Overall,
the results of this experiment show that web based mapping in VRML is possible.
In particular, the data support the use of interactive mapping as opposed
to a single forced static view of a map to increase map user’s estimation
accuracy. Although the data do not
clearly support pre-programmed or user controlled interaction, there is support
for the use of pre-set viewpoints within the VRML world.
Furthermore, the results do not support the use of a map in this
environment draped onto a 3D terrain. Finally,
the data indicate that though symbol rotation does not significantly worsen the
ability to read the symbols, there is support in favor of stationary symbols.
This study provides insight into user interaction with maps in a web based 3D dynamic environment. In particular, this study confirms that there is great potential in such mapping, which will hopefully be explored and expanded upon.
The results of this study suggest that, in a dynamic interactive 3D viewing environment, all available features may not be necessary. This applies particularly to 3D features such as the 3D terrain maps, which proved to be an obstacle to the participants. The flat maps provided more accurate estimates of cylinder height. Past experience with conventional flat maps may also contribute to such higher accuracy. Regarding animated features of the environment, the data suggest that using symbol rotation as a visual variable does not significantly obstruct the use of symbol height as a visual variable. Nonetheless, review of the data suggest that stationary cylinders provided more accurate estimates. The varying cylinder heights did not produce significantly different estimates, although the cylinders of height 50 tended to be less accurate.
The ability to interact with the maps was shown to be a particularly significant factor in this study due to the more accurate responses in the pre-programmed and participant controlled worlds. With even minimal exposure to the viewing software, map users will take advantage of the navigation tools, and may even prefer to use them rather than watch a pre-programmed navigation. Nonetheless, they do prefer to use pre-programmed options which they may choose themselves in the self-controlled environment. This was evident with the reliance on the pre-set viewpoints for making the estimates.
User reaction is also a key factor concerning participant interaction with these worlds. Participant frustration with the pre-programmed fly-by worlds and the desire to control the map rotation in those worlds enforce the finding in favor of user control over the navigation for accurate estimates. A world which combined the features of the pre-programmed and participant controlled worlds, may, in fact, be the most desirable environment according to these results. For example, this could include a circular fly-by panel which has a start/stop button, direction of rotation button, and a rate of rotation button so that the user may exercise total control over the map rotation. There may also be more pre-set viewpoints made available. Such pre-programmed options may reduce the time used for self navigation.
To
obtain higher accuracy in estimates with navigational ability, there was a cost
in time as the time required to make the estimates was longest in the
participant-controlled worlds. If
time is a serious factor to the map user, participant controlled interaction may
not be as desirable as pre-programmed interaction or an initial view showing all
important information. However, the
combination of the participant preference for control over the pre-programmed
fly-by, along with the participant reliance on the pre-set viewpoints, suggests
that a combination of environments, as discussed above, may reduce the time for
study of the map.
There was one main difficulty with the experiment that provided an
undesired challenge to running it effectively.
The computer used was not fast enough to display the dynamic features of
the VRML worlds smoothly. This was
first visible in the amount of time that was required to load each world.
As a result, the transition from world to world in the experiment created
a 15 to 30 second waiting period before each world was fully loaded. More
significantly, the rotations were not smooth.
The cylinder rotation was slightly jerky, which may have effected the
estimation of those cylinders. This
did alter the original concept of the experiment which was to also test ability
to estimate cylinder rotation. More
noticeably, the circular fly-by world with the 3D terrain map did not rotate
smoothly, moving one step at a time, intermittently.
Another
problem with processing power concerned the ability of the participant to
control the navigation within the 3D terrain map.
To successfully make a movement, the participant had to wait while the
computer registered the movement and then redrew the scene a few seconds later.
This was due to the large size of the VRML ElevationGrid used to program
the terrain of the map.
Based on both user reaction and user estimates, this study concludes that for dynamic web based mapping in VRML, the best scenario would be to use a flat map with a combination of navigation controls accessible including those in the VRML browser and those designed by the maker of the VRML world. This would allow the map user to take advantage of both pre-programmed and participant controlled environments.
There are several potential changes to the design of this experiment which could benefit future research on this topic. Most noticeably, the ability to use a computer with greater processing power would be a major improvement. This would also allow the study to expand upon itself with the ability to request estimates of rotation measurements. It would be interesting to study whether rotation could be used as a continuous variable with different rates, or if it is more effective as a classed variable.
Another follow-up would be to focus on the pre-programmed fly-by worlds, examining different rates of map rotation. The ability to control the fly-by may also be tested to see, for instance, if the users make their estimates while the view is static or rotating, at what rate they set the rotation, or how long they watch the rotation.
Yet another possibility for a future study would be to conduct a comparison between different groups of people and their ability to make the estimates. In particular, it would be intriguing to look at high school students who have grown up not only with computers, but playing highly interactive 3D games, yet with little map knowledge.
Finally, another potential modification would be to use actual data. Comparisons could then be made between using the local area where the study is being conducted to another location where the participants would not be so familiar with the demographic data. Actual data would also demonstrate the real use of such a mapping concept. There is still much to be studied on the potential of map users interacting with VRML maps.