| |
32
Chapter 3. Pilot Studies: Lessons Learned from Early Projects
main part of the full assembly plan. The graph consists of icons representing partial
states of the wardrobe and interconnecting lines (to be read from left to right).
The lines describe actions that need to be completed to move from one state the
next. Actions always consist of joining a previously unused board to the depicted
compound depicted or of joining the two compounds. The actions of preparing the
boards (adding dowels, cams, or screws) are not shown in the graph. The only
restriction for those actions is that they have to be completed before the board is
used. We had to distinguish between the situations of connecting the side board a
and the one of connecting side board b to the horizontal board. This is essential
since the orientation of the boards is important.
The dashed lines in the graph symbolize actions in which the user has to lift up one of
the compound parts before adding the extra board or joining the other compound.
This is due to the restriction that the compound has to be lifted before the two
horizontal parts are added to one sideboard (as it cannot be set upright anymore
otherwise). States from which one cannot continue are marked as dead-end states
with a lightning symbol. If the user reaches such a state he has to go one step back
before he can continue.
The plan o ers a total of 44 possible paths, and shows another 14 routes which lead
to dead-end states. The 4 routes which are marked with thick lines are the ones
proposed in the original assembly instructions by the manufacturer. This clearly
shows that the manufacturer’s instructions are limited to a subset of all possible
assembly paths.
For each step in time, the wardrobe is in one state of a column of Figure 3.2. An
action transfers the wardrobe to the next column on the right or to the end state.
Knowing with which state the user is occupied allows the system to give suggestions
about future steps (Section 3.1.3). The following section describes how the di erent
states can be recognized through sensors.
Perceiving User Actions with Sensors
The execution plan proposed in Figure 3.2 consists of actions and states. Table 3.1
lists the actions that have to be recognized in order to trace the full execution plan.
Various actions have to be recognized, such as preparing the di erent parts, lifting
a compound part into an upright position, and joining a side board or the base strip
to a horizontal board.
Most of the actions in Table 3.1 can be subdivided into partial actions. These actions
are relatively simple and self-contained, such as tightening a screw, hammering in a
dowel or a nail, turning a board or joining two parts together. In the third column,
the table gives the sensor configuration we used in our prototype. The fourth column
|  |
|
| |
|
|
