Way-Finding Study Proposal
(First of Two or Three: to be followed with a collaborative study, perhaps another with more explicit landmarks)
Jump To a particular section:
Purposes
Phase I: Driver Training
Phase II: Guided Tours
The 4 Conditions
Phase III: Independent Travel
Expectations
Data Collected
Discussion of Issues
Time Outline
Implementation Tasks
PURPOSES:
- To evaluate the flying interface of the FMVR system: learning, feedback, steering.
- To determine how well we internalize a 3D data space or terrain.
- To explore way-finding in a non-immersive virtual environment with 6 degrees of freedom.
PHASES:
I - Driver Training
This first phase will allow the subjects to gain experience with the flying interface. The purpose of this phase is to ensure that subjects are competent in the FMVR environment before starting the later tasks. Interesting results may be obtained from this phase by itself. It will consist of 3 sections: In-class, Practice, and Examination.
1. The In-Class Section
There will be a brief introduction to the system. This will explain the individual widgets of the flying interface, how to steer (adjusting your trajectory), and the related concept of how to accelerate (adjusting your velocity). This will be a scripted introduction, to be delivered either verbally or on paper.
2. The Practice Section
The subjects will then be provided a fixed amount of time to become familiar with the flying interface. They will be placed in an environment which encourages the exploration of all degrees of movement, such as a drastically different landscape from the one to be used in later phases (different in character, colour, size, or other), or a 3-dimensional object to examine. Suggested tasks will be provided (eg. view the object from directly above), but the activities of the practice section will not be monitored or analyzed in any detail.
3. The Examination Section
The subjects will proceed to the Driver's Test. They will be given a task to complete, in a landscape that is drastically different from the one to be used in later phases. For example, they may be given a flat landscape, with objects dispersed along a particular path through the terrain. Their task might be to steer to the first object, and drive through it. Upon successful completion, the next object in the path appears, at a different location and height, and so on to the end of the path. This will show their ability to steer themselves through the virtual environment. Measures that can be taken include:
- Time of traversal and its related velocity information
- Deviation from the "ideal path"
- Steering adjustments
Upon successful completion of the examination in a reasonable amount of time and along a reasonably efficient path, the subjects will advance to the final two sections. If the subjects fail to achieve reasonable success in this section, we can do one of two things:
- End their involvement with the experiment
- Permit a second chance, starting them afresh from the Driver Training phase.
II - Guided Tours
Phase II of the experiment will take the subjects to the experiment landscape for the first time. The purpose of this phase is to have them gain exposure to a given virtual environment, for them to internalize it. This phase will also introduce the controlled variable, as each subject is assigned to one of four conditions.
Common to all subjects, regardless of condition, will be a landscape with several pre-defined Departure and Arrival stations. The subjects will be guided along a set of paths through the terrain, each one from a specific Departure station, and to a particular Arrival station. For example, there may be 5 Departure stations (Red, Green, Blue, Grey, and Yellow) and 5 Arrival stations (same colour set, but light in colour, whereas the Departure stations will be dark). The subjects will know how many of each station there are, from a scripted scenario describing this phase. They will also know which stations they are traveling between, either from the visible representation on the screen, or from a text message of some sort. The set of paths will include intersections and some commonalities, and the subjects will learn the paths from observing the surrounding landscape (implicit landmarks) and the heading information (where provided). They will be guided by a "pilot car" which travels in front of them along the path. Perhaps each path on the guided tour should be repeated. (Q: Should the order of paths remain constant between subjects? or be randomly presented?)
Subjects will be randomly assigned to one of four conditions for this phase:
- Back-Seat Passenger
In this condition, the subject is "towed" along at a fixed distance behind the pilot car. As in the back of a chauffeured limousine, the passenger may passively watch the terrain go by, without steering and without the benefit of the heading clues provided by the flying widgets.
- Front-Seat Passenger
As in the previous condition, the subject is passive, doing no steering. They are towed at the fixed distance behind the pilot car. They may observe the passing scenery, but will also have the flying interface widgets as additional heading information. This is akin to taking the tour in the front seat, where you may observe the terrain and what the driver is doing to travel through it.
- Small Autonomy
This condition requires active following on the part of the subjects. The pilot car will travel along its pre-determined path, and the subjects must guide their vehicle behind it. They will have total motion control within a certain but small volume behind the pilot car. If they travel outside this volume, they will essentially be getting off the path they are to follow. When this occurs, we will apply sinusoidal clipping to the component of their velocity taking them off the approved path, down to zero at a pre-determined maximum deviation. This condition can generate data, regarding how well the subjects remain in the "ideal" zone behind the pilot car, and how often and severely velocity clipping must be applied.
This clipping is designed to keep subjects from wandering too far left or right of the path they are to learn, or from getting too high above the surface, or from sinking below the surface. It can also apply to the distance from the pilot car; as the subjects get too close and risk overtaking the pilot car, the clipping will gradually reduce their velocity in this direction until it matches the leader at some minimum distance. Likewise as they fall too far behind the pilot car, either their speed can be adjusted up to match it, or (preferable)its speed can be gradually slowed to match the subjects.
- Large Autonomy
This condition is the same as number 3, except the volume in which the subject has full autonomy is larger. It too can generate data of how well subjects followed in the "ideal" zone.
III - Independent Travel
In the final phase, we will be testing the subjects for their knowledge of the paths they have traveled, and their understanding of the entire environment. This will be accomplished by giving them two sets of tasks: (1) Travel along a selection of the paths that they were shown, known paths between Departure and Arrival stations; and (2) Find their way between a Departure-station:Arrival-station pair between which they have NOT been shown any paths. They will have multiple experiences with both stations in each case, but will have never been guided from one to the other.
Expectations are that the extra noise of condition 2 vs condition 1 will result in a greater ability to navigate through the terrain, but may detract slightly from understanding and internalizing it. It is also anticipated that the smaller autonomy will produce a greater understanding of the terrain. Conditions 3 and 4 are expected to produce better traversal times, as the subjects will have experience in steering along the paths.
How well the space is internalized will be assessed by teaching a set of paths through the space (by a follow-the-leader technique) to the subjects, and having them navigate the paths solo later. A close match between the intended and traveled paths will suggest an internalization of the data space.
This raises the issue of how to differentiate between the following:
- familiarity with the interface
- internalization of the paths
- internalization of the data space
Familiarity with the interface will be judged based on a period of experimentation, followed by some trials to navigate to specific points. This is related to some work done by Ware and Fleet, in which user familiarity with aspects of the FledermausVR interface were judged by having them navigate up to and inside boxes in the terrain, to identify an enclosed letter. See the Driver Training phase for more.
Familiarity with the paths will be evaluated by having the subjects navigate along a path that was shown to them in the training phase. Successfully navigating such a path would indicate a potential understanding of the data space, and a definite understanding of the paths.
Familiarity with the space, separate from the paths, can be tested by having the subjects navigate between start and end points with which they are already familiar, but for which they have never been shown a path.
The following data will be collected during each task trial:
- time to traverse Path from Departure to Arrival stations. (Normalize or remove individual variation by dividing this time by the average trial time for this subject, or perhaps by some value determined in their benchmark exercises for demonstrating interface abilities.)
- % of actual (known) path followed
- exact path traversed, with regular speed, height, and location samples
- # reversals of direction
DISCUSSION:
I would welcome any additions, changes, comments and feedback on the above. There are some outstanding issues to be discussed or decided, which this section indicates. They include:
- When velocity scaling is activated to prevent the subjects from wandering outside the constrained zone, should significant feedback of this scaling be provided? The visible direction of travel will change, and may not match the settings of the flying widgets (unless we change them as we affect the velocity). Is this sufficient feedback, or should there be additional indications, in the form of the sky changing colour, or a beep, or something else?
- Related to this is the special case of maximum and minimum height above the terrain. For the purposes of evaluating the subjects' understanding of the terrain from their perspective within it, it is necessary to limit their ability to obtain a bird's-eye view of the entire map. Thus, we will set a limit on their height above the terrain. An explanation of this restriction may be that it is practice guiding an underwater submersible, on the ocean floor, where popping up and down is not an option. It is also important that they be prevented from passing through the ground level. In these cases, is it not as important that feedback be given to the user, to indicate the automatic clipping of their velocity and heading?
- Initially, we discussed this as the presence/absence of landmarks, but since there are implicit landmarks in peaks, valleys, and ridges, it is still open for debate as to whether special landmarks should be identified. This issue be resolved in this study by relying on implicit landmarks in the terrain, and perhaps a future study will deal with the issue of explicitly identified landmarks.
- Navigation aids for this experiment will include the direction (heading) and velocity information provided by the widgets themselves, and is only available in Condition 2 (Condition 1 tests the results without the heading information) Similar information may be given with one or more of the following:
- compass readout, to identify heading, oriented around an arbitrary North
- mini map, 2D top view of terrain (may give too much information?)
- user-depositable markers for feedback of where I have been.
- The FledermausVR environment is set up for collaborative usage, but this study, with an automated pilot car, will be entirely individual. It precludes collaborative studies, with the guide being automated in this case to ensure a similarity across all subjects. It is easy to imagine this as an expert leader giving lead-by-example instructions (no words of explanation in this study).
An estimated outline of the test session would resemble:
Minutes Activity
====================
5 Driver Training: In-Class (introduction to the system)
10 Driver Training: Practice
5 Driver Training: Examination (benchmark of competency in interface)
5 -- break (and assess results)
15 Guided Tour: following pilot car in one of 4 conditions
5 -- break
10 Independent Travel: subject follows known paths in random order
5 -- break
20 Independent Travel: subject way-finds (unknown paths, known points)
total time: 1 hour 20 minutes (assume 0.5 minute per known path, makes 15-20 paths; assume 1.0 minute for unknown paths, makes 15-20 paths).
IMPLEMENTATION TASKS
This study will require several changes to the Fledermaus system. Please advise if any are left off the list. At a first glance, these will include the following tasks to be done, as implementation changes or data preparation:
- An object for the subjects to follow, capable of following a defined path
- A unique landscape or 3D object for the Driver Training phase
- A flat landscape for the Driver Examination, with objects of varying height through which the subjects must drive, and code that will make the next one appear only when the current one has been traversed.
- A unique landscape for the Guided Tour and Independent Travel phases.
- Arrival and Departure station objects that can be attached to this landscape. The program must notice when the subject has entered one of these stations, and stop the trial accordingly.
- A set of paths between the Arrival and Departure stations.
- The ability to cycle through the set of paths to follow with no more than a mouse click from the subject.
- Code to time the trials, and write positional and other subject trial information to disk.
- Sinusiodal velocity clipping based on deviation from some acceptable volume behind the pilot car.
- The ability to set the parameters for this clipping, both its severity, and the amount of autonomy (freedom of movement and velocity) before the clipping begins. These parameters should be settable from the command line and within the environment (menu options). Probably should be settable from the terminal of the system "expert" as a future useful feature.
- A method to prevent subjects from seeing the flying widgets (condition 1) and allowing them to watch but not influence them (condition 2). (is this non-trivial?)
- A tool for defining paths to be used in the study
- A tool to visualize the paths taken by the subjects (?)
- A method of analyzing paths the subjects took, against expected paths.