Indigo2 workstations running Alias Studio, Showcase, and Movie software; an Apple Macintosh running Photoshop; and a Canon scanner. Where subjects were required MA Automotive Design students were used.

2D sketch related techniques

A number of possible methods were examined and the following conclusions drawn:

The investigations involving the Automotive Design students where subjects were required, a number of possible methods:

(i) Paint programs can be used for concept sketching even if the interfaces are not ideal.
(ii) 3D modelling packages can also be used for concept sketching though it was not possible to paint onto a modelled surface.
(iii)Paper sketches can be satisfactorily scanned in as images to paint.
(iv) Geometry can be extracted from such freehand sketches although not completely
(v) It is useful to use underlays when sketching freehand to ensure geometric accuracy
(vi) With limited practice it was possible to use a computer based system for productive sketching though the differences of technique required often led to a different sketching style.

3D sketch related techniques

As previously stated to allow concurrent engineering a 3D model is a requirement. One of the limitations of traditional sketching is that it only produces two-dimensional representations of three dimensional concepts. It was established that no method was available to sketch non-planer lines (this would create automatically a 3-D model).

This was done by using geometry-extraction methods to create 2D profile lines from sets of 2D orthographic sketches, and then using a combination of extrusion and projection techniques to produce the implied 3D base models. An attempt was also made to produce styling-line models from perspective sketches, as styling-line models represent concept forms with greater fidelity than the orthographic base models.

The investigations established a number of possible solutions:

(i) Orthographic sketches of a concept can be imported to a modeller and placed in the correct relationship to one another. It is more efficient to extract the geometry from sketches and position that, rather than the sketches themselves. If the geometries are mutually correct, as is the the case when assisted with the use of underlays, the implied 3D linecan be constructed from them. In this way simple models can be produced from sketches in 20 minutes, though these models are fairly crude.
(ii) No existing way of constructing 3D styling line models from sets of perspective sketches was found. A possible method was explored but was found to be too slow to use practically without software support (and none of the packages examined gave this). The method was validated by tahng a 3D model, projecting it to make pseudo-sketches and then reconstructing the model from those. The resulting model was a good geometric match.

Matching and Combining sketches

The base models, while representing the design concept in their geometries, look nothing like sketches. It is already established that the tentative appearance of a sketch is an important factor for concept designers, and therefore ways were sought of adorning the base models with sketch-like surfaces. This set of studies examined several ways in which both olthographic and perspective sketches could be applied to simple surfaced models, such as the base models.

The following procedures were established:

(i) Mapping orthographic sketches onto 3D objects proved possible if solid textures were used and gave good results if solid textures were used.
(ii) Sketches, even with underlays, are inexact, because stylists tend to draw what looks correct, even if it is not strictly accurate, making it hard to extract accurate data from complete sketches. This is as conteracted by exploiting the midline symmetry of the car. This had the added, and unexpected benefit of effectively trebling the producible models by using each half, mirrored, as well as the original sketch. The sketch models produced in this way generally had the approval of the designers who had made the sketches.
(iii) The procedure for matching perspective sketches to models is similar to that for orthographic sketches, except that view matching is more difficult. This problem was solved by using templates derived from perspective views of the base models, so that projection characteristics were known and an inverse perspective mapping to apply the sketches to the object. A problem with peripheral distortion when mapping perspective sketches was discovered exploiting the symmetry techniques that had been developed before solved this problem. This technique gave models that looked 'good' and had designer acceptability. These practical investigations lead to three techniques for building models from sets of either orthographic or perspective sketches. These are described below.

Proposed Computer-Aided Concept Design method

The method uses the work designers currently perform at this stage in the design process, essentially sketching and uses the orthographic and perspective sketches to develop 3D sketches for initial evaluation and then developing the styling line model (identified in the CAVS research as the start for an engineering model).

• Produce and import orthographic sketches to a 3D modelling system;
• Relate the sketches to appropriate views in the system;
• Extract line information from the sketches and develop into surfaces;
• Assemble the surfaces into a model and map the sketches onto the surfaces;
• View and evaluate the rendered model.

Sketch mapping is the process by which line information is taken from orthographic sketches and used to create a simple surFace model, which is then combined with those orthographic sketches to give the first stage sketch model wah the implied feeling and detail of the orthographic sketches and the additional spatial attributes of a three-dimensional model. As in the conventional process, theme sketching will generally be followed by package-related sketching. The progression from the sketch to the model involves the following:

It is possible to move from a theme concept to a package-constrained concept, by either scaling the sketches to match the package constraints as 2D sketches before making a model, or by scaling or otherwise modifying the model to match the package constraints after creation.

• Using the model from the sketch mapping stage produce underlay views
• Sketch modifications using the views as underlays; and import them to the 3D modelling system;
• Relate the sketches to appropriate views in the system and map the sketches onto the appropriate surfaces;
• view and evaluate the rendered model.

Sketch projection is the second process where the designer uses the simple surface model created in the sketch mapping process as an underlay for perspective sketching. these sketches are then combined with the model to create the second stage sketch model for further evaluation.

• Take the perspective sketches already created and matched to views;
• Extract the line data from them
• Combine 2D line inforrnation into 3D styling lines;
• Create surfaces from tbe styling lines;
• Map the sketches onto the appropriate surfaces;
• View and evaluate the model.

Sketch combination uses perspective sketches 10 create styling-line CAD models which can be surfaced later. Assuming the resuits are satisfactory in design terrns the designer then uses Sketch combination where line information is taken from the perspective sketches created for the perspective projection process, this line information is then used to create a stling line model. These computer assisted methods will support both theme and package-related aspects of concept sketching, with the proviso that the resulting computer models are dependent on the quality of the sketches; a loose free theme sketch can only provide a loose free-theme sketch model. If surfaces are subsequently created, the sketch projection technique can then be used on the model to develop details.

The three dimensional models needed to be evaluated for their usefulness/ acceptability to designers as a visualisation and evaluation tool needed to be investigated. In order to do this a multi-media presentation of simulations of the techniques was developed and tested as both a pilot study and three case studies.

Pilot Trial

Introduction

Individual designers and managers were invited to take part in the pilot study which was conducted in the VIDE Visualisation laboratory. They were given an introduction and general background to the research, followed by the presentation data were collected by a combination of structured interviews during the presentation and completion of the checklists by the subjects. The checklist following areas for each of the techniques:

• Original sketch match to model
• Understanding of model
• Ability to make design judgments about the model
• Use as a design tool
• Use as an evaluation tool
• Speed of the approach

Feedback was positive about all aspects and the checklist was refined for use in the case studies.

Case Studies

Introduction

Three case studies were undertaken wah Rover (an automotive manufacturing company), Styling International and Krafthaus two design consultancies). The two texture mapping related techniques were investigated. We hoped that the case studies would give a greater insight into the effectiveness of the procedure, since designers would be able to input their own 2D sketches, about which they would have a better understanding of the expected 3D outcome.

Aims

To examine designer reaction to working with the procedure developed for creating 3D concept forms on a computer. Although the procedure for creating the forms was the same in each case, the method and background of each case study differed.

Rover

Four designers working in a manufacturing company with no prior knowledge of the procedure evaluated the sketch projection technique using perspective sketches.

Styling International techniques.

A designer working in an automotive design consultancy with a good background knowledge of the procedure tried both the sketch mapping and projection Krafthaus

As above, but the designer only evaluated the sketch projection technique.

Materials

The materials used in all the experiments were: Silicon Graphics Indigo and Indigo workstations running Alias Studio, Showcase, and Movie software: an Apple Macintosh running Photoshop; and a Canon scanner.

Method

The designers were asked to make perspective concept sketches over underlays, except that in one case (Styling International) the designer was also asked to produce freehand orthographic concept sketches. In all cases the designers completed their tasks. These were transformed to 3D sketches and the designers evaluated the results. The appropriate parts of the checklist developed earlier were used for the collection of information.

Resuits

All the designers who produced sketches were shown them combined wah the underlay model. Resuits were generally favourable. The 3D models appeared much as the designer had anticipated and in all cases felt as if they had gained a better understanding of their own concept designs and were able to make subsequent design decisions based on the model. One other major benefa was highlighted. Putting concept designs into 3D at this vew early stage in design allowed the designers to communicate their 3D ideas to other members of the development team e.g. engineers allowing them 3D comprehension of the idea.

The success of the case studies allowed a Computer aided Design to be specified.

Specification of a CACD system

Each of the three concept design techniques requires certain capabilities from the system in which it is to be performed. While many CAD programs each provide some of these capabilities, no one system of which we know provides them all; indeed, in the case of inverse perspective projection, none provides it.

Working environment within the 3D modelling component

The modelling component must support simuHaneous multiple perspective or orthographic views. Management of multiple windows on the screen at the same time will involve referencing the views with a facility to retain those references, and to retum to them at will, as well as the ability to lock a view to stop it being changed by accident. The user must be able to work in any window and be able to perform any of the actions necessary to the process in that window.

Image Creation

The system should allow sketches to be created externally and imported to the system, or to be created within the system, and it must allow transfer of the information to 3D programme.

Image transfer and the 3D world

The modelling component must allow the importing of pictures into both orthographic and perspective views. Once a sketch is imported there must be a facility to support matching the sketch to an appropriate view in the 3D world. There must also be a facility for converting the picture to line information, for example by manually tracing the sketch or by having the computer automatically trace the sketch.

Image manipulation

Since a concept will be represented by many sketches, there must be a facilRy to match these muRiple perspective sketches together; in other words the user must be able to scale the imported sketches and to position them correctly in the 3D world.

3D Development

The modelling component must support parallel projection of line information to combine 2D lines into 3D lines, as well as parallel projection onto a surface. Once created, the surface lines must be usable to trim the surfaces and to create new ones, as well as to support fillets on the trimmed intersections of surfaces.

It must also support perspective projection using the view vector and projection angle to create the projection characteristics, and the use of this projection to allow a similar combination of information, as above

Texture mapping

Our proposals require that a sketch can be mapped onto a model surface which involves treating a sketch as a cross-section of an extrusion. This extrusion method must be supported by both orthographic and perspective proiection techniques in the modeller.

References

Aish, R. 1990 CAD Soflware Design to Augment the Creation of Form. YRM ArchRects and Planners

Birtley, N.. 1990 The Conventional Automobile Styling Process. Internal Report Coventry University

Tovey, M. 1992 Intuitive and Objective Processes in Automotive Design. Design Studies,13(1)

Tovey, M. 1991 Computer Aided Vehicle Styling. Final Report grant number GR/F 68065 from the Control, Design and Production Group of the EPSRC.