Rapid tactical deployment capability of a transportable optical ground station

Abstract

In this paper we present TeraNet-3 (TN-3), a transportable optical ground station (TOGS) comprising the third node of the TeraNet optical ground station network in development by the University of Western Australia (UWA). In an effort to improve on the versatility granted by existing TOGS systems, TN-3 is capable of rapid and precise star-based calibration in daytime conditions in the absence of stability hardware. By using a consumer-grade inertial measurement unit and dual camera system, the TOGS can be deployed and ready for free space optical communications (FSOC) to low Earth orbit (LEO) within 5–10 minutes from arrival on-site. In testing the system has demonstrated on-sky root mean square (RMS) pointing errors of 3.54 arcseconds (17.2 $\mathrm{\mu }$rad) and a mean RMS closed-loop tracking error of 1.14 arcseconds (5.54 $\mathrm{\mu }$rad) across multiple LEO satellite passes. By eliminating the need for site-specific requirements like external power or level terrain, the design of TN-3 facilitates flexible deployment in remote locations. These capabilities introduce a range of novel ’tactical deployment’ scenarios, such as rapidly re-establishing communications in disaster-affected areas, communications operations in contested or GPS-denied environments, disrupting adversarial quantum communications, and utilizing the system’s 17-inch telescope for strategic space situational awareness observations. The versatility of the system is further bolstered by direct compatibility with UWA’s existing pointing, acquisition, and tracking systems for FSOC links to land-, air-, and sea-based targets. Together, these capabilities greatly extend the operational versatility granted by existing TOGS systems.

Introduction

By utilizing the same optical frequencies already used in fiber optic networks, free space optical communications (FSOC) can offer a host of advantages over traditional radio frequency systems. For wireless communication, an increase in carrier frequency from radio (kHz-GHz range) up to optical (THz range) significantly alters several key signal characteristics. The most advantageous of these is achievable data rate, which scales with frequency¹. As such, the increase to optical frequencies enables a dramatic increase in achievable data rate, with current FSOC records reaching Tb/s over a single laser². The shift to optical frequencies also dramatically affects beam propagation characteristics. Due to the inversely proportional relationship between carrier frequency and the diffraction limited divergence of a wireless transmission, optical frequencies produce beams orders of magnitude narrower than those of radio³. This has the consequence of increasing power density, enabling greater efficiency and lessened size, weight and power (SWaP) requirements, as well as a lowering probability of intercept for sensitive communications given the smaller footprint which must be intersected. This drastically narrower beam however, also places much stricter requirements on a pointing, acquisition and tracking (PAT) systems used for optical communications, which now need to be orders of magnitude more precise than their radio counterparts.

The largest practical difference between radio and optical communications lies in their level of atmospheric interaction. With their comparatively large wavelengths, radio transmissions have minimal interactions with atmospheric particles, allowing them to effectively pass through clouds and rain⁴. Optical wavelengths however, are highly susceptible to atmospheric scattering, absorption, and refraction. This leads to significant attenuation, particularly during adverse weather conditions such as thick clouds, heavy rain, or fog which can degrade or entirely disrupt the transmission of a signal^3,5,6.

Despite the challenges, the data rate and SWaP advantages of optical communications make them ideal for ground-to-space satellite communications, which are subject to increased demands each year as the amount of data generated by on-orbit sensors increases. To date, the construction of traditional optical ground stations (OGS) sites has been the focus of the vast majority of ground-to-space FSOC development. However, transportable optical ground stations (TOGS) have been presented as a versatile alternative with some key advantages^7–9. Given adverse weather conditions can disrupt the operation of any single OGS for potentially days at a time, a widely proposed solution to increase reliability and ensure communications with on-orbit assets is the development of OGS networks, comprised of multiple geographically-diverse fixed-OGS^10–13. These OGS are typically modeled on traditional astronomical observatories, requiring the procurement of land and development of permanent infrastructure. This process can take several years with costs reaching into the millions of dollars. The prospect of constructing several of these facilities to achieve reliable communications can make the transition to optical communications an unattractive prospect for a wide variety of operators across scientific, commercial and military industries. An appropriately designed TOGS system could grant an operator the ability to quickly relocate to avoid adverse weather conditions and maintain reliable communications for a fraction of the cost of a fixed-OGS site, especially in a region such as Western Australia, where a system could be redeployed as far as 2000 km away in under 24 hours.

While work has been conducted by various research groups and private companies into vehicle-based short-to-medium range FSOC using low SWaP terminals, only a small number of full-size TOGS with sufficient aperture sizes to perform the role of an equivalent fixed–OGS site have been reported^{7–9,14–17}. Additionally, all of these existing TOGS have been developed strictly with a focus on short-to-medium term scientific campaigns^7–9. Two of these systems have been designed to be delivered by road, in a truck or van, while in the case of Saito et al., the TOGS is centered around a cargo truck itself^7–9. Given the scope of their use, they can rely on curated deployment sites equipped with power connectivity and flat stable terrain, further supplemented by stability hardware such as auto-leveling platforms and pneumatic legs^7–9. In addition, as these systems are not designed for rapid deployments, they have not reported any specific daytime calibration capabilities, which greatly limits their versatility for same-day FSOC deployment operations.

At the University of Western Australia (UWA) we have developed TeraNet-3 (TN-3), a TOGS comprising the third node of UWA’s TeraNet OGS network. To build on the capabilities reported by existing TOGS systems, TN-3 is a completely modular system, capable of rapid and precise star-based calibration in daytime conditions without the need for external stability hardware, power, or level ground, allowing deployment in challenging and remote terrain. This facilitates a range of rapid tactical deployment scenarios previously inaccessible to TOGS systems. In this paper we discuss the hardware and software methodology of the system, before presenting results of the system’s rapid daytime deployment method and closed-loop tracking ahead of a discussion on the efficacy of these methods and potential future work.

Method

System overview

The TOGS is centered around a Jeep Gladiator, facilitating deployment to remote areas and in challenging terrain conditions (See Fig. 1). The telescope hardware consists of a PlaneWave Instruments CDK17 OTA and L-500 mount. To raise the telescope for use, the system features a bespoke lifting mechanism. This interface provides three states: stowed for transport, connected but lowered, and connected and raised. In the stowed state, the telescope system rests on a foam pad to lessen the shocks transferred to the OTA during transport. The transition between the stowed state to raised for use takes approximately one minute, when stowed, the system takes up 1.5 m$\times$1.5 m$\times$1.0 m in volume.

Fig. 1.

A graphic representing the operation of the TN-3 system in a daytime remote deployment scenario. Visible on the rear of the vehicle is the PAT systems inertial measurement unit (IMU), wide-field of view (FOV) camera system (denoted WF camera) and optical tube assembly (OTA). These elements are enumerated in order of use in the pointing model attainment procedure. Visible in the center of the image is a representation of an outgoing laser to the on-orbit asset.

The PAT design ethos of TN-3 was to avoid the need for specialized stability hardware, such as pneumatic legs and/or auto-leveling platforms, and instead rely on optical design and techniques such as closed-loop tracking coupled with tip/tilt corrections to compensate for the effects of wind or other small scale movement on the suspension. In addition to reducing setup time and complexity of the system, avoiding such hardware removes limitations regarding the range of inclinations at which the system can be deployed. Crucially, it also allows the system to be entirely modular and able to be removed from the tray of the Jeep and then be placed onto any suitably sized vehicle by using a forklift or pallet lifter. In this scenario, the main operational consideration is the angular movement of the host vehicle on its suspension caused by wind gusts. This can be accurately compensated for by suitable tip/tilt design within the utilised optical payload.

For the geopositioning required to accurately build a pointing model and track satellites via orbital elements, the system features a GPS receiver. This is supplemented by a celestial geopositioning process developed by UWA for use in GPS-denied or otherwise contested environments. While the details of UWA’s implementation are outside the scope of this paper, existing implementations have been reported in the literature (eg. Pierros¹⁸).

The PAT system features an IMU and a wide FOV camera to supplement vision through the OTA’s main aperture. The IMU consists of an accelerometer, gyroscope and magnetometer, whose measurements are fused using a Kalman filter on-board the device to output an absolute orientation estimate. The wide-FOV camera was previously used by UWA to demonstrate coherent FSOC links to land, air, and sea based vehicles on a system using a similar PlaneWave Instruments mount^14,15,19. As such, the software is directly compatible, and the PAT system of TN-3 is immediately capable of the tracking and acquisition required for these links, which can proceed in time-poor scenarios without a full-pointing model given the positive identification granted by the wide-FOV lens and initial IMU orientation estimate. Given the need for absolute heading values attained by the IMU’s magnetometer, the IMU has been placed on a 60 cm aluminum boom (visible in Fig. 1), in order to lessen the influence of nearby ferromagnetic elements of the vehicle, as well as the magnetic field generated by the direct-drive motors of the mount.

Control is administered via custom software developed in Python, which is run on an Intel NUC placed inside the base of the L-500 mount. All necessary functionality to operate the PAT of the TOGS is controlled via the software’s graphical user interface (GUI). This includes but is not limited to camera control, pointing model acquisition control as well as target identification for closed loop tracking. It is envisaged that the tip/tilt system of the optical payload will run on a separate single board computer to ensure its operation is computationally unburdened by the operation of the PAT system.

The system is powered via a portable power station that can be charged by the host vehicles engine. The PAT system consumes $\simeq$100 W, meaning the modest consumer grade power station used for initial testing (see Table 1) can facilitate over eight hours of continuous use. It should be noted total power consumption will be subject to the additional draw of computing and communications equipment. The details of components critical to the operation of the system are shown in Table 1.

Table 1.

The details of components critical to the operation of TN-3.

Component	Details
Platform	Jeep Gladiator
Telescope system	PlaneWave Instruments CDK17 (43 cm diameter) PlaneWave Instruments L-500
Hardware interface	Bespoke lifting mechanism
Power supply	VTOMAN Jump 1500x portable power station
Processing unit	Intel NUC12WSKi5
Geopositioning	GPS receiver: SparkFun GPS-RTK-SMA Breakout - ZED-F9P or Celestial based
Inertial measurement unit	YostLabs 3-space USB/RS232
Wide FOV camera	FLIR Blackfly S (BFS-U3-28S5M-C) Kowa LM100JC1MS 100 mm lens Midwest Optical Systems 780 nm longpass filter
Main aperture camera	New Imaging Technologies WiDy SenS 320

Pointing, acquisition and tracking

In order to achieve precise pointing calibration of the TOGS in daytime conditions with comparatively low-cost equipment, a three stage procedure has been developed, with each subsequent stage facilitating finer pointing calibration:

1. Consumer-grade IMU ($\simeq$ degrees)

2. Wide-FOV camera ($\simeq$ arcminutes)

3. Main aperture camera ($\simeq$ arcseconds)

The initial IMU-based pointing estimate is integrated by determining the equivalent altitude and azimuth ’pointing’ of the IMU. The device has been mounted to the OTA and aligned such that orientation of the IMU approximately corresponds to the pointing of the OTA. Using the system time and GPS location this pointing is then converted into right ascension and declination, which can then be introduced as a model point in PointXP 6 via the PWI4 Python API. Given each point is comprised of two coordinates (right-ascension and declination), two model points are required in order to uniquely solve pitch, roll, and heading values.

The ability of the subsequent camera stages to detect unresolved sources in daytime conditions is primarily limited by both the surrounding sky brightness (largely caused by Rayleigh scattering) and point spread function (PSF) size. To reduce the observed sky brightness, a 780 nm long-pass filter is used on the wide-FOV camera. In the case of the main aperture, the spectral response of the shortwave-infrared camera is sufficient in avoiding the bulk of the scattering component without modification. Given the PSF size determines the area over which light received from the star is distributed, the smallest possible PSF assures the highest possible signal-to-noise with respect to the background sky²⁰. In this regard, the large PSF of the wide-FOV camera limits its use to $\le$20 of the brightest stars in infrared regardless of exposure time, while the main aperture camera with its large OTA is capable of identifying over one hundred stars.

Given the limited number of stars available in daytime conditions, traditional plate-solving has been abandoned in favor of an approach utilizing the astronomical coordinates of individual stars as points in the model. By using the initial IMU stage estimate, the first two to three stars are identified in the FOV of the 100 mm camera (5.0$\times$3.8 deg) and guided onto the in-frame position corresponding to the much narrower FOV of the CDK17 main aperture (5.6$\times$4.5 arcminutes). The star is then centered before the point is added to the model

To aid in this initial stage, the GUI provides the ability to automatically slew to the next star based on a scoring system that determines the best candidate. The score is dependent on both the stars’ I-band apparent magnitudes and the calculated surrounding sky brightness using the CIE Standard General Sky model²¹. Star coordinates and magnitudes are provided by a locally stored copy of the XHIP catalog²².

After these initial stars, the accuracy of the model is sufficient in placing target stars in the FOV of the main aperture directly, eliminating the need for further use of the wide-FOV camera. At this stage, due to their comparatively low accuracy, the IMU points are removed. The desired number of stars can then be centered and added as points using the main aperture until a satisfactory model accuracy is achieved.

In order to streamline the process, this final stage has been completely automated. The aforementioned scoring system automatically determines and slews to acquire the most suitable target. The star is then identified in-frame using machine vision and centered by a proportional, integral, derivative (PID) controller sending corrective movements to the mount.

Once the pointing model has been attained, satellites can then acquired using their orbital parameters as is convention. Given the infrared spectral response of the main aperture camera, satellite communications and beacon lasers can be imaged through the main aperture. This allows for the system to utilize the lasers themselves as a target for closed-loop tracking to ensure precise alignment during the communications phase. This is performed using the same machine vision and PID controller technique described above which operates at a rate of $\simeq$30 Hz. This not only ensures received power is directed accurately into communications equipment on the ground segment, but also the accurate targeting of uplink lasers. To simulate this PAT strategy, all works conducted feature a 90/10 beam-splitter in the main aperture, with 90% of incoming light being discarded to simulate a realistic distribution of incoming light to both the PAT system and communications hardware.

Results

In order to test the efficacy of the PAT system, TN-3 has been repeatedly deployed in daytime conditions. To quantify the time required to build a model of varying accuracy, the automated model building algorithm was modified to survey open-loop pointing error of the cataloged stars. Two models were built on different days with points progressively removed at random in order to probe the accuracy of smaller models. A total of 1434 measurements were collected over the course of two days. During this period, the Australian Bureau of Meteorology recorded peak wind gusts of 18.5 km/h and 13 km/h during the first and second data runs respectively at the Perth Airport weather station located approximately 3 km from the observation site²³. The vehicle was deployed in the same location during both days, facing approximately 47 deg away from north and on a fairly low inclination of 2.7 deg to the north-west. The data set is presented in Fig. 2.

Fig. 2.

Convergence of open-loop pointing accuracy versus number of model points. The upper and lower dotted lines represent the 5th and 95th percentiles for a given number of points while the solid line represents the RMS. Secondary axes are approximate attainment time (top x-axis) and fraction of FOV (right y-axis). Approximate attainment time was calculated using an allowance of 90 seconds for the IMU stage and subsequent two points requiring the wide-FOV camera. Time per point for subsequent stars was taken as 15 seconds which is derived from the median time between subsequent points in the data.

Pointing models with 30 points yield root mean square (RMS) errors of 3.54 arcseconds (17.2$\mathrm{\mu }$rad). Given the automation of the final stage of the model procurement procedure has reduced the median time between points to $\simeq$15 seconds, this can be achieved in approximately nine minutes (This includes a 90 second allowance for the initial two to three stars). The data also demonstrates smaller models containing only four points (requiring an attainment time of approximately 2.5 minutes) are sufficient in placing targets within the central third of the main aperture FOV.

The closed-loop satellite tracking system has been extensively validated using sunlit satellites in low Earth orbit (LEO). Given the unresolved nature of both sunlit satellites (with certain exceptions) and the intended satellite optical signals, the machine vision algorithm developed for star identification can be applied without modification. The mean RMS tracking errors across 5 LEO passes was 1.14 arcseconds (5.54 $\mathrm{\mu }$rad). An example of such a tracking test is shown in Fig. 3.

Fig. 3.

Exemplar daytime LEO tracking data. Top left: Tracking errors in both altitude and azimuth axes. Bottom Left: Angular rate reconstructed from inter-frame telescope positions. Right: Polar plot illustrating the path of the satellite during the closed-loop tracking.

Discussion

The results presented demonstrate the ability of TN-3 to rapidly attain pointing calibration and closed-loop track LEO satellites in daytime conditions in the absence of stability hardware. This ability, in addition to the off-road capabilities of the host vehicle and lack of site requirements such as power or level ground, allows for the rapid tactical deployment of the system to unvetted remote areas in daytime conditions, a scenario previously inaccessible to existing TOGS systems. Use cases for such a capability include rapid deployments to disaster-affected areas to reestablish communications, as well as operations in contested military environments, where the system can employ its GPS-denied geopositioning technique. In these contested environments, the precisely calibrated aperture is suitable for both secure and ultra-high bandwidth FSOC communications as well applications in disruption of adversarial quantum FSOC^2,24,25. Benefits for more traditional OGS operators include the ability to rapidly deploy to the optimal location for a given satellite pass, either to maximize the communications window or to avoid adverse weather conditions. In addition to its FSOC capabilities, the ability to rapidly deploy a precisely calibrated 17-inch telescope to remote locations also grants unique space situational awareness capabilities, where an operator can strategically relocate to attain optimal observations of a given object of interest. Furthermore, with an approximate hardware cost of $121,550 USD, the construction of such a system represents a fraction of the cost of a fixed-OGS such as TeraNet-2, which is also under construction by the authors. The bulk of the cost of the TN-3 TOGS system comes from the host vehicle and lifting mechanism. For a fixed-OGS this is replaced by the costly procurement of land and construction of an observatory building. The additional hardware to enable remote deployment, geopositioning and to facilitate the three-stage pointing model attainment process only represents 1.8% of the total system cost. It should be noted the current estimated cost does not include future planned developments such as FSOC modems and the tip/tilt hardware, however as both TOGS and fixed-OGS incur these costs it does not impact any direct comparison.

The final observed on-sky RMS of 3.54 arcseconds after 10 minutes from arrival on site can be considered more than adequate for ’blind acquisitions’ where the satellite will not be visible as a target for closed-loop tracking, prior to accurate uplink beacon targeting. Although this error is larger than that of a contemporary fixed-OGS systems (i.e. 1.76 arcseconds RMS achieved by DLR’s Optical Ground Station Oberpfaffenhofen), it still represents an improvement on existing TOGS systems, with the TOGS described in Saito et al., achieving 7.92 arcseconds after 30 minutes of pointing analysis, post-mechanical setup^7,26. Data from the model size vs accuracy survey shown in Fig. 2 demonstrates a scalable approach to calibration, where the level of accuracy can be tailored to setup time and satellite acquisition type requirements. In time-critical situations, the system is capable of $\simeq$2 arcminute accuracy in approximately 1.5 minutes of calibration time. Despite the comparatively poor accuracy, such models can still be considered acceptable given a reasonably diverged uplink beacon (for example 1 mrad or $\simeq$3.4 arcminutes used by Rödiger and Schmidt²⁷) or in downlink-only scenarios such as DLR’s Flying Laptop where the target is immediately visible for closed-loop tracking²⁸. The reliability of such an acquisition strategy can be easily bolstered with the inclusion of a lens in the optical path of the main aperture camera to increase the FOV to provide additional compensation for inaccuracies in orbital parameters. Such a lens has been tested in the system and provided a 2.6$\times$ increase in FOV. The observed closed-loop tracking accuracy of 1.14 arcseconds can be considered comparable to fixed-OGS systems with DLR’s Neubiberg OGS achieving “below” 3 arseconds after tip/tilt corrections which have yet to be implemented on TN-3²⁹. For this reason, the authors anticipate comparable performance to fixed-OGS systems once the tip/tilt system has been fully integrated.

The inclusion of a high-precision IMU could potentially eliminate need for intermediary wide-FOV stage and act to simplify the procedure and further shorten the PAT calibration time. To avoid the need for additional scanning, which would counteract the calibration time benefits, the IMU would need to be accurate to within a few arcminutes. This is a difficult endeavor, especially in absolute heading.

During testing of the pointing model attainment process, it has been found that the limited number of stars visible in the wide-FOV camera greatly aids in positive identification of the desired stars. Given the large angular separation between sufficiently bright stars, the potential for misidentification is largely avoided. For deployments at night, the same benefit can achieved by sufficiently lowering the cameras exposure time.

For the subsequent main aperture stage where over one hundred stars are detectable, stars with $\ge$6 apparent I-band magnitude neighbors closer than 15 arcminutes were removed from the locally stored catalog to avoid the need for more sophisticated target identification. With this condition applied, the automated final stage of pointing calibration procedure is reliable and typically requires no intervention. However, one limitation of the current automation system is that regions of the sky which are obscured by trees and/or buildings in certain deployments are still considered as viable targets, as it lacks the necessary information to disregard them. While this can be considered an edge-case, an all-sky camera could vet portions of the sky, so stars whose apparent positions fall within regions deemed inaccessible can be eliminated as model point candidates. In addition, while computationally inexpensive, the simplicity of the currently employed target identification algorithms comes with a greater risk of misidentification than a more sophisticated procedure employing a trained neural network. It is the intention of the authors to investigate this as an upgrade to the system in the future. Future work can also aim to deploy the system in a greater range of inclinations to confirm the systems ability to attain accurate models, although the default ability of the mount to be deployed in equatorial configurations all but assures this, with the caveat of model-fit behavior.

Conclusion

We have demonstrated the ability of TN-3 to be rapidly deployed in daytime conditions, attaining an on-sky pointing error of 3.54 arcseconds (17.2 $\mathrm{\mu }$rad) RMS after 10 minutes from arrival on site and achieving an average 1.14 arcseconds (5.54 $\mathrm{\mu }$rad) RMS closed-loop tracking error across multiple LEO satellites. The versatility of the system is further bolstered by direct compatibility with UWA’s existing PAT systems for links to land-, air-, and sea-based targets. TN-3’s design facilitates tactical deployments in remote and contested environments by eliminating the need for site-specific requirements like external power or level terrain. Together, these features represent a notable advancement in the versatility of TOGS systems and enable deployment to a range of previously inaccessible scenarios.

Author contributions

A.M. wrote the manuscript with input from all authors, performed hardware research and procurement, developed the pointing model procedure, software, and automation, and conducted all data collection and analysis provided. N.M, S.M and S.S assisted in construction of temporary hardware to facilitate testing. S.W, D.G and S.S provided technical feedback during the development process. S.S conceived the experiment. All authors read and agree to the published version of the manuscript.

Data availability

The data analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.