a snapshot of the page taken as our search engine crawled the Web.
The web site itself may have changed. You can check the current page or check for previous versions at the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content.
Designing Effective Step-By-Step Assembly Instructions
Designing Effective Step-By-Step Assembly Instructions
Maneesh Agrawala Microsoft Research
Doantam Phan
Stanford University
Julie Heiser
Stanford University
John Haymaker
Stanford University
Jeff Klingner
Stanford University
Pat Hanrahan
Stanford University
Barbara Tversky
Stanford University
Abstract
We present design principles for creating effective assembly in-
structions and a system that is based on these principles. The prin-
ciples are drawn from cognitive psychology research which investi-
gated peoples conceptual models of assembly and effective meth-
ods to visually communicate assembly information. Our system is
inspired by earlier work in robotics on assembly planning and in vi-
sualization on automated presentation design. Although other sys-
tems have considered presentation and planning independently, we
believe it is necessary to address the two problems simultaneously
in order to create effective assembly instructions. We describe the
algorithmic techniques used to produce assembly instructions given
object geometry, orientation, and optional grouping and ordering
constraints on the objects parts. Our results demonstrate that it
is possible to produce aesthetically pleasing and easy to follow in-
structions for many everyday objects.
Keywords: Visualization, Assembly Instructions
1
Introduction
Many everyday products, such as furniture, appliances, and toys,
require assembly at home. Included with each product is a set of
instructions showing how to put it together [Mijksenaar and West-
endorp 1999]. For modular product lines, such as customizable of-
ce furniture, many different versions of the instructions are neces-
sary. As the number of customizable products and demand for task-
specic instructions increase, technology will be needed to produce
instructions more cost effectively. Already there is a high incidence
of poorly designed and out of date instructions.
The problem is that it is difcult and expensive to design as-
sembly instructions that are easy to understand and follow. Since
the instruction design process has not been systematized, skilled
human designers are needed to produce good instructions. As a re-
sult, the process of producing instructions is time-consuming and
labor-intensive. Computer support is currently limited to replac-
ing low-level tools such as pen and paper. Most high-level design
decisions are still made by human designers.
We have developed a system that provides higher-level tools for
designing assembly instructions. Figure 1 depicts instructions pro-
duced with our system. A broader goal of our work is to understand
how humans produce and use visual instructions. By codifying this
design knowledge in computer programs, we can make it easier to maneesh@graphics.stanford.edu
1
2
3
4
5
6
Figure 1: Assembly instructions for a TV stand. Our system plans the set of assembly
operations to show in each diagram and then renders action diagrams which explicitly
depict the operations required to attach each part.
produce clear drawings of 3D objects and more effective instruc-
tions [Tversky et al. Submitted].
The two primary tasks in designing assembly instructions are:
Planning: Most objects can be assembled in a variety of
ways. The challenge is to choose a sequence of assembly op-
erations that will be easy for users to understand and follow.
Presentation: There are many ways to depict assembly op-
erations. The challenge is to convey the assembly operations
clearly in a series of diagrams.
These tasks have been independently studied in the areas of
robotics and visualization. Assembly planning is a classic problem
in robotics [Wolter 1989; de Mello and Sanderson 1991; Wilson
1992; Romney et al. 1995]. Given the geometry of each part in the
assembly, an assembly planner computes all geometrically feasible
sequences of assembly operations. These plans are used by robotic
machine tools for automated manufacturing and are not meant to be
seen, understood, or carried out by humans. Most robotic assembly
plans would seem unnatural to people assembling everyday objects.
In contrast, automated presentation design systems have been de-
veloped in the domain of visualization [Feiner 1985; Mackinlay
1986], with the goal of producing diagrams that are easy for hu-
mans to understand. These systems assume that the information to
be portrayed is given as input and automatically design an effective
diagram to convey that information. Although some of these au-
tomated presentation systems have been developed to illustrate 3D objects and actions [Seligmann and Feiner 1991; Rist et al. 1994;
Butz 1997; Strothotte 1998], their primary focus has been on show-
ing the locations or physical properties of parts.
Our approach is inspired by a combination of ideas from these
previous systems. However, we believe that decisions involved in
planning and presentation are strongly intertwined. Therefore both
issues must be considered simultaneously.
The contributions of our work include:
Cognitive design principles for effective assembly instructions:
We performed cognitive psychology experiments to identify how
people conceive of the assembly process and to characterize the
properties of well-designed instructions. Based on the results of
these experiments and prior cognitive psychology research, we
identify design principles for effective assembly instructions. These
principles connect peoples conceptual model of the assembly task
to the visual representation of that task.
A system instantiating these design principles:
Our assembly
instruction design system consists of two parts: a planner and a
presenter. The planner searches the space of feasible assembly se-
quences to nd one that best matches the cognitive design princi-
ples. To do this the planner must also consider many aspects of
presentation. The presenter then renders a diagram for each step of
the assembly sequence generated by the planner. The presenter also
uses the design principles to determine where to place parts, guide-
lines and arrows. In particular, the presenter can generate action
diagrams which use the conventions of exploded views to clearly
depict the parts and operation required in each assembly step.
2
Design Principles for Assembly Instructions
Before we can develop automated tools for designing assembly in-
structions, we must understand how people think about and com-
municate the process of assembling an object. Cognitive psycholo-
gists have developed a variety of techniques to investigate how peo-
ple mentally represent ideas and concepts. We recently performed
human subject experiments based on these techniques to determine
the mental representations underlying assembly [Heiser and Tver-
sky 2002]. We briey describe our experimental setup.
In the rst experiment, we asked participants to assemble a TV
stand, given only a photograph of the completed stand as a guide.
After they assembled the TV stand, we asked them to create a set
of instructions that would show another person how to assemble it.
Examples of the diagrams they drew are shown in Figure 2. In the
second experiment, we asked a new group of participants to rank
the effectiveness of a subset of the instructions produced in the
rst experiment. Finally, the third experiment tested whether the
highly ranked instructions were more effective. Yet another group
of participants used instructions ranked in the second experiment to
assemble the TV stand, while experimenters recorded task comple-
tion time and error rates. We found that in general the highly rated
instructions were easier to understand and follow. Participants spent
less time assembling the TV stand and made fewer errors.
Based on these experiments, as well as earlier cognitive research,
we identify a set of design principles for creating assembly instruc-
tions that are easy to understand and follow.
Hierarchy and grouping of parts:
People think of assemblies as
a hierarchy of parts. At the base level, parts are segmented by per-
ceptual salience indexed by contour discontinuity; that is, parts that
are disjoint are more likely to be segmented. Typically, the dis-
joint parts are also grouped by different functions (e.g. the legs of
a chair or the drawers of a desk) [Tversky and Hemenway 1984].
When possible, people prefer that parts within a group are added
to the assembly at the same time, or in sequence one after another.
The part groups are usually considered as hierarchical structures,
which parallel the subassembly structure of the object.
Structural Diagram
Action Diagram
Figure 2: Hand-drawn assembly diagrams for the TV stand. The action diagram is
preferable to the structural diagram because it depicts the operations required to attach
each part. In this case the action diagram shows how the shelf is fastened by the screws.
Hierarchy of operations:
People think of the attachment