IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
9
Real-Time Noncoherent UWB Positioning Radar With Millimeter Range Accuracy: Theory and Experiment Cemin Zhang, Michael J. Kuhn, Student Member, IEEE, Brandon C. Merkl, Member, IEEE, Aly E. Fathy, Fellow, IEEE, and Mohamed R. Mahfouz, Senior Member, IEEE
Abstract—In this paper, we propose a novel architecture for ultra-wideband (UWB) positioning systems, which combines the architectures of carrier-based UWB systems and traditional energy detection-based UWB systems. By implementing the novel architecture, we have successfully developed a standalone noncoherent system for positioning both static and dynamic targets in an indoor environment with approximately 2 and 5 mm of 3-D accuracy, respectively. The results are considered a great milestone in developing such technology. 1-D and 3-D experiments have been carried out and validated using an optical reference system, which provides better than 0.3-mm 3-D accuracy. This type of indoor high-accuracy wireless localization system has many unique applications including robot control, surgical navigation, sensitive material monitoring, and asset tracking.
TABLE I COMPARISON OF CURRENT RESEARCH AND COMMERCIAL HIGH-ACCURACY POSITIONING SYSTEMS
Index Terms—Localization, noncoherent, positioning, ranging, ultra-wideband (UWB).
I. INTRODUCTION S THE technologies utilized in businesses, hospitals, and manufacturing facilities have become more advanced, a pronounced need within the RF identification (RFID) market (total projected revenue in 2009 of 5.56 billion USD) has developed for high accuracy, indoor, reliable, real-time location information solutions to track people, assets, etc. [1], [2]. In many cases, knowing the location of resources/assets can be the difference between success and failure and can have serious effects in applications such as automated bomb detection/disablement and real-time 3-D tracking for computer-assisted surgery [3]. Therefore, there is a great demand to develop wireless local positioning technologies as they have many diverse applications and have received considerable attention [4]. While global positioning systems (GPSs) use ultra-high-precision atomic clocks to measure the time-of-flight, a more standard method for indoor localization systems is the use of time difference of arrival
A
Manuscript received May 08, 2009; revised September 26, 2009. First published November 20, 2009; current version published January 13, 2010. C. Zhang is with the Hittite Microwave Corporation, Chelmsford, MA 01824 USA (e-mail:
[email protected]). M. J. Kuhn and M. R. Mahfouz are with the Mechanical, Aerospace, and Biomedical Engineering Department, The University of Tennessee, Knoxville, TN 37996 USA. (e-mail:
[email protected];
[email protected]). B. C. Merkl is with Medtronic Navigation, Louisville, CO 80027 USA (e-mail:
[email protected]). A. E. Fathy is with the Electrical Engineering and Computer Science Department, The University of Tennessee, Knoxville, TN 37996 USA (e-mail:
[email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org Digital Object Identifier 10.1109/TMTT.2009.2035945
(TDOA), where all of the base stations or receivers are synchronized, and the difference in time is measured between each pair of receivers to triangulate the position of an unsynchronized tag. Competing technologies for high accuracy indoor positioning include frequency modulated continuous wave (FMCW), impulse-based (i.e., carrier-free) ultra-wideband (UWB), and carrier-based UWB. Table I provides a summary of the various research groups and commercial systems utilizing these three approaches for high-accuracy indoor positioning. Similar accuracy levels (0.5–20 cm) have been achieved for both carrier-based [5]–[8] and impulse-based [9]–[12] UWB positioning systems, although carrier-based systems have shown the potential for millimeter and sub-millimeter range accuracy even for 3-D indoor environments (this study and [6]–[8]). FMCW has proven to be a successful competing technology for high-accuracy positioning systems [13]–[19]. In the 5.8-GHz band for industrial, scientific, and medical (ISM) applications,
0018-9480/$26.00 © 2009 IEEE
10
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
Fig. 1. ED-based UWB receiver architecture. Fig. 2. Carrier-based noncoherent UWB receiver architecture.
documented accuracy of 5–20 cm for 2-D has been achieved [14]–[18]. FMCW systems operating at higher frequencies including 35 and 77 GHz have achieved accuracy levels of 0.1 mm [13], [19] with the system described by Feger et al. working at ranges of up to 10 m [19]. The most recent FMCW trend is a European-wide push to create low-power wireless sensor networks built on 5.8-GHz FMCW technology [15]–[17]. It should be noted that in Table I many of the reported error are standard deviation error, whereas our reported 3-D error are root mean square error (RMSE) referred to an optical system, which provides better than 0.3-mm 3-D accuracy. RMSE is a good measure of error resulting from both the accuracy and precision, i.e., the true unbiased error when data values fluctuate above and below zero. UWB is a promising technology for use in short-range indoor local positioning and wireless data communications. It is well known that very high spatial resolution can be achieved using UWB due to the wideband nature of the signals. A widely used and low complexity UWB architecture is energy detection (ED), where the UWB signal is transmitted directly through the UWB antenna without up-conversion [10], [20]–[25]. Meanwhile, at the receiver side, the ED of the signal is achieved by passing the signal through a square-law device, usually a Schottky or tunnel diode, followed by an integrator and sampler. Fig. 1 shows the typical UWB receiver architecture using ED. However, due to the large bandwidth of the received UWB signal, it is difficult and costly for ED-based receivers to operate at above the Nyquist rate [20]. Typically a fast comparator is used as the sampling device. For example, Buchegger et al. realized an UWB communication link with a data rate of 1.2 Mb/s using the tunnel diode detector [21]. Lie et al. realized the leading-edge pulse detection method using a tunnel diode combined with an envelop detector to maintain high accuracy for UWB ranging in a multipath environment [22]. Recently, Fujii et al. achieved a 0.3-ns time resolution using the ED-based impulse detector, corresponding to 0.1-m distance resolution [10]. Another widely used UWB architecture is carrier-based impulse radio (IR) UWB, where the transmitted UWB signal is up-converted through a microwave carrier, then down-converted at the receiver side [26]–[28]. Fig. 2 shows the carrier-based low-complexity UWB receiver architecture. In [26], the demodulator is implemented with a 2-FSK scheme, whereas in [27], the demodulator is realized through a low-cost two-channel sub-sampling mixer with an equivalent sampling rate of over 100 GS/s [29], followed by a digital signal-processing unit implemented through a standard field programmable gate array (FPGA). However, a problem exists in the amplitude and phase differences between the in-phase (I) and quadrature (Q) channels due to hardware variation between the I and Q receiver chains, which causes distortion of the demodulated signal. Recently, Treyer et al. [30] corrected the amplitude and
Fig. 3. Noncoherent UWB receiver architecture using FLL.
Fig. 4. Proposed UWB receiver architecture which combines the carrier- and ED-based UWB receiver schemes.
phase error by using the Hartley phasing-type single-sideband modulator, but this method requires relatively complex circuitry and is limited to narrowband applications. One type of noncoherent UWB receiver architecture was implemented in [8] by McEwan, as shown in Fig. 3. Upon receiving the RF bursts from the transmitter, a timing circuitry called frequency-locked loop (FLL) locks the receiver pulse repetition frequency (PRF) clock to the transmitted PRF clock so that both clocks have the same frequency and phase. However, the required receiver timing circuitry design is quite complex and the system is limited to operate at a relatively low carrier frequency to avoid using the expensive samplers with high sampling bandwidth. This paper presents analysis and development of a novel UWB receiver architecture for a low-complexity noncoherent real-time UWB localization system. As shown in Fig. 4, the proposed UWB receiver architecture combines the carrier-based and ED-based UWB receiver schemes. The down-conversion requires only one channel instead of I and Q channels, as compared to Fig. 2, which lowers system complexity and overall cost. In this study, a noncoherent real-time localization system in an indoor environment has been developed. The developed system has been built on our previous work [27], [28] and is based on transmitting and receiving picosecond pulses and then carrying out a complete narrow-pulse, signal detection, and processing scheme in the time domain. The challenges in developing such a system include: generating UWB pulses, pulse dispersion due to antennas, modeling of complex propagation channels with severe multipath effects, need for extremely high sampling rates for digital processing, synchronization between the tag and receivers’ clocks, clock jitter, local oscillator (LO) phase noise, frequency offset between the tag
ZHANG et al.: REAL-TIME NONCOHERENT UWB POSITIONING RADAR WITH MILLIMETER RANGE ACCURACY
and receivers’ LOs, and antenna phase center variation. For such a high precision system with millimeter or even sub-millimeter accuracy, all these effects should be accounted for and minimized. Many of these effects were addressed in [27]; however, the reported 3-D localization results were based on utilizing a Tektronix TDS8200 oscilloscope and the system was coherent, i.e., the transmitter and receiver were wired. A comprehensive simulation framework has also been utilized in quantifying the accuracy of the system in realistic multipath indoor environments in terms of the overall sensitivity to the mentioned challenges in achieving high accuracy [31], [32]. Some of the recent measurement results were reported in [33], which, for the first time, we demonstrated millimeter-range dynamic accuracy with 1-D and 3-D experiments based on a noncoherent real-time UWB system. This paper is a substantial extension of [33], which focuses on revealing the fundamental theories behind the achieved accuracy. This paper is organized as follows. In Section II, the background of our previous positioning system is introduced including a block diagram of major system components, the potential issues of using I/Q down-conversion and the derogatory effects of high phase-noise carriers on overall 3-D accuracy. In Section III, the single-channel receiver approach is presented and how phase noise relates to overall system performance is examined. An inherent timing error source called the “shoulder” effect is also discussed with focus given to how the ED can minimize this “shoulder” effect. In Section IV, the 1-D experimental results with the proposed UWB receiver architecture are outlined and compared to our previously published results. In Section V, various real-time 3-D experiments are conducted including a tag moving randomly in a 3-D space and a tag attached to a robot arm with preplanned motion. Both dynamic and static results are reported using an optical tracking system with 3-D accuracy of 0.3 mm for reference. Finally, Section VI presents a conclusion. II. BACKGROUND The complete setup of our previously developed localization system is shown in Fig. 5 [27]. In the previously developed system, we transmit a modulated narrow Gaussian pulse with a carrier frequency and demodulate it at the receiver side. The source of our UWB positioning system is a step-recovery diode (SRD) based pulse generator with a pulsewidth of 300 ps and bandwidth of greater than 3 GHz. A detailed discussion of the pulse generator design can be found in [34]. The Gaussian pulse is up-converted with an 8-GHz carrier and then transmitted through an omni-directional UWB antenna. Multiple base stations are located at distinct positions to receive the modulated pulse signal. The received modulated Gaussian pulse at each base station first goes through a directional Vivaldi receiving antenna and then is amplified through a low-noise amplifier (LNA) and demodulated to obtain the I/Q signals. After going through a low-pass filter (LPF), the I/Q signals are sub-sampled using an UWB sub-sampling mixer, extending them to a larger time scale (i.e., s range) while maintaining the same pulse shape [29]. The PRF clocks are set to be 10 MHz with an offset frequency of 1–2 kHz between the tag and base stations, which
Fig. 5. Block diagram of localization system showing one tag and tions, which feed into the main system controller.
11
N base sta-
corresponds to an equivalent sampling rate of 50–100 GS/s. Finally, the extended I/Q signals are processed by a conventional analog to digital converter (ADC) and standard FPGA unit. According to the analyses in [27], the 300-ps Gaussian pulse could theoretically be recovered by the I/Q down-conversion at the receiver side. The and signals after I/Q downconverter become (1) (2) is the Gaussian pulse signal, is the carrier signal where is the offset frequency of relative to leakage factor, and from the tag. The filtered and the carrier generated by data are then sub-sampled and ac coupled, which are given by (3) (4) where is the pulse signal after time extension while maintaining the same pulse shape as and are extended is the equivalent offset I/Q signal from sub-sampler, and frequency after sub-sampling, which can be expressed as (5) where is an integer. The extended and signals are then processed by an FPGA, and the reconstructed received signal is given by (6) However, the above analysis is based on two assumptions, which are: 1) there is no phase difference between the I and Q channels and 2) the phase noise of both the tag and base station carrier are neglected, which leads to a fixed offset carrier without variation with time and temperature. In frequency practice, the utilized I/Q mixer has a phase difference of up to 4 between the I and Q channels. This is combined with unknown phase differences inherent to the designed sub-samplers
12
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
TABLE II PHASE-NOISE PERFORMANCE OF DIFFERENT CARRIER SOURCES
frequency after the sub-sampling process, which makes it extremely difficult to calibrate the unknown phase difference error. For illustration, we define a modulation factor to be bandwidth of
(7)
is the time extension where is the 10-dB bandwidth of the transmitted factor and UWB Gaussian pulse signal. For example, in Fig. 6(a), the and signal features a modulation factor . Considering the complexity and difficulties of applying the I/Q scheme, and based on the results in either (3) or (4), the or channel signal, single-channel scheme, i.e., either the can be utilized for localization purposes as long as the equivaremains unchanged. Howlent offset carrier frequency ever, the phase noise of the carriers in the transmitter and receiver are translated and included in the equivalent offset carrier , leading to an unstable offset carrier frequency frequency that varies with time and temperature and causes jitter and localization error. To minimize this effect, a phase-locked loop (PLL) or LO sources with extremely low phase noise and minimal temperature sensitivity should be used in this system architecture, e.g., the HMC764LP6CE, which is a PLL with integrated voltage-controlled oscillator (VCO) from the Hittite Microwave Corporation, Chelmsford, MA [35]. Fig. 6. (a) Simulated results of I/Q mismatch. (b) Reconstructed pulse signal from mismatched I/Q signal, failed to recover the original Gaussian pulse. (c) Measured I/Q mismatch.
III. SINGLE-CHANNEL SCHEME A. Carrier Phase-Noise Effects
and other phase unbalanced sources such as operational amplifier, LPFs, and cables in the I and Q channels. This combination of factors can easily build up and lead to a nonnegligible phase difference between the I/Q channels, which is termed and signals I/Q mismatch. Fig. 6(a) shows simulated at the output of the two sub-samplers using Agilent Technologies’ Advanced Design System (ADS), where a phase unbalance is introduced as a time difference between the I/Q channels of 0.2 ns before feeding into the sub-samplers. The time extension factor is set to be 1000 for this simulation example. As can be seen in Fig. 6(b), the reconstructed pulse from the phase unbalanced I/Q channels was distorted and failed to recover the original Gaussian pulse. As shown in Fig. 6(c), the experimenand signals demonstrate an even larger tally measured mismatch. Unfortunately, such phase differences between and channels are unknown and different for each base staand sigtion due to hardware variations. The extended nals have also been modulated with the equivalent offset carrier
Carriers with high and low phase-noise performance have been studied, and the impact of carrier sources with different phase-noise performances on the resulting localization system jitter has been compared through simulation and measurement. Table II lists the phase-noise performance of three types of is a low-cost commercial free-runLO sources where ning monolithic microwave integrated circuit (MMIC) VCO and are high-performance signal generator and bench-top instruments. We first consider a static scenario; under this situation, the tag is at a fixed position from the base stations. The unsynchronized simulation/experimental setup has been carried out, as shown in Fig. 7, to study the relationship between the carrier phase noise and system jitter. The jitter was calculated by recording 40 continuous time positions when the comparator is triggered at the rising edge of a fixed voltage threshold setting at around 50% of the maximum pulse served amplitude. The sub-sampled output signal from as the trigger signal for the Tektronix TDS340A oscilloscope.
ZHANG et al.: REAL-TIME NONCOHERENT UWB POSITIONING RADAR WITH MILLIMETER RANGE ACCURACY
Fig. 7. Experimental setup to study the relationship between the carrier phase noise and system jitter.
TABLE III JITTER PERFORMANCE WITH DIFFERENT CARRIER SOURCES
The simple received signal strength (RSS) method with a fixed voltage threshold has been used for pulse detection. Table III lists the results of simulated and measured rms jitter under different combinations of carrier configurations at the transmitter and receiver. In case II, only replacing the receiver with a low phase-noise carrier source has little effect improving the system jitter since the carrier with high phase-noise signal . is included in the equivalent offset carrier frequency Both the simulated and measured results show a similar trend, except for case III. In case III, when low phase-noise carriers were used at both the transmitter and receiver, the measured jitter was not reduced as expected according to the simulated results. This is due to the “shoulder” effect, a result of carrier LO phase shift superimposed on the equivalent frequency offset after the down-conversion and sub-sampling process. Maintaining the same setup as in case III, the measured rms jitter reduced to 7.2 ps when both the modulation factor and the threshold voltage were tailored to minimize the “shoulder” effect. B. “Shoulder” Effect The experimental results from Table III reveal an interesting problem of the noncoherent carrier-based UWB system using the single-channel approach. Fig. 8 shows the single-channel output after the sub-sampling process when the tag is put at a static position. The “shoulders” come from the equivalent , which modulates the time exoffset carrier frequency , as given by (3) or (4). The impact tended pulse signal of phase noise from the carrier source has been translated not only as the timing jitter, but also as the “shoulder” amplitude variation of the modulated signal. When the RSS method with a fixed voltage triggering threshold was used, the random “shoulder” amplitude variation produces another source of error. Such error created the large measured jitter in case III of Table III even when low phase-noise carriers were used at
13
Fig. 8. “Shoulder effect” in static scenario: the received single-channel subsampled signal modulated by the equivalent offset carrier frequency.
both the transmitter and receiver. By adjusting the equivalent and the threshold voltage, the offset carrier frequency amplitude variation induced error can be eliminated and the measured rms jitter was reduced to 7.2 ps, as stated earlier. However, such an optimization process is not practical since the threshold has to be set close to the peak amplitude to fully mitigate the nearby “shoulder” amplitude variation. When the tag is moving, the peak amplitude may vary, and the threshold needs to be readjusted. To further understand the impact of the “shoulder” effect, the dynamic scenario has been investigated. Under this situation, the tag is considered moving continuously away from the base station up to one wavelength at the carrier frequency. The modulation factor defined in (7) was set to be 5 and 15, respectively. In the simulation model, in order to study how the “shoulder” effect responds only to the tag dynamic movement, no phase noise was included in the carriers. Fig. 9 shows the simulated results where the “shoulder” amplitude experiences a large variation, while the tag is moved from 0 to 360 at 60 increments. By setting a fixed threshold, i.e., 0.4 V in both examples, the triggered time delay does not vary linearly with tag movement. As can be seen in Fig. 9, a large time position error occurs at certain tag positions during the tag movement, which is caused by the “shoulder” amplitude variation. C. Envelope Detection Based on the discussion in Section III-B, in the single-channel approach, in order to reduce system jitter and dynamic error, which directly translate into localization error, the “shoulder” effect, for both static and dynamic situations, needs to be minimized. According to (3) or (4), the UWB pulse signal informa, which is the down-converted pulse tion is contained in signal after sub-sampling. The useful information, i.e., the ex, is modulated by the equivalent offset tended pulse signal frequency and represents the envelope of the received pulse. Although the modulated single-channel signal, i.e., or , suffers from “shoulder” amplitude variation, and results in large timing trigger error, the envelope of the modulated signal remains relatively constant and less sensitive to the “shoulder” effect. A simple Schottky diode-based envelope detector is added following the sub-sampler as an energy collector for the time extended single-channel signal. The same dynamic simulation setup has been investigated with the tag
14
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
Fig. 9. “Shoulder” effect when tag is moving one wavelength. (a) . (b)
= 15
= 5.
moving continuously away from the base station up to one wavelength at the carrier frequency. Fig. 10 shows the simulated results of the same signal, as shown in Fig. 9, but after ED. As shown in Fig. 10, the triggered time position varies linearly with tag movement. The localization error reduced significantly as compared to Fig. 9. The “shoulder” effect has been minimized and the time position output is insensitive to trigger threshold voltage. Table IV compares and summarizes the four cases from Figs. 9 and 10. The simulated time position errors have been recorded as the tag moves a distance of one wavelength, and the standard deviation errors have been calculated. It shows that the ED minimizes the “shoulder” effect and substantially reduces the standard deviation error. No carrier phase-noise effects are included in the simulated results. It should be noted from Table IV that for the case without ED, the error decreases substantially as the modulation factor increases. However, in reality this is not true when phase noise is present since when is large, there will be an increased the modulation factor number of “shoulders” close to each other, thus an increased sensitivity to the “shoulder” amplitude variation. This also explains why in case III of Table III, both the modulation factor and the threshold voltage need to be optimized to reduce the “shoulder” effect introduced jitter error. D. Experimental Validation Noncoherent experiments were conducted to validate the simulated results from Section III-C in order to show how the combination of low phase-noise carriers and ED reduces the “shoulder” effect and minimizes jitter noise. The tag was put at a fixed position, and 1000 continuous data points were
Fig. 10. “Shoulder” effect is minimized by using ED after the sub-sampler when tag is moving one wavelength. (a) . (b) .
=5
= 15
TABLE IV SIMULATED STANDARD DEVIATION ERROR FOR DYNAMIC SCENARIO
recorded. The signal from the sub-sampler went through an A/D converter and was fed into an FPGA, where the triggering threshold was set to 50% of the pulse peak amplitude, and the optimized modulation factor was set to be around 7. Table V compares and summarizes three measured jitter results, which (in Table II) without ED; are: 1) high phase-noise carriers and without ED; and 2) low phase-noise carriers 3) low phase-noise carriers and with ED. As can be seen from Table V, the rms jitter error reduced from 18.82 to 5.73 mm by applying low phase-noise carrier sources and ED after the sub-sampler. Fig. 11 plots the measured TDOA raw data error over 1000 measurement points for these three cases. One interesting effect can be noticed in Fig. 12, which is a zoomed-in plot of the circled portion of Fig. 11, representing case II of Table V. It shows that using the low phase-noise carriers, but without ED, the error is oscillating between 12 mm. Such measurement results validate the theory of “shoulder” effect in the static mode, as shown in Fig. 8, where even carrier sources with extremely low phase noise can produce a large error due to the “shoulder” effect. The error distributions for each of the three cases are plotted in Fig. 13. In Fig. 13(b), using the low phase-noise carriers and without ED shows two distinct Gaussian distributions, corresponding to two different shoulders. After introducing ED,
ZHANG et al.: REAL-TIME NONCOHERENT UWB POSITIONING RADAR WITH MILLIMETER RANGE ACCURACY
15
TABLE V COMPARISON AND SUMMARY OF MEASURED JITTER ERROR
Fig. 11. Measured TDOA raw data error for different carrier source configurations.
Fig. 12. Zoomed-in view from Fig. 11 for measured TDOA raw data error using low phase-noise carriers LO and LO without ED.
the “shoulder” effect has been minimized and the error distribution demonstrates a single Gaussian shape with a standard deviation near 6 mm, as seen in Fig. 13(c). The resulting raw TDOA data error with Gaussian distribution could be easily reduced by increasing the number of samples averaged, which has been validated through a static 3-D experiment, as shown in Table VI. By increasing the number of TDOA samples averaged from 1 to 106, the 3-D static rms error reduced from 12.21 to 1.98 mm. However, further increase of the number of TDOA samples averaged could not effectively improve rms error since the 1.98-mm 3-D rms error is mainly from other sources such as position dilution of precision (PDOP) and clock instability [27]. Fig. 14 shows the measured sub-sampled output waveforms before and after going through the energy detector. After passing through the energy detector, the equivalent offset carhas been filtered, leaving only the envelope rier frequency signal, thus substantially increasing the system dynamic range.
Fig. 13. Measured TDOA raw data error distribution. (a) High phase-noise carriers LO without ED. (b) Low phase-noise carriers LO and LO without ED. (c) Low phase-noise carriers LO and LO with ED. TABLE VI MEASURED 3-D rms ERROR VERSUS NUMBER OF SAMPLES AVERAGED
IV. 1-D NONCOHERENT EXPERIMENT Two 1-D experiments with unsynchronized LOs and PRF clock sources were carried out to test the robustness of our
16
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
Fig. 14. Measured sub-sampled output waveforms and the signal after ED.
Fig. 16. Measured error of the 1-D unsynchronized experiment. (a) LOs with high phase noise at the tag and receiver, no ED. (b) Low phase-noise LOs at the tag and receiver, with ED after sub-sampling.
Fig. 15. Experimental setup for 1-D unsynchronized positioning measurement. (a) LOs with high phase noise at the tag and receiver, no ED. (b) Low phasenoise LOs at the tag and receiver with ED after the sub-sampler.
system. The two experimental setups are shown in Fig. 15, where only two base stations are needed for the 1-D measurements. The differences between the two 1-D experimental setups in Fig. 15 are listed as follows. 1) In Fig. 15(a) free-running VCOs with a relatively high phase noise ( 80 dBc/Hz@10 kHz) are used at both the transmitter and receiver, whereas in Fig. 15(b), low phasenoise LO sources ( 100 dBc/Hz@10 kHz) are used at both the tag and receiver, respectively. 2) The envelope detectors are used following the sub-sampler at the receiver in Fig. 15(b), whereas in Fig. 15(a), no envelope detectors are used. For both cases, millimeter-range accuracy was consistently achieved for the 1-D unsynchronized measurements at eight separate locations along a Newport optical rail with 5-cm dis-
tance between any two successive measurements. As shown in Fig. 16(a), the system jitter causes noticeable short-term variation in the error at each static point of roughly 19 mm. This short-term variation was mitigated by averaging 32 pulses at each static point. For the single-channel scheme with low phasenoise carriers and ED, results shown in Fig. 16(b) demonstrate the jitter has a much smaller short term variation of roughly 6 mm at each static point, compared to the 19 mm shown in Fig. 16(a). This small short-term variation was mitigated by averaging only four pulses at each static point, thus speeding up the processing time. In Table VII, we compare the results between both cases, and it is clear that the single-channel scheme with low phase-noise carriers and ED requires less times averaging and produces less 1-D error. Compared to the coherent experimental results reported in our previous study [27], the mean error in measuring the 1-D static data increases from 1.49 to 2.38 mm. The increase in error of 0.89 mm is comparable to the measured error of 1.05 mm due to the PRF clock jitter discussed in [27]. V. 3-D NONCOHERENT EXPERIMENT Two 3-D experiments with unsynchronized LOs and PRF clock sources were carried out, where a minimum of four base
ZHANG et al.: REAL-TIME NONCOHERENT UWB POSITIONING RADAR WITH MILLIMETER RANGE ACCURACY
17
TABLE VII COMPARISON OF NONCOHERENT 1-D EXPERIMENTAL RESULTS
Fig. 20. 3-D dynamic mode with ED. x-; y -; and z -axes error compared to Optotrak measurements.
Fig. 17. 3-D experiment setup of unsynchronized localization system using a single-channel demodulation with ED.
Fig. 21. Experimental setup of robot tracking using the developed noncoherent UWB system.
Fig. 18. 3-D unsynchronized localization experiments, four base-station distribution with locations for each base station.
shown in Fig. 17, compared to our previous approach in Fig. 5 are as follows. 1) Single channel of down-conversion has been used instead of I/Q, which lowers the cost and reduces the system complexity. 2) Following the sampling mixer, the energy detectors are added, which helps in getting rid of the carrier offset due to the frequency difference between the transmitter and receiver LOs. 3) Low phase-noise LOs are used at both the transmitter and receiver. A. 3-D Dynamic Free Motion
Fig. 19. 3-D dynamic random mode with ED. UWB trace is compared to Optotrak trace.
stations are needed for the 3-D measurements. The signals from the sub-sampler are then fed to the FPGA, which uses the newly developed leading-edge detection algorithm to locate the pulse positions [36]. The major improvements of the system setup, as
Fig. 18 shows a four base-station setup where the 3-D positions were measured for each base station utilizing the Optotrak 3020 system, which also serves as a reference for comparing the 3-D real-time accuracy of our UWB localization system. The Optotrak 3020 has 3-D real-time accuracy of better than 0.3 mm. It should be noted that the spatial spread of the base stations along the -axis is the largest (2498 mm), while the -axis is the smallest (1375 mm). In the dynamic mode, the tag is moving randomly inside the 3-D space indicated in Fig. 18. The 3-D motion of the tag is then plotted and UWB measurements are compared with Optotrak measurements. RMSE is used to report the
18
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
Fig. 23. 3-D robot tracking at static positions. UWB points compared to Optotrak points.
TABLE VIII ERROR SUMMARY—3-D UNSYNCHRONIZED LOCALIZATION EXPERIMENTS
Fig. 20 shows the 3-D dynamic errors in the - - and -axes over 1000 measured points. The overall 3-D RMSE is 6.37 mm. The error along the -axis contributed most to the overall distance error, which can be explained by the limited spatial spread of base stations along the -axis and can be calculated using the PDOP definitions in [27]. Such error can be easily mitigated through better arrangement of the base stations along the -axis. B. 3-D Robot Tracking The next noncoherent 3-D experiment is to dynamically track the robot position. The experimental setup is shown in Fig. 21. The monopole antenna and the reference Optotrak probe are tied together and then fixed to the arm of the CRS A465 robot. The robot was pre-programmed to specifically cover 20 distinct static positions in a 3-D volume, stopping for 3 s at each position and then moving to the next position and so on. The measured traces by the UWB system are compared to the Optotrak reference system, as shown in Fig. 22. Fig. 23 shows the 20 distinct static positions taken by both the UWB and the Optotrak systems. The overall dynamic 3-D robot tracking RMSE is 5.24 mm. In Table VIII, we summarize the real-time noncoherent 3-D experimental results under various scenarios. The reported RMSE are based on 1000 continuous data points recorded and compared to the Optotrak 3020 system, which served as the real-time reference of our UWB localization system and provides a 3-D accuracy of better than 0.3 mm. Fig. 22. 3-D dynamic robot tracking. UWB trace compared to Optotrak trace. -plane. (c) -plane. (d) -plane. (a) 3-D view. (b)
VI. CONCLUSION
error since it is the true unbiased error when data values fluctuate above and below zero. Fig. 19 plots the UWB trace and Optotrak trace in the 3-D dynamic mode.
A novel architecture for UWB positioning systems has been presented, which combines the single-channel carrier-based UWB system and traditional ED-based UWB positioning system. The UWB localization system is equipped with low
XY
XZ
YZ
ZHANG et al.: REAL-TIME NONCOHERENT UWB POSITIONING RADAR WITH MILLIMETER RANGE ACCURACY
phase-noise carrier sources at both the transmitter and receiver and the advanced sub-sampling mixer at the receiver for equivalent time sampling of the incoming pulse train. A proper modulation factor is intentionally chosen between the transmitter and receiver carrier and is combined with an ED step to eliminate the requirement of carrier synchronization. We have addressed step-by-step the main challenges, which led us to the finalized system architecture, including the I/Q mismatch, jitter errors due to phase noise in carrier offsets, the “shoulder” effect in static and dynamic scenarios, etc. Both simulation and measurement results show the robustness of the proposed system architecture with a reduced timing jitter error and improved system dynamic range. By comparing two 1-D experiments, the ranging error has been improved significantly with the reduced timing jitter and “shoulder” effect through application of a low phase-noise carrier-based UWB architecture together with advanced sub-sampling and ED. To further validate the theories, extensive 3-D static and dynamic experiments have been performed, including a tag moving randomly in a 3-D space and a tag attached to a robot arm with preplanned motion, where constant millimeter-range accuracy in both static and dynamic scenarios has been demonstrated. This is a milestone in UWB and wireless positioning systems and opens up many new and exciting applications for the future.
REFERENCES [1] “RTLS and wireless technologies,” Venture Development Corporation, Natick, MA, 2005–2006. [Online]. Available: http://www.pencomputing.com/news/news_rtls.html, RFID Business Planning Service [2] “RFID forecasts, players and opportunities 2009–2019,” Summary, IDTechEx, Cambridge, MA, 2009. [Online]. Available: http://www. idtechex.com/research/reports/rfid_forecasts_players_and_opportunities_2009_2019_000226.asp [3] J. Kowal, F. Langlotz, and L. Nolte, “Basics of computer-assisted orthopaedic surgery,” in Navigation and MIS in Orthopedic Surgery. Berlin, Germany: Springer, 2007, ch. 1, pt. I, pp. 2–8. [4] M. Vossiek, L. Wiebking, P. Gulden, J. Wieghardt, C. Hoffman, and P. Heide, “Wireless local positioning,” IEEE Microw. Mag., vol. 4, pp. 77–86, Dec. 2003. [5] B. Waldmann, R. Weigel, and P. Gulden, “Method for high precision local positioning radar using an ultra wideband technique,” in IEEE MTT-S Int. Microw. Symp. Dig., Atlanta, GA, 2008, pp. 117–120. [6] C. Meier, A. Terzis, and S. Lindenmeier, “A robust 3D high precision radio location system,” in IEEE MTT-S Int. Microw. Symp. Dig., 2007, pp. 397–400. [7] C. Meier, A. Terzis, and S. Lindenmeier, “Investigation and suppression of multipath influence on indoor radio location in the millimeter wave range,” in Wave Propag. Commun., Microw. Syst. Navigat. Conf., Chemnitz, Germany, 2007, pp. 21–24. [8] T. E. McEwan, “Radiolocation system having writing pen application,” U.S. Patent 6 747 599, Jun. 8, 2004. [9] G. Ossberger, T. Buchegger, E. Schimback, A. Stelzer, and R. Weigel, “Non-invasive respiratory movement detection and monitoring of hidden humans using ultra wideband pulse radar,” in IEEE Int. UWB Syst. Tech. Conf., Kyoto, Japan, 2004, pp. 395–399. [10] A. Fujii, H. Sekiguchi, M. Asai, S. Kurashima, H. Ochiai, and R. Kohno, “Impulse radio UWB positioning system,” in IEEE Radio Wireless Symp., 2007, pp. 55–58. [11] Z. N. Low, J. H. Cheong, C. L. Law, W. T. Ng, and Y. J. Lee, “Pulse detection algorithm for line-of-sight (LOS) UWB ranging applications,” IEEE Antennas Wireless Propag. Lett., vol. 4, pp. 63–67, 2005. [12] R. Zetik, J. Sachs, and R. Thomä, “UWB localization—Active and passive approach,” in Proc. 21st IEEE IMTC, 2004, vol. 2, pp. 1005–1009. [13] A. Stelzer, C. G. Diskus, and H. W. Thim, “A microwave position sensor with sub-millimeter accuracy,” IEEE Trans. Microw. Theory Tech., vol. 47, no. 12, pp. 2621–2624, Dec. 1999.
19
[14] A. Stelzer, K. Pourvoyeur, and A. Fischer, “Concept and application of LPM-a novel 3-D local position measurement system,” IEEE Trans. Microw. Theory Tech., vol. 52, no. 12, pp. 2664–2669, Dec. 2004. [15] F. Ellinger et al., “Local positioning for wireless sensor networks,” in IEEE Globecom Workshops, Washington, DC, 2007, pp. 1–6. [16] R. Mosshammer, M. Huemer, R. Szumny, K. Kurekt, J. Hittner, and R. Gierlichli, “A 5.8 GHz local positioning and communication system,” in IEEE MTT-S Int. Microw. Symp. Dig., Honolulu, HI, 2007, pp. 1237–1240. [17] P. Tragas et al., “Resolution: reconfigurable systems for mobile local communication and positioning,” in Mobile Wireless Commun. Summit, Budapest, Hungary, 2007, pp. 1–5. [18] “LPR-2D,” Symeo, Munich, Germany, 2009. [Online]. Available: http://www.symeo.com/cms/upload/PDF/Datasheet_LPR-2D.pdf [19] R. Feger, C. Wagner, S. Schuster, H. Jäger, and A. Stelzer, “A 77-GHz FMCW MIMO radar based on an SiGe single-chip transceiver,” IEEE Trans. Microw. Theory Tech., vol. 57, no. 5, pp. 1020–1035, May 2009. [20] I. Guvenc, Z. Sahinoglu, and P. V. Orlik, “TOA estimation for IR-UWB systems with different transceiver types,” IEEE Trans. Microw. Theory Tech., vol. 54, no. 4, pt. 2, pp. 1876–1886, Apr. 2006. [21] T. Buchegger, G. Obberger, A. Reisenzahn, E. Hochmair, A. Stelzer, and A. Springer, “Ultrawideband transceivers for cochlear implants,” EURASIP J. Appl. Signal Process., vol. 2005, no. 18, pp. 3069–3075, 2005. [22] J. P. Lie, C. M. See, and B. P. Ng, “UWB ranging with high robustness against dominant jammer and multipath,” IEEE Microw. Wireless Compon. Lett., vol. 15, no. 12, pp. 907–909, Dec. 2005. [23] A. Rabbachin and I. Oppermann, “Synchronization analysis for UWB systems with a low-complexity energy collection receiver,” in Proc. IEEE Ultrawideband Syst. Technol. Conf., Kyoto, Japan, May 2004, pp. 288–292. [24] R. J. Fontana and F. J. Larrick, “Ultra wideband receiver with high speed noise and interference tracking threshold,” U.S. Patent 5 901 172, May 4, 1999. [25] C.-C. Chong, S. K. Yong, and S.-S. Lee, “UWB direct chaotic communication technology,” IEEE Antennas Wireless Propag. Lett., vol. 4, pp. 316–319, 2005. [26] D. Barras, F. Ellinger, H. Jäckel, and W. Hirt, “A robust front-end architecture for low-power UWB radio transceivers,” IEEE Trans. Microw. Theory Tech., vol. 54, no. 4, pp. 1713–1723, Apr. 2006. [27] M. Mahfouz, C. Zhang, B. Merkl, M. Kuhn, and A. Fathy, “Investigation of high accuracy indoor 3-D positioning using UWB technology,” IEEE Trans. Microw. Theory Tech., vol. 56, no. 6, pp. 1316–1330, Jun. 2008. [28] C. Zhang, M. Kuhn, B. Merkl, M. Mahfouz, and A. E. Fathy, “Development of an UWB indoor 3-D positioning radar with millimeter accuracy,” in IEEE MTT-S Int. Microw. Symp. Dig., 2006, pp. 106–109. [29] C. Zhang, A. Fathy, and M. Mahfouz, “Performance enhancement of a sub-sampling circuit for ultra-wideband signal processing,” IEEE Microw. Wireless Compon. Lett., vol. 17, no. 12, pp. 873–875, Dec. 2007. [30] D. M. Treyer and W. Bächtold, “Investigation of a self-calibrating SSB modulator,” IEEE Trans. Microw. Theory Tech., vol. 53, no. 12, pp. 3806–3816, Dec. 2005. [31] M. Kuhn, C. Zhang, B. Merkl, D. Yang, Y. Wang, M. Mahfouz, and A. Fathy, “High accuracy UWB localization in dense indoor environments,” in IEEE Int. Ultra-Wideband Conf., Hannover, Germany, 2008, vol. 2, pp. 129–132. [32] M. Kuhn, C. Zhang, S. Lin, M. Mahfouz, and A. Fathy, “A system-level design approach to UWB localization,” in IEEE MTT-S Int. Microw. Symp. Dig., Boston, MA, 2009, pp. 409–412. [33] C. Zhang, M. Kuhn, M. Mahfouz, and A. E. Fathy, “Real-time noncoherent UWB positioning radar with millimeter range accuracy in a 3D indoor environment,” in IEEE MTT-S Int. Microw. Symp. Dig., Boston, MA, 2009, pp. 413–416. [34] C. Zhang and A. E. Fathy, “Reconfigurable pico-pulse generator for UWB applications,” in IEEE MTT-S Int. Microw. Symp. Dig., 2006, pp. 407–410. [35] “PLLs w/integrated VCOs, HMC764LP6CE datasheet,” Hittite Microw. Corporation, Chelmsford, MA, 2009. [Online]. Available: http://www44-.hittite.com/ [36] B. Merkl, “The future of the operating room: Surgical preplanning and navigation using high-accuracy ultra-wideband positioning and advanced bone measurement,” Ph.D. dissertation, Dept. Biomed. Eng., Univ. Tennessee, Knoxville, TN, 2008.
20
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 58, NO. 1, JANUARY 2010
Cemin Zhang was born in Chengdu, Sichuan, China, in 1978. He received the B.S. and M.S. degrees in information science and electronic engineering from Zhejiang University, Hangzhou, China, in 2001 and 2004, respectively, and the Ph.D. degree in electrical engineering from The University of Tennessee, Knoxville, in 2008. In 2003, he was an RF Engineer with the UTStarcom Corporation Ltd, Hangzhou, China, where he was involved with the development of antenna switch and mobile base station hardware. In early 2004, he was a Product Engineer with the Intel Corporation, Shanghai, China, where he was involved with the development of flash memory. In November 2008, he joined the Hittite Microwave Corporation, Chelmsford, MA, as a Monolithic Microwave Integrated Circuit (MMIC) Design Engineer, currently involved in the research and development of various cutting-edge MMIC components including low phase-noise VCOs. He has established a novel unsynchronized UWB system architecture to achieve the real-time millimeter-range 3-D localization accuracy and developed various microwave components including a tunable picosecond pulse generator, high-speed sampler, and UWB antennas for such system. He has authored or coauthored over 30 journal/conference papers and presented at numerous international conferences. Dr. Zhang is a member of Phi Kappa Phi and Sigma Xi. He has served as a reviewer for many journals/transactions including the IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES. He was the recipient of the 2007 URSI Student Fellowship and 2008 UT Chancellor’s Citation for Extraordinary Professional Promise. Michael J. Kuhn (S’06) was born in Wheat Ridge, O, in 1982. He received the B.S. degree in electrical engineering and B.S. degree in computer science from the Colorado School of Mines, Golden, in 2004, the M.S. degree in engineering science from The University of Tennessee, Knoxville, in 2008, and is currently working toward the Ph.D. degree in biomedical engineering at The University of Tennessee. Since 2005, he has been a Researcher with the Center for Musculoskeletal Research, The University of Tennessee. He has authored or coauthored and presented papers at numerous international conferences in the fields of biomedical engineering and also microwave and antenna engineering. His current research interests include medical applications of UWB, numerical techniques in microwave engineering, and orthopedic surgical navigation. Mr. Kuhn was the recipient of a Ph.D. Fellowship from the College of Engineering, The University of Tennessee, in 2005. Brandon C. Merkl (S’06–M’09) received the B.S. degree in electrical engineering and B.S. degree in computer science from the Colorado School of Mines, Golden, in 2004, and Ph.D. degree in biomedical engineering The University of Tennessee, Knoxsville, in 2008. From 2005 to 2008, he was a Researcher with the Center for Musculoskeletal Research, The University of Tennessee. He is a founding member of Sapientia Technologies Inc. He is currently a Senior Software Engineer for Medtronic Navigation, Louisville, CO, where his work with the Advanced Development Group consists of research and development of image processing algorithms, computer vision applications, and novel extensions to the medical imaging and information analysis domains.
He has authored or coauthored publications in the fields of image processing, forensic anthropology, biomedical engineering, clinical orthopedics, and signal processing. His research interests consist of such disparate topics as high-performance computing, machine learning, anatomical modeling/analysis, fuzzyneural systems, multivariate statistics, computer vision, information theory, and nonlinear optimization. Dr. Merkl is a member of the IEEE Computer Society, the IEEE Computational Intelligence Society, and the IEEE Engineering in Medicine and Biology Society. Aly E. Fathy (S’82–M’84–SM’92–F’04) received the B.S.E.E. degree, the B.S. degree in pure and applied mathematics, and the M.S.E.E. degree from Ain Shams University, Cairo, Egypt, in 1975, 1979, and 1980, respectively, and the Ph.D. degree from the Polytechnic Institute of New York, Brooklyn, in 1984. In February 1985, he joined the RCA Research Laboratory (currently the Sarnoff Corporation), Princeton, NJ, as a Member of the Technical Staff. In 2001, he became a Senior Member of the Technical Staff. With the Sarnoff Corporation, he was engaged in research and development of various enabling technologies such as high-T superconductors, low-temperature co-fired ceramic (LTCC), and reconfigurable holographic antennas. He was also an Adjunct Professor with the Cooper Union School of Engineering, New York, NY. In August 2003, he joined The University of Tennessee, Knoxville, as an Associate Professor. He has authored or coauthored numerous transaction and conference papers. He holds 11 U.S. patents. His current research interests include wireless reconfigurable antennas, see-through walls, UWB systems, and high-efficiency high-linearity combining of digital signals for base-station amplifiers. He has developed various microwave components/subsystems such as holographic reconfigurable antennas, radial combiners, direct broadcast antennas (DBSs), speed sensors, and LTCC packages for mixed-signal applications. Dr. Fathy is a member of Sigma Xi and Eta Kappa Nu. He is an active member of the IEEE Microwave Theory and Techniques Society (IEEE MTT-S) International Microwave Symposium (IMS) Technical Program Committee, the IEEE Antenna and Propagation Symposium, and the IEEE Radio and Wireless Steering Committee. He was the general chair of the 2008 IEEE Radio and Wireless Conference. He was the recipient of five Sarnoff Outstanding Achievement Awards (1988, 1994, 1995, 1997, and 1999). Mohamed R. Mahfouz (S’98–M’01–SM’06) received the B.S.B.M.E. and M.S.B.M.E. degrees from Cairo University, Cairo, Egypt, in 1987 and 1992, respectively, the M.S.E.E. degree from the University of Denver, Denver, CO, in 1997, and the Ph.D. degree from the Colorado School of Mines, Golden, in 2002. From 1998 to 2002, he was the Technical Director with the Rocky Mountain Musculoskeletal Research Laboratory, Denver, CO. In 2002, he became both Technical Director for the Center for Musculoskeletal Research and an Associate Professor with The University of Tennessee, Knoxville. He has authored many journal articles, conference papers, and book chapters. His current research interests include medical applications of UWB, biomedical instrumentation, medical imaging, surgical navigation, microelectromechanical systems (MEMS) bio-sensors, and 3-D bone and tissue reconstruction. Dr. Mahfouz was the recipient of numerous National Institutes of Health (NIH) and National Science Foundation (NSF) grants.
