Processing math: 0%
X
By Topic

IEEE Quick Preview
  • Abstract
  • Authors
  • Figures
  • Multimedia
  • References
  • Cited By
  • Keywords

A Precise Pointing Technique for Free Space Optical Networking

In free space optical communication networks, pointing, acquisition, and tracking (PAT) techniques are needed to establish and maintain optical links among the static or mobile nodes in the network. This paper describes a precise pointing technique to steer the local directional laser beam of an optical transceiver to a target optical transceiver at a remote transceiver node. The pointing technique utilizes Real-Time Kinematic GPS coordinates, local angular sensors, and a reference baseline, to retrieve accurate navigation information (roll, pitch, yaw) of the mobile or static platform that carries an optical transceiver. Through experiments using gimbal pointing stages, we have demonstrated "dead-reckoning" pointing accuracy in the milliradian range in our outdoor testbed.

INTRODUCTION

Free Space Optical (FSO) communication has been recognized as a high-speed bridging technology to current fiber optic networks [1], and a valuable technology in commercial and military backbone networks [2]. However its bright prospects depend on the performance of pointing, acquisition, and tracking (PAT) technique, and autonomous reconfiguration algorithms dealing with the effects of node mobility and atmospheric obscuration.

Precise laser beam pointing requiring microradian to milliradian accuracy is a challenging problem unless both nodes are close to each other. In this case manual alignment is straightforward, and can be guided by the use of optical beacons, or image based pointing. However, if the link distance is more than a few kilometers, then these techniques become increasingly difficult to implement. Instead, we need complete information as to where nodes are (their position coordinates) as well as FSO transceiver angular pointing coordinates (pointing vectors). The use of various kinds of position and angular sensor devices is therefore natural in pointing techniques, as described previously [3] [4] [5] [6] [7].

References [3] [4] have described a coarse pointing system using a 180� field-of-view (FOV) fisheye camera and a 30� FOV regular camera. The omnidirectional fisheye camera first acquires the target (fixed or mobile node) of interest by image extraction. Next, the 300 FOV camera is rotated toward the target of interest based on an homographical computation. The rotation angles are used to generate a radial trifocal tensor, which is applied to estimate the movement of the FSO transceiver on the target (tracking). In this work a pointing error around 0.2� was reported.

References [5] [6] describe low-cost and lightweight PAT systems for mobile nodes such as aircraft or ground vehicles. Their pointing system (MOCT: Mobile Optical Communication Terminal) used both Differential GPS (DGPS) and inertial navigation system (INS) to measure a coarse pointing vector as well as a camera to provide a fine pointing vector. However, due to the low accuracy of DGPS (5 m) and INS (3�), the coarse pointing error appeared as 0.79� (mean) in azimuth and 0.3� (mean) in elevation; the standard deviations of the pointing error in azimuth and elevation were 1.25� and 0.2�, respectively. They employed an optical beacon in measuring a fine pointing vector. The camera on the MOCT continuously tracked the beacon on the ground, and the real-time image was used to compute the fine pointing vector. Their tracking system was based on an optical camera (17� FOV) and DGPS. During their tracking experiment, an aircraft regularly broadcast its error corrected position; then an optical ground station tracked the airplane by estimating the next position and velocity of the aircraft.

Reference [7] describes a method of pointing to a designated stellar target (e.g., star or satellite) with known position coordinates from a mobile platform (e.g., airplane, terrestrial vehicle, or ship). An INS mounted on a pointing instrument (e.g., telescope, antenna, sensor, laser, missile launcher, etc.) measures three dimensional attitude angles (roll, pitch, yaw); then the angle difference between the desired pointing angles to the target and the attitude angles from the INS provide pointing command signals as input to the pointing instrument. The pointing system utilizes GPS position updates to compensate for drift errors in the INS measurements.

The above PAT systems depend on optical devices ([3] [4] [5] [6]) or INS combined with GPS ([5] [6] [7]); GPS or DGPS acts as a secondary device for providing position updates and correcting INS drift error, whose position accuracy is at the meter level. In this paper, we describe a precise PAT technique that is applicable to link initiation in FSO networking; the technique employs Real-Time Kinematic (RTK) GPS and local angular sensors (e.g., tilt sensors or INS). The RTK GPS provides centimeter level positioning accuracy (e.g., 2 cm for the NovAtel RT2W); it has been used in applications requiring precise location, navigation, and tracking [8] [9] [10]. Because we use highly accurate GPS to provide primary pointing information, our pointing method is distinct from previously described PAT systems [3] [4] [5] [6] [7].

DESIGN OF POINTING SYSTEM USING RTK GPS

A. Background

We have been developing a pointing technique that is applicable to link initiation in FSO networking. At link initiation, each node measures its position by stand-alone GPS or DGPS. Then, it broadcasts this position information through a wireless RF data transceiver. In a centralized network (such as a ring network), a central node collects all position information from all nodes within the RF coverage region. Next, a central control node decides on the best ring topology from the available information, and sends out pointing commands to each node for the establishment of links. In a decentralized or distributed network, however, each node can make its own link decision based on the GPS location information from its neighbor nodes. Since the proposed pointing technique aims a point-to-point interconnection between any two nodes, it can be applied to centralized or decentralized networking.

Our pointing technique is based on measuring i) three dimensional attitude angles (roll, pitch, yaw) of an FSO transceiver mounted on a two-axis gimbal and ii) pointing vector of the transceiver (i.e., where the FSO transceiver is directing its laser beam). The attitude angles and pointing vector are measured on the local tangent plane or navigation frame, such as East-North-Up (ENU) coordinates. For example, as illustrated in Figure 1, the pointing procedure works as follows:

 Pointing Procedures

  1. The precise position of two FSO transceivers at A and B is measured referenced to a local origin in ENU coordinates.

  2. The displaced angles, \theta_{1} and \theta_{2} , to the baseline \overline{AB} are determined; they become a control input to the gimbal.

  3. The FSO transceivers at A and B are aligned to be on the baseline \overline{AB} ; two transceivers point to each other (and become interconnected).

Figure 1. A pointing scenario

Once the baseline vector \overline{AB} or \overline{BA} is known in Step 1, each pointing system at A and B continues to Step 2, independently. The only necessary information for the two pointing systems at A and B is each other's location; then the two systems operate independently to align their FSO transceivers on the baseline. By this scheme, we can interoperate with another pointing system placed at B (whose system components may be different from those used at A ) only if we know the coordinates of B .

B. Positioning and Local Angular Sensors

RTK GPS is used to measure the precise position of FSO transceiver and generate a pointing vector ofthe transceiver in ENU coordinates. RTK GPS position data ([X,Y,Z]^{T}) in the World Geodetic System 1984 (WGS-84) coordinates is transformed to ENU coordinates [11]. The pointing vector provides yaw (\psi) information which is defined as a deviation angle from the N-axis in ENU coordinates. Roll (\phi) and pitch (\theta) are measured by tilt sensors for a fixed node or by inertial sensors (INS or IMU) for a mobile node.

C. Method of Pointing Vector Measurement

A vector is defined by two points in a coordinate frame; if the coordinates of the two points are P_{1} and P_{2} , respectively, then the vector is computed by the difference of the two coordinates, P_{2}-P_{1} , which is the vector from P_{1} to P_{2} . Likewise, knowing the pointing vector of an FSO transceiver requires measurement of two points on the path through which the laser beam of the transceiver passes. Figure 2 illustrates our method for pointing vector measurement. Figure 2a shows a GPS antenna mounted on FSO transceiver placed on a mobile platform which is stationary at a location C_{1} ; then, the mobile platform moves forward in a straight line to another location C_{2}.C_{1} and C_{2} are RTK GPS coordinates for a GPS antenna mounted on the FSO transceiver, whose coordinates are C_{1}=[E_{1},N_{1},U_{1}]^{T} and C_{2}=[E_{2},N_{2},U_{2}]^{T} . Therefore, the pointing vector between C_{1} and C_{2} is calculated as: u={[E_{2}-E_{1},N_{2}-N_{1},U_{2}-U_{1}]^{T}\over D}, \eqno{\hbox{(1)}} View Source Right-click on figure for MathML and additional features. where D=\mid\mid[E_{2}-E_{1}, N_{2}-N_{1},U_{2}-U_{1}]^{ T}\mid\mid and \mid\mid\bullet\mid\mid is the Euclidean vector norm. Figure 2b shows a second way of measuring the pointing vector: align FSO transceiver located at C_{1} to a pre-determined target at C_{2} (with previously determined position coordinates).

From the pointing vector (u) in Equation (1) and position coordinates of C_{2} ,we can obtain the pointing vector centered at C_{2} as shown in Figure 2a (or C_{1} in Figure 2b), which is the current location of the FSO transceiver.

Figure 2. Pointing vector measurement

The RTK GPS position coordinates C_{1} and C_{2} contain measurement errors at the centimeter level. If we denote {\varepsilon}_{1} and \varepsilon_{2} as the three-dimensional measurement errors in C_{1} and C_{2} , respectively, then the unit vector u is expressed as: u={[E_{2}-E_{1},N_{2}-N_{1},U_{2}-U_{1}]^{T}+(\varepsilon_{2}-\varepsilon_{1})\over D_{\varepsilon}}, \eqno{\hbox{(2)}} View Source Right-click on figure for MathML and additional features. where D_{\varepsilon}=\mid\mid[E_{2}-E_{1}, N_{2}-N_{1}, U_{2}-U_{1}]^{ T}+(\varepsilon_{2}-\varepsilon_{1})\mid\mid and D_{\varepsilon}\leq D+\mid\mid\varepsilon_{2}-\varepsilon_{1}\mid\mid by the Triangle Inequality [12].

If we assume that \mid\mid\varepsilon_{1}\mid\mid and \mid\mid\varepsilon_{2}\mid\mid are less than 3 cm, then \mid\mid\varepsilon_{2}-\varepsilon_{2}\mid\mid\leq\mid\mid\varepsilon_{2}\mid\mid+\mid\mid\varepsilon_{1}\mid\mid\leq <6cm. Therefore, if D is much larger than both \mid\mid\varepsilon_{1}\mid\mid and \mid\mid\varepsilon_{2}\mid\mid({\rm e.g}., D=6m) , then the following relations are hold: \eqalignno{&{(\varepsilon_{2}-\varepsilon_{1})\over D_{\varepsilon}}\leq{\mid\mid\varepsilon_{2}-\varepsilon_{1}\mid\mid\over D_{\varepsilon}}\leq{\mid\mid\varepsilon_{2}\mid\mid +\mid\mid\varepsilon_{1}\mid\mid\over D_{\varepsilon}}\simeq{\mid\mid\varepsilon_{2}\mid\mid+\mid\mid\varepsilon_{1}\mid\mid\over D}\simeq 0&{\hbox{(3)}}\cr &u\simeq{[E_{2}-E_{1},N_{2}-N_{1},U_{2}-U_{1}]^{T}\over D} &{\hbox{(4)}}} View Source Right-click on figure for MathML and additional features.

Hence, as D increases, the pointing vector becomes closer to the true one without measurement errors.

The East and North components of the pointing vector determine yaw (\psi) . With roll and pitch from local angular sensors and yaw from RTK GPS, we have complete attitude angle information of the FSO transceiver which is necessary to convert the pointing vector in the navigation frame (ENU coordinates) to the one in the body frame by the following equations: P_{B}=C_{ENU}^{B}P_{ENU}=C(\phi)C(\theta)C(\psi)P_{ENU}, \eqno{\hbox{(5)}} View Source Right-click on figure for MathML and additional features. where

  • (\phi,\theta,\psi) : Attitude angles

  • C(\phi),C(\theta),C(\psi) : Rotation matrices

  • C_{ENU}^{B}=C(\phi)C(\theta)C(\psi) : Transformation matrix from the ENU coordinates to the body frame.

  • F_{ENU} : Position in the ENU coordinates

  • F_{B} : Position in a body frame.

With the baseline vector {AB} or {BA} transformed to the body frame by Equation (5), the control input to the gimbal (i.e., heading and elevation angles in Figure 1) is computed from the pointing vector and the baseline vector transformed to the body frame (we skip the details here).

The advantages of this pointing method are summarized as:

  1. Conceptually simple and easy to implement.

  2. A complete attitude angle can be measured only by RTK TK GPS, if the the surface is flat or as long as we can keep the surface level using a stabilizer.

  3. RTK GPS consists of a GPS unit and a wireless RF data transceiver; the precise position data will be sufficient to track mobile nodes. Its high accuracy will improve the performance of position position and velocity estimation ofa mobile node.

  4. The precise heading information from RTK GPS can be combined with the roll and pitch outputs from INS mounted on an aircraft [13] [14]; thus enabling us to use the pointing technique in a dynamic environment.

Since RTK TK GPS requires the observation of at least five GPS satellites to yield such high accurate position coordinates, our basic assumption in this pointing method is that there is no significant GPS signal blockage on the site where the pointing system is being operated.

EXPERIMENTAL RESULTS

A. Overview of Experiments

We have carried out a medium range pointing experiment between two distant buildings (roof-to-roof) on the University of Maryland campus at College Park; its purpose was to test how the length of D affects pointing accuracy of the proposed method.

By keeping roll and pitch close to zero ({\rm i.e}., \phi=0, \theta=0) , Equation (5) is simplified to P_{B}=C(\psi)P_{ENU} CQ()PENU, which means that RTK GPS positioning error is a major error source influencing the pointing accuracy. As D increases, the measurement error in the pointing vector becomes smaller, which renders measurement error in yaw (\psi) smaller; consequently the pointing vector transformed to the body frame becomes more accurate. Hence, we can control the heading and elevation angles of a two-axis gimbal more accurately.

In this experiment, the pointing vector was measured by the pointing method depicted in Figure 2b. Figure 3 shows the diagram ofthe experiment. Because ofthe low load bearing capacity of our small two-axis gimbal, we used a helium-neon laser pointer (\lambda=633nm) to mimic an FSO transceiver; the laser pointer was located at A . The position coordinates at B was pre-determined by RTK GPS (Model: NovAtel RT2W). A pointing target mimicking an FSO transceiver placed at a remote site C (equivalent to the location B in Figure 1). The distance between the two points A and C was 264 m .

Figure 3. Diagramof PAT experiment

B. Results of the Experiment

Table 1 presents the pointing accuracy obtained from the experiment. Figure 4 shows our definition of the pointing error in Table 1. The pointing error is defined as the distance from the pointing target C . Because the distance between A and C is 264 m , a distance of 0.264 m corresponds to a 1 milliradian pointing error.

Figure 4. The pointing target used in the experiment.

In Table 1, the first column represents the distance between A and B . The position coordinates of A were [50.621, 259.58, 1 4734]^{T} with \sigma_{E}=0.26 cm, \sigma_{N}=0.36 cm, and \sigma_{U}=0.88cm . For each distance, we conducted the pointing experiment at four different sites of B with one site of A ; thus, the total number of pointing trials was sixteen. The second column shows the pointing error range corresponding to the different locations of B . The third column summarizes the standard deviation values of the ENU coordinates of B ; they show that the horizontal and vertical positioning errors of RTK GPS coordinates are within \pm 1 cm and \pm 2 cm with 95% confidence limits, respectively. The last column displays the Position Dilution of Precision (PDOP) for the RTK GPS position coordinates, in which PDOP was within the normal range between 1 and 6. As we mentioned at the previous section, the pointing error decreases with increasing D . We observed that fifteen of the sixteen pointing trials satisfied the 1 milliradian pointing accuracy; two cases with D=6m and 9 m were outside the 1 milliradian radius.

Table 1. Pointing Experiment Results

FUTURE RESEARCH AND DISCUSSION

A. Speed and Reliability ofthe Pointing Method

We have described an accurate pointing method applicable to link initiation in FSO networking; the link initiation should occur in less than 1 second with milliradian accuracy. We have shown that the 1 milliradian pointing accuracy can be achieved with D=9\sim 40m by using RTK TK GPS. How quickly the pointing method can obtain such accuracy is still questionable. We will answer this question by conducting pointing experiments with increased position update rate using RTK GPS (NovAtel I RT2W: max. 20 Hz update rate). The experiment will be combined with a reliability test with bi-axial tilt sensors (\pm 0.01^{\circ} repeatability, \pm 0.004^{\circ} resolution); the reliability test will provide us with the success rate of the pointing method. To realize the precise pointing method with actual FSO transceiver, the current two-axis gimbal with low load bearing capacity will be replaced with the one with high load capacity.

B. GPS Signal Availability

RTK GPS receivers (at base station and rover) should keep track of at least the same 5 GPS satellites above a mask angle (> \geq 10^{\circ} in elevation; preset inside the receiver) on both LI (1,575 MHz) and L2 (1,227.60 MHz) to guarantee its RTK performance. In the GPS Modernization plan, a new L5 (1,176.45 MHz) signal will be available for GPS users [15]. The new signal will provide signal redundancy to the users so that they can choose any combination ofthe signals (e.g., LI/L2, L2/L5, or LI/L5) for their RTK operation.

The European global positioning satellite system (GNSS), Galileo, will also increase signal availability in the future. Galileo will be composed of a constellation of 30 satellites; the current GPS constellation consists of 30 satellites [16] [17]. Since both GPS and Galileo are designed to be in sight of at least four satellites by anyone anywhere in the world, GPS users will be able to keep track of at least eight GNSS satellites, which subsequently satisfies the RTK GPS requirement on the minimum number of observable satellites (i.e., five satellites).

C. Extended RTK Service

For RTK operation, a RTK GPS base station (or reference station) regularly broadcasts its differential corrections compensating various GPS error sources to the rover (i.e., RTK GPS receiver on a fixed or mobile platform). RTK GPS positioning accuracy is dependent on the distance between base station and rover. For example, the NovAtel RT2W provides 2 cm horizontal accuracy to the rover within 10 km from the base station; however, the accuracy is degraded to 8 cm when the distance is over 10 km due to GPS error sources such as the ionospheric (dominant error over 10 km) and tropospheric delays. To provide accurate RTK GPS accuracy at a range of over 10 km, we might need more RTK GPS base stations to form a network RTK [18]. Or the Nationwide Differential GPS (NDGPS) could provide an alternative.

NDGPS is an expanded service of the Maritime Differential GPS designed to cover the entire surface area of the United States and provide 10 m positioning accuracy to surface users [19]. To date, there are 37 operational National DGPS sites (Figure 5). The service is operated to the RTCM SC-I04 broadcast standard (RTCM: The Radio Technical Commission for Maritime Services) via radiobeacon frequencies (300KHz); the standard reserves its Type-18 and Type-19 messages for RTK operation. If the 37 operational NDGPS sites (i.e., DGPS base station) broadcast the message types, then terrestrial and coastal areas will be covered by the RTK service.

Figure 5. NDGPS coverage (Source: USCG Navigation Center)

D. Application of the RTK Pointing Method

A bi-connected FSO network (a ring network) will be one of the applications for the RTK pointing method. Each node in the bi-connected ring networks has two FSO transceivers. An FSO network is characterized by frequent changes in its link states due to the dynamic performance of optical wireless links, which depend on the effect of node mobility and atmospheric obscuration (e.g. dense fog, dust, or snow). Thus it must be capable of autonomous physical and logical reconfiguration responding to degradation in one or more links. We call this Topology Control [1]. This reconfiguration occurs both physically, by means of pointing, acquisition and tracking (PAT), and logically, by using autonomous reconfiguration algorithms and heuristics. For example, Figure 6 illustrates how topology control works to handle a degradation problem. Because of a sudden change in link or traffic states, physical reconfiguration is necessary. Then, the reconfiguration algorithms yield an optimal topology that minimizes cost or congestion in the network. Subsequently, the PAT technique creates a new topology in accordance with the solution. The problem of finding an optimal topology is known to be a computationally hard problem [20] [21]. References [22] [23] describes a fast heuristic method to find the optimal topology in near-real time with respect to achieving single objective performance (i.e., network congestion minimization) [22] or of multi?objective performance combining network cost and congestion minimization at the same time [23].

Figure 6. Figure 6. Degradation scenario and respective action taken by topology control (Example bi-connected ring topology with directional optical link).

ACKNOWLEDGEMENT

This research was supported by the National Science Foundation.

Footnotes

"No Data Available"

References

1. Davis CC, Smolyaninov II, Milner SD, 2003, "Flexible optical high data rate wireless links and networks", IEEE Communications Magazine, March.

2. Milner SD, Thakkar S, Chandrashekar K, Chen W, 2003, "Performance and Scalability of Wireless Base-Station-Oriented Networks", Invited paper Mobile Computing and Communications Review

3. Ho TH, Davis C, "Three-dimensional Optical Pointing System Encoded by Radial Trifocal Tensor", Proceedings of SPIE 2006

4. Ho TH, Trisno S, Smolyaninov II, Milner SD, Davis CC, "Studies of Pointing. Acquisition and Tracking of Agile Optical Wireless Transceivers for Free Space Optical Communication Networks", Optics in Atmospheric Propoagation and Adaptive Systems VI Proceedings of SPIE, vol. 5237, pp.137-158

5. Epple B, "Using a GPS-aided Inertial System for Coarse-Pointing of Free-Space Optical Communication Terminals", Proceedings of SPIE 2006

6. Wilkerson BL, Giggenbach D, Epple B, "Concepts for Fast Acquisition in Optical Communications Systems", Proceedings of SPIE 2006

7. Yee R, Robbins F, 1998, "Inertial Pointing and Positioning System"

8. Cohen C, McNally BD, Parkinson BW, "Flight Tests of Attitude Determination using GPS compared against an Inertial Navigation Unit", ION National Technical Meeting

9. Lachapelle G, Cannon ME, Lu G, Loncarevic B, 1996, "Shipborne GPS Attitude Determination During MMST-93. ST-93", IEEE Journal of Oceanic Engineering 21, no.1

10. Buick R, 2006, "White Paper: RTK base station networks driving adoption of GPS +1/−1 inch automated steering among crop growers", Trimble Navigation Limited

11. Leick A, 1995, Wiley-Interscience, John Wiley & Sons

12. Meyer C, 2000, "Matrix Analysis and Applied Linear Algebra", SIAM

13. Lee S, Yoo C, Shim Y, Kim J, 2001, "Performance Testing of Integrated Strapdown INS and GPS", KSAS International Journal, vol. 2, no.1

14. Lee S, Shim Y, Kim D, Kang C, Yoo C, Tunik AA, Kim J, "RDGPS-based Automatic Landing System for Light and Commuter Aircraft", The 14th IFAC Symposium on Automatic Control in Aerospace

15. Enge P, "GPS Modernization: New Capabilities of the New Civil Signals", Invited Paper for the Australian International Aerospace Congress Brisbane 29 July–1 August 2003

16. http://www.esa.int/esaNA/ESAAZZ6708D_galileo_0.html

17. http://tycho.usno.navy.mil/gpscurr.html

18. http://www.network-rtk.info

19. http://www.navcen.uscg.gov

20. Desai A, Milner S, 2005, "Autonomous Reconfiguration in Free-Space Optical Networks", IEEE Journal on Selected Areas in Communications, vol. 23, no.8

21. Llorca J, Desai A, Vishkin U, Davis CC, Milner SD, "Reconfigurable Optical Wireless Sensor Networks Optics in Atmospheric Propagation and Adaptive Systems VI", J. D. Gonglewski, K. Stein, Proc. SPIE, vol. 5237, pp.136-146

22. Shim Y, Gabriel SA, Desai A, Sahakij P, Milner S, 2007, "A Fast Heuristic Method for Minimizing Traffic Congestion on Reconfigurable Ring Topologies", Journal of the Operational Research Society (JORS). Published online 7 February 2007

23. Gabriel SA, Shim Y, Llorca, Milner SD, 2007, "A Multiobjective Optimization Model for Dynamic Reconfiguration of Ring Topologies with Stochastic Load", Networks and Spatial Economics. Accepted

Authors

No Photo Available

Yohan Shim

No Bio Available
No Photo Available

Stuart D. Milner

No Bio Available
No Photo Available

Christopher C. Davis

No Bio Available

Cited By

Cited by IEEE

1. Lower-order adaptive beam steering system in terrestrial free space point-to-point laser communication using fine tracking sensor

A. A. B. Raj, J. A. V. Selvi

Signal Processing, Communication, Computing and Networking Technologies (ICSCCN), 2011 International Conference on, pp. 699-704, 2011

Corrections

None
This paper appears in:
Military Communications Conference, 2007. MILCOM 2007. IEEE
Conference Date(s):
29-31 Oct. 2007
Conference Location:
Orlando, FL, USA
On page(s):
1 - 7
E-ISBN:
No Data Available
Print ISBN:
978-1-4244-1512-0
INSPEC Accession Number:
Digital Object Identifier:
10.1109/MILCOM.2007.4454831

Text Size