Authors: Gregory Baratoff and Scott Blanksteen
Tracking devices allow a virtual reality system to monitor the positionand orientation of selected body parts of the user. Many interaction devices (see "Interaction Devices") incorporate a tracking device ofsome sort in order to measure the position and orientation of the bodypart they're attached to. In an HMD, the position and orientation ofthe head is measured. This information defines the user's viewpoint inthe virtual world, and determines which part of it should be renderedto the visual display, as well as influencing the generation ofacoustic stimuli. Attached to a glove, a tracking device measures theposition and orientation of the hand. Based on this information, a handcan be rendered in the virtual world at the same position with respectto the user, providing feedback that is often necessary for dextrousmanipulation.
Tracking devices, also called 6-degree-of-freedom (6-DOF) devices, workby measuring the position (x, y, and z coordinates), and theorientation (yaw, pitch, and roll) with respect to a reference pointor state. In terms of hardware, the following three components are ingeneral required : a source that generates a signal, a sensor thatreceives the signal, and a control box that processes the signal andcommunicates with the computer. Depending on the technology used, eitherthe source or the sensor is attached to the body, with the other placedat a fixed spot in the environment, serving as a reference point.
The usefulness of tracking devices in virtual environments depends toa large degree on whether the computer can track the movements of theuser fast enough to keep the virtual world synchronized with the user'sactions. This ability is determined by the lag, or latency, of thesignal, and the sensor's update rate. The signal lag is the delaybetween the change of the position and orientation of the target beingtracked and the report of the change to the computer. Lags above 50milliseconds are perceptible to the user and affect human performance.The update rate is the rate at which measurements are reported to thecomputer. Typical update rates are between 30 and 60 updates per second.
The precision with which actions can be executed in the virtual worlddepend on the resolution and accuracy of a tracking device used.Whereas the resolution is fixed for a given device, the accuracyusually decreases with the distance of the sensor from the source. The range of a tracking device is the maximum distance between sensorand source up to which the position and orientation can be measuredwith a specified accuracy.
Interference, or sensitivity to environmental factors, can limit theeffectiveness of tracking devices. Depending on the technology used,they can be sensitive to large metal objects, radiation from displaymonitors, various sounds, and objects coming between source and sensor. During the design of the physical environment, these factors should becarefully considered, so that the user doesn't have to stay aware ofthe properties of the physical environment while engaged in a task ina virtual environment. This is especially important in fully immersivevirtual environments where the outside view is completely blocked.
Most currently used tracking devices are active, in that the sensor,or sometimes the source, is attached to the target to be tracked.In passive tracking the target is monitored from a distance by one orseveral cameras. Although this approach is to be favored from the user'sperspective, at this point in time this is not a technologically viablesolution.
Current tracking devices are based on electromagnetic, acoustic,mechanical, or optical technology. A presentation of each of theseapproaches follows, together with a discussion of their advantagesand disadvantages.
Mechanical tracking devices
These devices measure position and orientation by using a directmechanical connection between a reference point and the target.Typically, a light-weight arm connects a control box to a headband,and encoders placed at the joints of the arm measure the change inposition and orientation with respect to the reference point.The lag for mechanical trackers is very short (less than 5msec),their update rate is fairly high (300 updates per second), and theyare accurate. Their main disadvantage is that the user's motion isconstrained by the mechanical arm. An example of such a mechanicaltracking device is the Boom developed by Fake Space Labs.Inertial tracking devices represent a different mechanical approach,relying on the principle of conservation of angular momentum. Thesetrackers use a couple of miniature gyroscopes to measure orientationchanges. If full 6-DOF tracking ability is required, they must besupplemented by some position tracking device. A gyroscope consistsof a rapidly spinning wheel suspended in a housing. The mechanicallaws cause the wheel to resist any change in orientation. This resistance can be measured, and converted into the yaw, pitch, androll values. Inertial tracking devices are fast and accurate, andsince they don't have a separate source, their range is only limitedby the length of the cable to the control box or computer. Theirmain disadvantage is the drift between actual and reported valuesthat is accumulated over time, and can be as high as 10 degres perminute.
Optical tracking devices
Most optical tracking devices currently used in virtual environmentsare for tracking head position and orientation. Basically, they comein two variants. In the first one, one or several cameras are mountedon top of the HMD, and a set of infrared LEDs is placed above thehead at fixed locations in the environment. In the alternative setup,the cameras are mounted on the ceiling, or a fixed frame, and a fewLEDs are placed at fixed and known positions on the HMD. In bothapproaches, the projections of the LEDs on the cameras image planescontain enough information to uniquely identify the position andorientation of the head. Various photogrammetric methods exist for computing this transformation.The optoelectronic ceiling tracker developed at the University ofNorth Carolina is an example of a head tracker. The system consiststhree cameras mounted on the HMD, and 1000 infrared LEDs placeduniformly across the ceiling. The computer pulses the LEDs sequentiallyand processes the images to detect the flashes. Based on the locationsof the flashes, the position and orientation of the head are calculated.The range of the optoelectronic ceiling tracker is limited only by thearea of the ceiling covered by LEDs, and is thus easily expandable.Its update rate is between 50 an 80 Hz, and the lag varies between 20and 80 ms. The resolution is 2 mm in position and 0.1 degrees inorientation. A problem with this tracker is that for some positionsof the head very few or none of the LEDs are visible to the cameras,leading to 'blind spots' of the tracker.
The alternative design solution is exemplified by the Honeywell LEDarray system. Four infrared LEDs, arranged in a prescribed pattern onthe HMD, are monitored by a camera mounted at a fixed position in theenvironment. As with the UNC system, the LEDs are pulsed one at a time,and the positions of the resulting flashes on the camera images togetherwith the known relations between the LEDs are used to compute the position and orientation of the head. A variant of this system avoidsthe use of active elements, that is the LEDs, on the HMD. A reflectingpattern is placed on the HMD, and is illuminated by an externalinfrared source to make it visible to the camera.
Optical trackers in general have high update rates, and sufficientlyshort lags. However, they suffer from the line of sight problem, in thatany obstacle between sensor and source seriously degrades the tracker'sperformance. Ambient light and infrared radiation also adversely affectoptical tracker performance. As a result, the environment must becarefully designed to eliminate as much as possible these causes ofuncertainty.
Electromagnetic Tracking Devices
Electromagnetic tracking devices function by measuring the strength ofthe magnetic fields generated by sending current through three smallwire coils, oriented perpendicular to one another. These three coilsare embedded in a small unit that is attached to whatever the systemneeds to track - typically, the user. The current has the effect ofmaking each wire into an electromagnet while the current is flowingthrough it. By sequentially activating each of the wires, andmeasuring the magnetic fields generated on each of three otherperpendicular wire coils, it is possible to determine the position andorientation of the sending unit.These tracking units may experience interference operating in thevicinity of CRTs or other devices that produce magnetic fields, aswell as metal objects, such as office furniture, that disrupt magneticfields. Another disadvantage to these tracking devices is that theworking volume tends to be rather small.
The most well-known producer of electromagnetic sensor technology isPolhemus. Their systems provide extremely low latency (on the order of5 milliseconds) and have the ability to track multiple objectsconcurrently.
Acoustic Tracking Devices
Acoustic tracking devices use ultrasonic (high-frequency) sound wavesfor measuring the position and orientation of the target object. Thereare two ways of doing this: so-called time-of-flight tracking andphase-coherence tracking.Time-of-flight tracking works by measuring the amount of time that ittakes for sound emitted by transmitters on the target to reach sensorslocated at fixed positions in the environment. The transmitters emitsounds at known times, and only one is active at a time. By measuringwhen the sounds arrive at the various sensors, the system candetermine the length of time it took for the sound to travel from thetarget to the sensors, and thereby calculate the distance from thetarget to each of the sensors. Since there will only be one pointinside the volume delimited by the sensors that satisfies theequations for all three distances, the position of the target can bedetermined. In order to find position, only one of the transmitters isneeded. Orientation is determined by the differences in locationindicated by these calculations for each of the three sensors.
Time-of-flight trackers typically suffer from a low update rate,brought about by the low speed of sound in air. Of course, anotherproblem is that the speed of sound in air is affected by suchenvironmental factors as temperature, barometric pressure, andhumidity.
Phase coherence tracking works by measuring the difference in phasebetween sound waves emitted by a transmitter on the target and thoseemitted by a transmitter at some reference point. The phase of a soundrepresents the position on the sound wave, and is measured in degrees:360 degrees is equivalent to one wavelength difference. This is clearif one thinks of a sound that is a pure sine wave. The graph of thesine and cosine describes a circle as the angle progresses from 0degrees to 360 degrees. After 360 degrees (one cycle, or wavelength),the graph returns to its starting point. As long as the distancetraveled by the target is less than one wavelength between updates,the system can update the position of the target. By using multipletransmitters, as with time-of-flight tracking, orientation can also bedetermined.
Since they work by periodic updates of position, rather than bymeasuring absolute position at each time step, phase-coherencetracking devices are subject to error accumulation over time.
Conclusion
Through our investigation of tracking devices, we see that manydifferent approaches have been tried, all of which have their ownadvantages and disadvantages. It is clear that the mouse and keyboardof virtual reality have yet to be discovered. All of the devices wehave described are good for some environments and tasks, and fail onothers, and, while we don't claim that the mouse and keyboard areperfect, they are certainly effective in a broad range of tasks, easyto use, not cumbersome, and inexpensive. Researchers are stillactively seeking the tracking device with a large working volume, highaccuracy and resolution, very short lag time and high update rate,that is convenient for the user. Until such devices are found, it willbe hard to achieve the virtual reality goal of having the computerdisappear.
