Technical Report Virtual Reality Tracking System. John P. Baker, Andrew P. Paplinski, Member, IEEE. November 4, 1994

Transcription

1 Faculty of Computing and Information Technology Department of Robotics and Digital Technology Technical Report Virtual Reality Tracking System John P. Baker, Andrew P. Paplinski, Member, IEEE November 4, 1994 Enquiries:- Technical Report Coordinator Robotics and Digital Technology Monash University Clayton VIC 3168 Australia

2 Abstract Tracking systems are used in virtual reality applications where the orientation and/or the position of a real physical object is required. A prototyped tracking system based on ultrasonic sensors using triangulation techniques will be discussed, as will the problems associated with existing tracking systems. Key words: Virtual Reality, Tracking Systems, Headtracking. 1

3 Contents 1 The Author 3 2 Introduction 3 3 Tracking Systems available Polhemus magnetic AC tracker : : : : : : : : : : : : : : : : : : : : : : : : Ascensions mechanical boom : : : : : : : : : : : : : : : : : : : : : : : : : : Ascensions magnetic DC tracker : : : : : : : : : : : : : : : : : : : : : : : : Shooting Star Technology arm based position tracker : : : : : : : : : : : : Logitechs ultrasonic tracker - Red Baron : : : : : : : : : : : : : : : : : : : VRN mechanical tracker : : : : : : : : : : : : : : : : : : : : : : : : : : : : Summary : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 6 4 A New Tracking System Specication : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Implementing the System : : : : : : : : : : : : : : : : : : : : : : : : : : : 6 5 Ultrasonic sensors Ultrasonic Receiver Hardware : : : : : : : : : : : : : : : : : : : : : : : : : Ultrasonic Transmitter Hardware : : : : : : : : : : : : : : : : : : : : : : : 11 6 The controller specication Controller Hardware : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Controller Software : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 13 7 Results Filtering : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 17 8 Conclusion 17 2

4 1 The Author The rst computer I ever programmed was the TRS-80 Colour Computer 2 with 32kB of RAM and 32kB ROM. I started programming in Motorola BASIC at the age of 12 and my rst publication was at the age of 16, a drawing package written in BASIC supporting speech output it was called Superdraw [1]. Two months later I had two more publications, the rst one was a security program written in the 6809 Assembly language [2], and the next program made use of the fact that programs for the TRS-80 were saved to a cassette tape and the cassette tape could be controlled by the computer. Being fond of the rock group Dire Straits I wrote a graphical demo that controlled the cassette tape to play the music in time with the demo. The whole thing was automated so the program was saved rst on the tape then the music. The program was titled In Concert [3]. My fourth publication one month later was a puzzle program called Aliens [4]. I progressed up to the Amiga 500 computer and started programming in Motorola Assembly Language. In 1988 I then started programming in C. I began the Bachelor of Computing(Digital Technology) in 1989 and immediately began programming in Pascal, Fortran and dipping back to 6801 assembly. In 1992 in a group of 2, I constructed and programmed a Micro-mouse and participated in the IEEE November Micro-Mouse competition being placed rst in Australia and awarded the IEEE Best Oz-Mouse prize. In 1993 I achieved class 1 Honours and thus began my interest in Virtual Reality where I designed and prototyped a HMD 1 System, and tracking system for the HMD. In addition, 3D Sound sourcing software and a 3D graphics engine were also programmed on an Amiga 1200 computer. 2 Introduction Immersive Virtual Reality requires a computer generated environment to react to a users movements. In the simplest form this requires the ability to track a users head for orientation and possibly position. This adds realism to the virtual reality system. An improvement to this situation would be the ability to track the entire users body orientation and position realtime with a high degree of accuracy. With current technology it is only realistic to track multiple point objects rather than a person's body. The need for a good tracking system in Virtual Reality exists to make the virtual experience more believable. Several factors are involved in designing a tracking system. Position requirements Orientation requirements The number of samples per second the system can provide. Accuracy Reliability Cost 1 Head Mounted Display 3

5 Current o the shelf products are readily available but they all have problems in one or more of the areas outlined above. Section(3) will describe the currently available tracking systems provided by commercial companies and the problems associated with these systems. Section(4) will describe a new tracking system design, the prototype built and the results of this tracking system. 3 Tracking Systems available The following commercially available tracking systems have been investigated in this report: The Polhemus magnetic AC tracker. Ascensions mechanical boom tracker. Ascensions magnetic DC tracker. Shooting Star Technology arm based position tracker. Logitech ultrasonic tracker. VRN mechanical boom tracker. 3.1 Polhemus magnetic AC tracker Polhemus Navigation Sciences currently provide magnetic based tracking systems. Sensors are used to measure magnetic interference in three dimensions. Any coil conducting with an electrical current generates an electromagnetic eld. The eld is strong in the direction of the coil's radius, and it is relatively weak in the perpendicular direction. Similarly, a magnetic eld passing through a coil of wire generates an electric current proportional to the eld's strength [5]. The tracker uses a transmitter with three coils of wire, each perpendicular to the other two. A similar receiver has the same arrangement. The tracker's controller pulses each of the transmitter's coils in turn and reads the current generated in each of the three receiving coils, for a total of nine readings. This allows one to determine the receiver's orientation and distance. Knowing that the strongest readings come from the coils that lie on the same plane as the transmitter, a microprocessor can determine the orientation of the receiver in space (relative to the transmitter), as well as the distance in X, Y and Z directions. This system has a relative position resolution of 3mm to 4mm and 0.5 degrees within a 1 metre radius from the transmitter. The receiver is a small, lightweight plastic cube, about 1cm square. The transmitter, is a slightly larger cube. Both receiver and transmitter connect to a control unit that handles the pulsing and sensing. The control unit can be connected to a host computer via a serial or parallel port. (See Figure 1). 4

6 Controller Receiver Transmitter Figure 1: Polhemus tracker conguration 3.2 Ascensions mechanical boom Ascensions mechanical boom is used to support, and track the movement of a boom mounted viewer. There are six potentiometers on the boom. The potentiometers analog values are converted to digital values and available via a serial port for calculation of the orientation and position. 3.3 Ascensions magnetic DC tracker Otherwise known as ock of birds is used for tracking multiple targets. Utilising a pulsed DC magnetic eld eect. Distortion generated by eddy-currents are drastically reduced using the DC system [6]. 3.4 Shooting Star Technology arm based position tracker This a boom mount type tracker with sensors at each joint for determing position and orientation. This tracker is used specically to track the head. 3.5 Logitechs ultrasonic tracker - Red Baron Uses ultrasonic line of sight to provide 3D 6DOF 2 tracking system. This is used to track an object such as the head or hand. 3.6 VRN mechanical tracker This an inexpensive boom mount type tracker for determining position and orientation. It is primarily used for tracking the head. 2 Degrees of freedom 5

7 3.7 Summary Problems exist with each of the trackers described. The Polhemus tracker is very expensive, prone to servere noise and made use of electromagnetic waves close to the human operator. It is known that the transmitter must be placed at least 2 metres away from any cables and computer monitors to avoid interference. The unit cannot be used in environments surrounded by metallic objects. The same problems exist with the Ascension magnetic tracker where noise from eddycurrents and local magnetic interference from surrounding objects are a problem. Boom trackers are too restrictive and oer only one device to be tracked. For some applications they may be of use where the operator needs to look away from the display system. The Logitech ultrasonic method is similar to the prototyped system described in section(4.1). These systems were developed using a similar method to the Mattel Powerglove [5]. Another type of system that has been used is infrared sensing, although there are problems with line of sight and movement restrictions, and the amount of CPU processing required. A commercial unit currently does not exist. Several universities have made prototyped systems. 4 A New Tracking System 4.1 Specication A new system needed to be designed to t the following specications and satisfy the following criteria. Orientation yaw allow the user a certain amount of movement reliable fast operation Aordable - Low cost Light Future expansion for orientation pitch and roll Future expansion for position X, Y, Z. Future support for multiple input devices 4.2 Implementing the System The method used involves triangulation in determining the orientation of the user. Triangulation works on the following principles. Assume two sensors located in the 2-D plane at point S A and S B, see gure(2). given, 6

8 X 2 SA da P db SB 0 X 1 S A = (S A1 ; S A2 ) S B = (S B1 ; S B2 ) P = (X 1 ; X 2 ) Figure 2: 2D - Plane The position of a point P is to be determined from the distances d A and d B of the point P from the sensors S A and S B. The point P must be located at the intersection of two circles: Subtracting, we have (X 1? S A1 ) 2 + (X 2? S A2 ) 2 = d 2 A (1) (X 1? S B1 ) 2 + (X 2? S B2 ) 2 = d 2 B (2) (X 1? S A1? X 1 + S B1 )(X 1? S A1 + X 1? S B1 ) thus, + (X 2? S A2? X 2 + S B2 )(X 2? S A2 + X 2? S B2 ) = d 2 A? d2 B (3) (S B1? S A1 )(2X 1? S A1? S B1 ) + (S B2? S A2 )(2X 1? S A1? S B1 ) = d 2 A? d2 B (4) Assume that the sensors are parallel to the X 1 axes, ie., S A2 = S B2, see gure(3). then (S B1? S A1 )(2X 1? (S A1 + S B1 )) = d 2 A? d2 B (5) giving, 2X 1 = d2 A? d2 B S B1? S A1 + S A1 + S B1 (6) X 1 = 1 d 2? A d2 B + S 1M (7) 2 S 1L Similarly for a pair of sensors parallel to X 2 axes, see gure(4), 7

9 X 2 P db da S1L 0 SA1 S1M SB 1 X 1 Figure 3: 2D Plane S 2M X 2 S 2L S C db dc P da S B S A 0 X 1 Figure 4: 2D Plane X 2 = 1 2 d 2 C? d2 B S 2L + S 2M (8) Now the point P(x,y) has been determined. A second point Q is introduced and its position is also determined from equations (7) and (8). We now have two points P(X 1,X 2 ) and Q(X 1,X 2 ). (See Figure 5). To determine the orientation about one axis requires the following calculation. X 2 S C P Alpha Q S B S A 0 X 1 Figure 5: Triangulation = arctan P X2? Q X2 P X1? Q X2 (9) 8

10 The resultant angle, yaw, is between -180 degrees and 180 degrees. 5 Ultrasonic sensors Ultrasonic waves occur above the human hearing range, greater than 20kHz. Transducers operating in this range transmit ultrasonic waves at the speed of sound in air. Hence using a transmitter and receiver the time taken for a wave to travel can be determined. Two types of transducers are available in this range, 25kHz and 40kHz transmitter and receiver pairs, their periods being 40us and 25us respectively. Hence the wavelength can be calculated from(10) giving, and = V s f s (10) 40 = 316: = 7:92 10?3 metres 25 = 316: = 13:00 10?3 metres Knowing that the velocity of sound in air is m/s and allowing for an approximate range of 3 metres between the transmitters and receivers will enable approximately 100 samples per second to be recorded for maximum throughput. The rst trial was to connect each of the transmitters up to a function generator and see the receiving signal on a oscilloscope. Firstly the 40kHz transducers were tried. A sinusoidal signal was generated at 40kHz and the received signal was observed on the oscilloscope. The signal received was in the 0.01mV range. The transducer was found to be very directional. The input voltage at the transmitter was approximately 14V. The second 25kHz transducer pair was then tested. This pair was found to have a very small range. The voltage was also about the same in the 0.01mV range. The 25kHz signal has a period of 40us which is nearly double the period of the 40kHz signal of 25us. It was decided that with the increased range and the shorter period that the 40kHz transducer pair would be used. There are two pieces of information available using the ultrasonic transducers. Distance Magnitude It was decided that both pieces of information would be made available by the receiver hardware. 5.1 Ultrasonic Receiver Hardware The initial specication required that the receiver be capable of generating a CMOS level signal when detecting a 40kHz signal. It was also required that an output be available for Analogue to Digital Conversion of the magnitude. The receivers are comprised of a 9

11 Filter AC amplier AC to DC converter Comparator Voltage level adjustment See Figure 6 for the receiver block diagram. When a received signal reaches the receiver sensor the voltage is in the range of 0.01mV. The rst task was to construct an AC Amplier incorporating a band pass lter. Much of the problem was in AC amplication of the very weak signal. A Darlington pair of BC547 transistors are used. A resistor capacitor pair act as a band pass lter with cuto frequencies of 38kHz and 42kHz. The next stage was to convert the signal to a DC level. This is achieved using a diode to clamp the signal and a capacitor resistor pair to smooth the waveform. The DC level signal is then fed to a voltage level adjustment and to a comparator. The voltage level adjustment is an output for an Analogue to Digital Converter. The comparator is used to generate a CMOS level output when the DC level exceeds the threshold for a detected single. The threshold may be adjusted with the potentiometer provided in the circuit. Signal in Amp 38-42KHz (1) Ultrasonic Receiver Volts Freq Band pass Filter Ac Amplifier (1) Time Ac to Dc Converter Voltage Converter Comparator Output 1 Output 2 To Input Capture To Analogue Digital Converter Figure 6: Receiver block diagram 10

12 5.2 Ultrasonic Transmitter Hardware The transmitter circuit uses a transistor that is voltage switched from ground to the positive rail. A controller is connected to the base of the transistor and must provide the correct timing to produce a 40kHz output pulse. The transmitter sensor will work in a very narrow frequency band around 40kHz 1kHz. The optimum value being within 100Hz. See Figure 7 and 9. Magnitude 0 Hz 39Khz 41Khz Frequency Logarithmic Hz Figure 7: Ultrasonic Transmitter Spectrum 6 The controller specication To coordinate the transmitters and receivers required a controller board external to a host computer. A MCU 3 must be able to communicate with a host computer and control the transmitting and receiving of signals to/from the sensors. The tracking system can then be used with any host computer system without the need for new low level software to be written. 6.1 Controller Hardware The controller is responsible for controlling the transmitters, and processing signals from the receiver units. The processor chosen had to support a system that allowed inputcapturing of external signals. The capture had to record the time between transmitting and receiving of a signal. Many microcontrollers support this system. The two microcontrollers that provide the good functionality are the 14MHZ MC68HC16 3 Micro Controller Unit 11

13 2MHZ MC68HC11 Both these MCU support realtime interrupts [7], [8], [9] from multiple external sources. The MC68HC16 supported up to eight external sources and the MC68HC11 supported four. The MC68HC16 is the better processor and would have been chosen except for the fact that the MC68HC16 was available only as surface mount and not many software tools were available during the development stage. The MC68HC11 provided all the functionality necessary for implementation and also a wide range of software tools. The MC68HC11A8P1 is the actual MCU being used. This MCU contains only 256 bytes of RAM hence external RAM was required. This MCU is congured for Expanded Mode of operation which allowed RAM and EPROM to be added. The MCU has two PORTs which are used for controlling the transmitters and connecting signals from the receivers. The block diagram is shown in Figure 8. To Transmitters From Receivers BUFFER From I/O Port CPU 68HC811A8 BUFFER To Input Captures To A/D Serial Port Host Computer Data Bus RAM 68HC24 EPROM Latch Clk Data To/From Powerglove Figure 8: Controller block diagram Transmitter control lines are connected to output pins of the MCU. These lines can be changed by controlling the register assigned to them. The receivers time signals are connected to the input capture pins of the MCU. The receiver analogue signals are connected to the Analogue to Digital converter input pins on the MCU. The MC68HC24 shown 12

14 in Figure 8 is used for controlling other devices such as a Mattel Powerglove. This is a replacement PORT for the MCU's PORTS B and C when in expanded mode. Both the tracking system and powerglove can operate from this board. The hardware is now in place for receiving a signal from a transmitter and calculating the distance and magnitude. 6.2 Controller Software The controller software is used to control the transmitter and calculate the time and magnitude of the signal received. First the current internal MCU time is recorded. Then a 40kHz pulse is generated using instruction loops. The instruction loops were used as a timed interrupt was taking too long on the 2MHz 68HC11. The instruction loop is calculated exactly from the fact that 1 E clock = 0:5 us Hence the instructions can be added up to give the exact time frame required before changing the state of the transmitter pin from low to high or high to low. A burst at 40kHz is required to allow the ultrasonic transmitter to reach its maximum output value. A burst of 10 cycles was found to reach this limit. Extra cycles did not increase the strength of the signal. If less than 5 cycles are used, the transmitter does not reach maximum strength and the range is dramatically reduced. When the pulse is received by the ultrasonic receiver a high signal is received by the MCU which causes an input capture interrupt. The current MCU timer value is recorded by the interrupt routine. See Figures 9 and 10. The distance for the signal to travel from the transmitter to the receiver can be calculated as T ran Distance = Rec timer? Start timer (11) Three transmitters are connected to the MCU and three receivers are connected to the MCU. Each transmitter is pulsed and three dierent interrupts are created at three dierent times, depending on the location of the receivers to the transmitter. This gives a total of nine timed readings. This allows the transmitter's position to be determined. See Figure 11. If the receivers are connected in a triangle and placed upon a user, as shown in Figure 12 the position of the user relative to the transmitter can be determined. If two transmitters are placed above the user, two points can be found and the orientation of the users head with respect to the transmitters can be established as described in Section( 4.2). This can be extended to three transmitters, so that the pitch and roll can also be calculated. Currently the system is used to calculate the roll only. This will be extended in the future to calculate pitch and roll. The host computer is currently used to do the calculation, all of the information is transmitted serially to the host computer which then calculates the angle. The MCU has a lot of time spent waiting for the signal to arrive. During this time constant Analogue to Digital Conversions take place on the waveform, recording the maximum input amplitude during the transmission. Currently no calculations are performed 13

15 Amplitude (Volts) 25us 12 V Transmitter 0 V Time (usecs) Amplitude (Volts) Receiver 5 V 0 V Time to reach receiver Time (usecs) Figure 9: Timing Trace 14

16 Turn on Input Capture Interrupts Input Capture Interrupt Record Timer Value Check record flag Send 40KHz Burst Record Time Wait Max Delay Return Check Serial Port Send Data to Host Figure 10: Flowchart while waiting. The system could be improved so that the angle for the previous transmission is calculated and transmitted serially to the host while waiting for the ultrasonic signal to arrive at the receiver. The current system consists of 7 PCB modules: 3 transmitter 3 receiver 1 controller 7 Results A host computer was then used to test the tracking systems. The Amiga 1200 computer was used and a 3D graphics skeleton was written. The rst test involved placing the transmitters on a small board in a triangle conguration and mounting the board to a prototyped HMD 4 system. The receivers were then placed in a triangle conguration and mounted to a large board and the board was then mounted above the user. This resulted in many problems, mainly the receivers did not receive a strong enough signal from the transmitters and hence no signal was detected. This was a line of sight problems, as the user tilted their head the receiver would never receive the transmitted signal. The previous problem was drastically reduced by simply swapping the positions of the sensors. The transmitters were placed above the user pointing to a central point where the 4 Head Mounted Display 15

17 RD R1 R2 RD R3 Receiver Type : 40kHz Ultrasonic TD T1 T2 TD T3 Transmitter type : 40kHz Ultrasonic Figure 11: Tracking System Layout - Top View Transmitter Board Travel Distance Receiver Board Helmet Figure 12: Tracking System Setup - Side View 16

18 user stands or sits under the transmitter and board and the receivers were placed on the small board mounted to the HMD. The receiver's almost always receive the transmitted signal in this conguration. The HMD must be under the transmitter board and between 0.7 metres to 1.5 metres distance between receiver and transmitter board. Now that the receivers were receiving a signal, the next problem was accuracy. The system tended to jump a lot, which was very nauseous when using the HMD. So a ltering system was developed. 7.1 Filtering A simple moving average lter was designed to remove the high frequency components from the tracking angle. It was also tested on the raw values however the lter worked best on the nal angle being ltered. A four tap lter was found to provide reasonable results where, n being the sample number Y n the current output sample the current input sample X n Generalising, Y n = 1 4 (X n + X n?1 + X n?2 + X n?3) (12) Y n = 0 X n + 1 X n?1 + 2 X n?2 + 3 X n?3? 1 Y n?1? 2 Y n?2? 3 Y n?3 (13) where, n being the sample number coecients of [ ] and coecients of [0:25 0:25 0:25 0:25] Y n the current output sample the current input sample X n A small lag time was created from the time the user moved their head to the time that the tracking system nally reached the destination. This could be improved by having more samples going through the lter per second and modifying the lter coecients. 8 Conclusion The tracking system developed achieved what it was designed to do. Limitations exist in where the tracker can be used. The tracker requires a clear line of sight between the transmitters and the receivers. The tracker is prone to 40kHz noise such as that generated when keys are rattled. The system could be extended to have multiple controller boards with multiple receivers so that more than one device may be tracked at a time. A faster MCU will not necessarily make the system faster. To increase performance of the system a faster Universal Asynchronous Receiver Transmitter (UART) could be used from a port with a large 17

19 First-In-First-Out (FIFO), so that all the data for transmission can be buered to this device and free the MCU for other operations. A next step would be to use a MCU with many more input captures allowing for more receivers. Noise is currently a problem, requiring a lter described earlier to remove it. A better system would encode the signal which is detected by the receiver rather than just a small burst. The receiver hardware could be improved to be more accurate. The transmitters could be boosted to give a larger output signal. The current system support has all the hardware and software necessary for generating the orientation about one axis and position in space X and Y. Improvements in the layout of the sensors would allow this information to be more accurate. The largest limitation as with all current tracking systems is the required reference point. A better system would be self contained, ie., no transmitter-receiver pair would be required. An idea that could be explored would be the use of accelerometers. The measured acceleration could be used to calculate the distance moved over time gaining the position of the device. If three of these are used for 3 dierent axis, the position of the accelerometer could be calculated with respect to its starting position in three dimensions. The solution may be found with the combination of tracking technologies, using the ultrasonic solution combined with the accelerometer system. References [1] J. P. Baker, \Superdraw," Australian COCO, vol. 3, pp. 47{49, August [2] J. P. Baker, \Auto exec and password," Australian COCO, vol. 4, pp. 19,24, October [3] J. P. Baker, \In concert," Softgold, pp. 27{31, October [4] J. P. Baker, \Aliens," Softgold, pp. 21{23, November [5] H. Eglowstein, \Reach out and touch your data," Byte, pp. 230{290, July [6] A. T. Corporation, The Flock of Birds [7] Motorola, ed., MC68HC11 Reference Manual [8] Motorola, ed., MC68HC11A8 Technical Data [9] Motorola, ed., MC68HC11E2 Technical Data