by Glenna A. Satalich
This study provided a surprising, counter-intuitive and perhaps important results. People who explored a virtual environment were compared to a control group who were given a map of the same environment but no direct experience of it. Subsequently, all subjects were asked to perform navigational tasks in the same virtual environment. By all measures used, people who had the virtual experience either performed equivalently or worse than the control group. This result certainly serves as a caution to enthusiasts (e.g. Rheingold, 1991) who appear to believe that virtual environments may be superior training environments.
But, there maybe explanations for the poor performance by those who explored the virtual environments. One possibility is that the subjects simply did not have enough time in the virtual environment. Subjects in this experiment experienced a virtual environment for 30 minutes. Navigation studies in the real world have shown that egocentric exploration of an environment, results in more accurate orientation and route estimations than those who have only seen a map of the area. This is opposite to the findings of this study. However, real world studies often contrasted map study to weeks or months of primary experience. Further, in at least some cases people can fail to gain Euclidean knowledge of an environment even after years of experience (Moeser, 1988). It is likely that repeated exposures to a virtual environment will increase subject proficiency to that of a control group (with 5 minutes of map study), but at what cost.
A more interesting possibility is that the VR experience was a distraction; not because VR is inherently a poor training medium, but because is was simply unfamiliar. The majority of subjects (50 or 57) were novices to VR, and therefore the interface problems of the hardware may have had a negative effect. A recent study by the University of Wales (1995) , where subjects were exposed to a virtual environment replicating a real environment, stated that subjects took four to six hours for there performance to equal subjects trained in the real environment.
Another explanation is that the virtual environment training was compromised by the intrusiveness of the VR interface. Subjects could not behave naturally. Removing all artifactual differences in the interface would approach an experience similar to that seen in the holodeck of the TV science fiction show Star Trek. Presumably training in a holodeck environment would be equivalent to training in a real environment but with the advantage of built in safeguards. But rather than dreaming of this fictional environment a more relevant question is how can we use the virtual interfaces the can build in the next ten years. As of today, we know that many factors can hinder the virtual experience, such as lag time between the user's action and the system's response, the weight, resolution, and field-of-view of an HMD, the weight of a wand and the unnaturalness of using the hand or button presses to initiate walking. All of these engineering and user-interface design problems are currently being acted upon,but it is unknown at what point they have to reach ,so that VR will not interfere with a user's natural behavior.
This study also examined the use of different tools and cues during exploration. The results indicate that in some cases a map during exploration may have interfered with learning. These results are consistent with the argument that any tool that attracts attention to itself may interfere with the learning it is supposed to facilitate. Clearly, we need, virtual environment tools that concentrate attention upon the development of an appropriate Euclidean representation. For instance, it might help to modify the map to show position and orientation. An intriguing tool has been demonstrated by Randy Pausch (Stoakley, Conway and Pausch, 1995) that may improve navigational awareness and wayfinding in virtual environments. This tool produces 3-D miniature image of the virtual environment occupied by a user, which is held in a virtual hand. With the other hand the user holds an image of a body (representing the user). As the body image is positioned in the miniature environment the user is passively transported to the new spot. At present the tool transports the user through walls causing some disorientation, but with some modifications the shortest route through virtual doorways could be achieved.
The paths used in these experiments could also be modified. One possibility is to elevate the path above the floor. In this experiment the path was placed on the floor as it would appear in a real world environment. In the active guided condition this placement caused the participants to look down constantly, making it difficult to scan the environment around them. This was not the case for those who were passively guided as they did not have to constantly monitor the line. With so many options for tools and cues that can be implemented in virtual environments, the challenge is to find a tool that equals or surpasses performance in the real world for navigational awareness and wayfinding.
The most important conclusion of this thesis is that there are differences or artifacts between virtual and real environments that affect performance in simple navigation and wayfinding tasks. This begs the question of what these differences are and how might the quality of virtual interface hardware and environments be changed to improve the propensity of virtual technologies to train real world tasks. Further research is needed to ascertain these artifacts. The navigation and wayfinding task, and the metrics presented in this thesis may be useful tools for assessing the goodness of virtual interfaces and determining these artifacts. If virtual reality is to be used effectively for training spatial awareness in real world environments, these issues must ultimately be addressed and resolved.