Annual Reviews in Control 29 (2005) 247–260 www.elsevier.


Control and navigation aspects of the new Earth observation missions of the European Space Agency
Pierluigi Silvestrin
Earth Observation Programmes Directorate, European Space Agency (ESA), ESTEC EOP-FPP, Noordwjik 2200-AZ, The Netherlands Received 2 November 2004; accepted 3 May 2005

Abstract Several new space missions are under development within ESA’s Earth Observation (EO) programmes. By 2009 the following satellites will be launched: Cryosat (ice altimetry); GOCE (Gravity field and steady-state Ocean Circulation Explorer); SMOS (Soil Moisture and Ocean Salinity); Aeolus (atmospheric winds); Swarm (magnetic field). Other prepared projects include the SPECTRA (Surface Processes and Ecosystem Changes Through Response Analysis) mission. These missions are briefly reviewed, emphasizing control/navigation aspects when instrumental to the novel EO data. Two representative examples, GOCE and SPECTRA, are described in more detail. Relevant sensor and actuator technology aspects are also outlined. # 2005 Elsevier Ltd. All rights reserved.
Keywords: Space vehicles; Attitude control; Navigation systems; Spacecraft autonomy

1. Introduction The observation of the Earth from space is a rapidly expanding field. It relies on remote sensing techniques (observations of optical or microwave electromagnetic radiation, in either passive or active systems) or on the use of in situ measurements (e.g., of the magnetic and gravity fields) collected in low Earth orbits, with the goal to derive information on the different components of our planet and on their interactions. In recent years the trend towards a fully quantitative approach to Earth observation (EO) with ever increasing accuracy has accelerated. This is driven not only by the development of observation payloads with markedly improved performance for the more traditional sensors (e.g., imagers and sounders), but also by a renovated scientific approach to the study of the Earth as a complex dynamic system, which triggers the development of new sensing techniques. Control and estimation techniques play a major role both in the spaceborne EO systems and in the related on-ground data processing and interpretation. This paper will provide a brief review of some representative aspects of control and estimation issues with

E-mail address: 1367-5788/$ – see front matter # 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.arcontrol.2005.05.004

reference to the future EO missions of the European Space Agency (ESA). Though a classification of this kind is always rather arbitrary, one can broadly identify a first group of well-established Earth sciences, for instance space geodesy and space oceanography, which set increasingly demanding requirements on space observations. These disciplines have a long tradition not only in the derivation of geophysical information from space data – indeed space geodesy was born with the analysis of the first orbital tracking data in the late 1950s – but also in terms of the mathematical modelling approaches underlying the EO data analysis. The studies of the ‘solid Earth’ and of the fluid parts of the Earth are based on geophysical quantitative models that cover different scales (global, regional, local) and are founded on a well-defined set of physical interactions. This has also been the case for the atmospheric sciences, for which progress in modelling has been initially driven by the desire to improve numerical weather predictions and, more recently, to study the evolution of climate. The difficulties caused by the very different scales involved, ranging from those of turbulent phenomena (spatial and time scales of a few metres and tens of seconds, respectively) to the planetary spatial scale and centennial time scale associated to climate variations, not to mention some related fundamental issues like the intrinsically chaotic nature of weather (Lorenz, 1993), have not impeded progress.

Also these Earth sciences. It is worth noting that a frequent characteristic of the biosphere/land missions is the desire for additional implementation flexibility in defining observation and calibration approaches and plans. In space oceanography. like in the SMOS (Soil Moisture and Ocean Salinity) project (Silvestrin.g. This is mainly driven by the difficulty to define precise system requirements. data storage and transmissions. In the first case they relate to the precise control of the relative positions and pointing of the antennas and the complete baseline(s). A second group of Earth science disciplines differs from those mentioned above in that analysis tools and. these well-established EO sciences now require data of very high quality when traditional sensing techniques are used. with accuracies of small fractions of an arcsec required in particular for the knowledge of the attitude of the baseline. is the preferred technique. is not the only reason why automation is increasingly important in space EO. etc. The SMOS satellite. In order to provide an overview of the topic. in geodesy and solid-Earth geophysics the need to observe cm-level or even mm-level surface displacements over vast areas leads also to interferometric concepts. the Moon).. One specific example is constituted by missions to collect data for hydrological studies. possibly in differential mode. as well as during calibrations carried out pointing at specific reference targets on ground and possibly also in the sky (e. many issues of precise control and estimation arise. setting new requirements on performance and flexibility. Added to this. mesoscale phenomena characterised by spatial/temporal scales of tens of kilometres/days are now of great interest in both scientific research and application terms. Both solutions lead to challenging control problems. & Font. there is a strong ‘technology pull’. Berger. which is deployed and steered (by means of the satellite attitude control system) during the measurement phase. 1. In both cases.g. This will be followed by short reviews of control aspects pertinent to two missions that can be taken as representative of the two groups presented above: the GOCE mission (Section 3) and the SPECTRA mission (Section 5). Another example (SPECTRA) will be shown later in more detail. The latter can be implemented with either a long-baseline (4–12 m) antenna array on a single space platform or with multiple cooperating satellites flying in formation. which comes with its set of control issues related to orbit and formation control (e. 2001). The related space missions require new measurement techniques. the control challenges are evident. methodologies in mathematical modelling are to a large extent still under development. in ground operations and mission data exploitation. which are often fundamental for applications of high societal impact. based for instance on precise spectro-radiometric image acquisitions in a large part of the optical spectrum (‘hyperspectral imagery’) or on microwave radiometry using antenna aperture synthesis. realisations based on antenna arrays distributed over several spacecraft flying in tight formation have also been studied for follow-on soil moisture missions (Astrium. Kerr. especially for multi-instrument satellites where the constraints in managing the limited on-board resources (power supply. which must cope with phenomena difficult to model and observe (from space as well as on ground) such as evapotranspiration in vegetation and moisture variations in the soil. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 Because of the advances made possible by many years of modelling effort and EO data analysis. Alternative Fig. as well as observing systems based on innovative concepts. though now for imaging purposes. SMOS uses a large antenna array (Fig. In both cases. a demanding control of the positions and orientations of the satellites in the formation is needed. & Fielding. The fundamental character- .) are most felt. for instance. in a sense.2 m long arms. The drive towards cost-effectiveness has also made a wider application of automation essential. Burgmann.248 P. carrying a Y-shaped L-band antenna array with three 4. The expanding range of scientific and operational applications.. are rapidly advancing in quantitative model-based investigations and applications. 1). Rosen. They require observations of the sea surface topography achievable only with constellations of radar altimeters or with interferometric (synthetic aperture) radar approaches. the EO programme of ESA will be introduced in Section 2. In the second case. Some brief examples can illustrate this point. 2000). but also by the experience from past missions showing the value of being able to operate these EO sensors in ways not anticipated during the design phase. for instance with the advent of system-on-chip (SOC) approaches and of detector technologies such as active pixel sensors (APS) or those based on the application of micro/nano electro-mechanical systems (MEMS/NEMS). Representative disciplines in this group are those that deal with the biosphere and specifically with land processes. as they face tough problems of complexity of the physical and chemical interactions to be modelled and of scaling from local to global models. Synthetic aperture radar (SAR) interferometry. 2001). Similarly. More examples will be given later. notable especially for sensors and related signal processing.

designed for a costeffective implementation. However it would be perhaps more appropriate to take the launch of the low. 2 gives a basic overview of those most relevant to the climate). It is therefore very apt that one of the first Earth Explorers selected was the ‘‘Gravity field and steady state Ocean Circulation Explorer’’ (GOCE). dedicated to mapping variations of the sea-ice thickness and mass. These were extended with the launch of ERS-2 in 1995. Earth system components of key relevance to climate. 2. The Earth Explorers address the problem of modelling and understanding the Earth as a dynamic system. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 249 istics of these missions will be outlined. These interact in a complex manner. These include in particular the gravitational field. especially for what concerns climate evolution (e.P. augmented by across-track interferometry. 3). more application-oriented. Fig. of which the operational meteorological Meteosat and MetOp satellites developed for Eumetsat can be considered the precursors. . relies on simple cold-gas thrusters and magnetic torquers as actuators. user-driven (hence often single-instrument) missions with emphasis on science. In Cryosat a key requirement for interferometry is that the antenna baseline orientation be precisely known. A basic issue noted early on in the formulation of the programme is the importance of establishing at the outset what can be regarded as the fundamental reference frames for precise quantitative Earth modelling.g. After ERS. GOCE will determine the gravity field to 1 mGal (10À5 m sÀ2) accuracy at 100 km spatial resolution and the geoid. the approach for future EO missions had been re-focussed to include on one hand smaller. adding in particular ozone measurements but also enabling a set of (originally unplanned) observations of land surfaces and their fine displacements with the technique of SAR interferometry (Burgmann et al. to 1–2 cm. This requires using SAR altimetry along-track. the ESA EO programme embarked in the development of Envisat. 2004). traditional satellite control functions can be used to improve the system operation. followed by a discussion of the related control and estimation issues. the sea-level changes or the ice-mass balance if the Earth’s gravity field is poorly known... 2. 1997). The Cryosat interferometric altimeter operates in the Ku-band and uses two antennas on a baseline of $1 m. the hypothetical equipotential surface of the oceans at rest defined by the only action of gravitation. the Earth Explorers. comprising a suite of 10 instruments and extending the ERS science goals especially with respect to atmospheric composition and to spectroscopic observations of the ocean. The first Explorer to be launched will be the smaller mission Cryosat (Fig. and. with orbit track control to $100 m. operational missions. ESA’s Earth observation programme The beginning of EO at ESA could be traced back to the development and launch in 1977 of the first Meteosat geostationary satellite for operational meteorology. It will therefore be described later in some detail. the attitude sensing provides an example of how. By the time of the Envisat launch in March 2002. the land surfaces.8 and $3 cm/year. the largest EO satellite to date. which then included mainly observations of the oceans.. Rial et al. Though the satellite attitude control. as well as some complementary information on the sensors and actuators used or under development. with non-linear processes and feedbacks that are often poorly understood. the ice surfaces and. to be launched within 2005 into a 720 km high polar orbit. in a focussed mission. ERS-1 and ERS-2 were operated in formation. 1992. polar-orbiting ERS-1 (European Remote Sensing Satellite) in 1991 as the start point of a more comprehensive European EO programme. to a smaller extent. will measure the variations in the ice sheet interior and margins as well as in the total ice sheet area. To achieve this. depending on measurement objective and spatial resolution. ESA. on the other hand. Cryosat. comprising many components (Fig. with an error between $0. 2000. since there is no way of modelling accurately the ocean circulation. GOCE provides an excellent example of a mission for the mentioned well-established Earth sciences that requires automatic control applications on several fronts. Peixoto & Oort.

still ensure a pointing stability <3 arcsec/s. Also in SMOS an interferometric technique is used. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 Fig. SMOS (Fig. of newly developed fiber-optic gyroscopes (FOG) with very low angular random walk (<2 Â 10À48/h1/2. in this case a bi-dimensional one. The Cryosat satellite. Fig. as will be outlined. the deviations of the field from that due to a reference ellipsoid). starting from the early models derived from the orbit tracking data of various satellites.4 GHz). The inset shows the placement of star sensors on opposite sides of the antenna bench. especially. The Aeolus mission. an Earth Explorer using the Doppler wind lidar technique to determine vertical profiles of air wind speed up to 20 km height. These show. Its lidar instrument operates in the ultraviolet (355 nm wavelength). which requires a coarse formation control. which are the lesser expensive Earth Explorers. The altitude of its sunsynchronous circular orbit is a rather low 400 km. GRACE (Gravity Recovery And . The GOCE mission The GOCE mission will map the global gravity field to an unprecedented accuracy and will rely on multiple applications of control technology.e. employing radiometry with aperture synthesis in the L-band (1.. The satellite is based on the Proteus ‘standard platform’. based on DORIS (Doppler Orbitography by Radiopositioning Integrated by Satellite) tracking data. 3) provides knowledge of the antenna bench distorsions to $2 arcsec RMS. With the recent missions CHAMP (CHAllenging Mini-satellite Payload) and. 3. is the SMOS mission. in addition to the usual attitude data.6 arcsec over 10 s. and the control requirements are alleviated by the longer wavelength. GOCE and SPECTRA. will be launched in 2009. and the fact that the attitude and orbit control system (AOCS). An interesting feature of the Aeolus’ AOCS is the controlled platform slew to compensate for the Earth’s rotation so that the frequency shift experienced by the lidar signal back-scattered by aerosols and air molecules is due to their local speed. a small constellation of three satellites in near-polar orbits to accurately survey all the sources of the magnetic field of the Earth and its close environment. the improvement in the knowledge of gravity anomalies (i. among others. will be launched after Cryosat. on an amplified scale. though lacking gyroscopes. especially in the gradiometer payload and in the AOCS. The next mission to be launched after Cryosat within the so-called Opportunity missions. are discussed in more detail in the following sections. 3. consisting of a Y-shaped array of 69 radiometer receivers. the AOCS of which had to be adapted because of the large inertia moments of the payload. but the payload is now passive. Swarm. 4 shows a view of the satellite and sketches of various gravity field models. as well as to ensure safe operation and performance in the unusual attitudes for calibration and for normal measurements. Making the two star trackers tightly thermo-mechanically coupled with the antennas (Fig. which does not prevent reaching the desired relative pointing error of 1. being exemplary missions from the two groups identified in Section 1. with the help. 1) will be launched in 2007 into a sun-synchronous circular orbit at 763 km altitude. The two lower Swarm satellites will fly in formation at $450 km altitude and provide an East–West baseline of $150 km at equator crossing. Among the larger Earth Explorer missions prepared for a launch after the above mentioned missions. Other Cryosat features include its precise orbit determination to <3 cm RMS in the radial direction. the SPECTRA (Surface Processes and Ecosystem Changes Through Response Analysis) mission is especially interesting in the present context since it represents the archetypal mission for land processes and generates interesting and broad issues of automation. To eliminate any collision risk at polar crossings. GOCE and SMOS. going from the specific AOCS solution to satellite autonomy and mission planning.250 P. 1s). which include a $328 off-nadir tilt of the axis normal to the antenna array plane in order to improve the instrument swath. the phasing of the two satellites will be such that their equator crossings are separated by at least 3 s. Attitude determination of the on-board magnetometer axes to $2 arcsec will be ensured by a three-star sensors set placed at the tip of the magnetometer boom.

The GOCE satellite and the refinement of gravity models (left). The error signal for this one-axis control is derived from the Fig. via the CHAMP and GRACE models. only with GOCE the field knowledge will reach such a definition to enable in-depth investigations of: the Earth interior geophysics. separated by $180 km.01 E HzÀ1/2 (note: 1 E (Eotvos) = 10À9 sÀ2) in the gradiometer measurement bandwidth (MBW) of 0. from the early models (top) to the GOCE model (bottom). 6. The tracking of CHAMP. GOCE will on the other hand benefit from (along-track) drag compensation and fly at an altitude of $240 km. based on GPS. Ion thrusters and tail fins can be seen on the right. where the more sensitive axes of each accelerometer are indicated by solid arrows). Climate Experiment). placed around the spacecraft centre of mass (diagram in Fig. The gravity gradient off-diagonal components are very weak and most of the gravity information is obtained from the radial component (index ZZ in the figures). Two sensing principles are applied in GOCE: satellite-tosatellite tracking and gravity gradiometry by differential accelerometry. This compensation is achieved by electric propulsion with a modulatable ion thruster of 20 mN maximum thrust (a second thruster is flown for fault tolerance). CHAMP and GRACE are more sensitive to the gravity field harmonics at long wavelengths. . 5. GOCE will fly in a sun-synchronous orbit at a very low altitude where drag is substantial ($10 mN). The system goal.005–0. is to retrieve the gravity gradient tensor components such that the error spectrum does not exceed ¨ ¨ 0. whose principle is sketched in Fig. and the dynamics of ice sheets. The five independent components of the local gravity gradient tensor. which is symmetric and traceless. and their relative motion is observed with a precision of $1 mm in range and $10 mm/s in range rate by means of a Ka-band microwave tracking system. In these missions no active compensation of the drag force is applied. allows precise orbits to be recovered ($3 cm RMS in the radial direction) despite the relatively low altitude ($350 km). realised mainly as cooperations between Germany and the USA. 4. The innovative measurements from GOCE are those of the gravity gradiometer. the change in sea-level. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 251 Fig. However. The GOCE measurement principles: high-low tracking with GPS and gradiometry from differences of accelerations caused by gravity anomalies. derived from the mission goals stated in Section 2. are retrieved from measurements of the tiny differential accelerations between pairs of well-shielded proof masses (PM’s).P. the (steady state) ocean circulation. This high-low tracking concept is augmented in GRACE by that of low-low tracking: the two identical GRACE satellites fly on the same orbit. Satellite-to-satellite tracking is applied in the CHAMP and GRACE missions. gravity models have much improved.1 Hz. so active compensation in the flight direction is necessary to provide both orbit maintenance and reduction of the residual gradiometric errors due to differences in alignments and gains in accelerometer pairs. 5. so enhancing the sensitivity to gravity anomalies of the gravity gradiometer and GPS tracking data.

which are based on phase synchronous detection. The six accelerometers are arranged in a diamond configuration with baselines of 0. This is also limited by the quantisation noise of the 24-bit output analogue-to-digital converter and by the CDS internal noise. The three solar panels. Satellite length and body diameter are $5 and $1 m. with force actuation provided by electric fields applied using the same CDS electrodes. To reach the specified performance. are fixed. Stabilizing tail fins shift the centre of pressure and ensure aerodynamic stability close to the nominal attitude. a major limiting factor for the sensitivity. The configuration of the satellite (Fig.1. The configuration of the proof masses of the gradiometer accelerometers (solid arrows indicate the ultra-sensitive axes). For the planned launch in 2006. Each accelerometer contains a platinum-rhodium alloy PM of 0. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 Fig. 7. which is electrostatically suspended and kept precisely at the centre of the accelerometer electrode cage. Coupled with the measurement range of Æ6 Â 10À6 m sÀ2. while the mass is $1100 kg. Control authority. 3. baseplate (left). The GOCE gravity gradiometer payload Besides the gravity gradient components. Capacitive displacement sensors (CDS’s) using eight electrode pairs (for redundancy) measure the PM position and orientation. PM.252 P. The general layout can be seen in Fig. a lifetime of $20 months is expected. with sensitivities of $10À9 and $10À10 m sÀ2 HzÀ1/2. Foulon. this introduces undesired mechanical stiffness and damping of the PM motion. The electrostatic suspension voltages provide a direct measurement of the accelerations required to maintain the PM at the centre of the electrode cage. one of which is mounted on the satellite body. each accelerometer has a sensitivity of $10À12 m sÀ2 HzÀ1/2 on its two sensitive axes. close to the minimum of the solar activity and therefore also of drag. including electrode plates. The accelerometers have been developed through ESAsupported research at the French institute ONERA (Touboul. limits the mission duration. . hence of the non-gravitational accelerations applied on the satellite. this results in a challenging dynamic range close to 107 and a correspondingly tight linearity requirement. 6. derive from the early Fig. 4) has a slender shape to reduce the air drag. 7. 6). so that its 6 degreesof-freedom (DOF) are controlled in closed-loop. and the electrode arrangement ensuring 6-DOF control (right). which will depend on the actual drag levels experienced.32 kg mass. Cross-talk between detection and actuation signals is avoided since the CDS’s. showing an accelerometer model and the electrode arrangement. The accelerometer core. the gradiometer data is used on-board to estimate the satellite angular rates and the perturbing angular accelerations. Each PM is electrically grounded by a gold wire of 5 mm diameter in order to eliminate charging effects and consequent parasitic (Lorentz) forces on the PM. rather than on-board fuel. However. & Willemenot. in-orbit fine loop tuning and injection of calibration signals. providing flexibility for e. 1999).g. operate at 100 kHz.5 m (Fig. The controllers for all the PM control loops are implemented in digital signal processors. common-mode accelerometer measurements in the flight direction. far above the accelerometer measurement bandwidth. The accelerometers used on CHAMP and GRACE. respectively. respectively.

flow valves). This is within specifications. in turn affecting the recovery of the radial and across-track gravity gradient tensor components. Gradiometer engineering and drag-free and attitude control The gravity gradiometer (Fig. The panel in the middle is precisely controlled in temperature. is shown in Fig. The expected performance for the angular accelerations. In addition.5 mN thrust each). The gradiometer configuration (Fig.g. 1999). 2003). which are beyond the scope of this paper. The GOCE gradiometer with the accelerometers fixed on the top ultrastable plate. a magnetometer. to minimise the disturbances caused. by: variations in satellite self-gravity. 8) calls for the application of control and estimation techniques in several ways.3. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 253 models developed for the GOCE gradiometer. The precise on-board estimation of attitude. The critical requirements for the DFACS call for angular accelerations and angular rates to be <7 Â 10À8 and 5 Â 10À7 rad sÀ2 HzÀ1/2. but depends critically on the assumptions on the drag fluctuations in the GOCE orbit. 6) is such that this impacts the recovery of the roll angular components. in the MBW. An ion thruster compensates the drag in the flight direction in the bandwidth 0–0. angular rates and accelerations uses the accelerometer and star-tracker data. Only 4 DOF (3 rotations. 9. The accelerometers include interesting control aspects by themselves (e. so tightly coupling spacecraft and payload. No gyroscopes are used. angle of 2 arcsec). 9 (Catastini et al. and the position errors. among others. other measurements are provided by two star-trackers (with noise equivalent Fig. microvibrations (from e. besides the lower roll inertia moment. For instance. 1 translation) will therefore be controlled. it was not possible to develop cold-gas proportional microthrusters or field emission electric thrusters with reliable performance over the full mission lifetime on time for the planned launch in 2006.P.1 Hz. The low noise of this actuator at <30 mN HzÀ1/2 contributes to achieving a residual along-track acceleration <2 Â 10À8 m sÀ2 HzÀ1/2.2. respectively. based on the current estimator and controller designs. provide sufficient performance. scale factors) differences. The measurements used by the drag-free and attitude control system (DFACS) include those of the payload.. about which data is very scarce.g. the one in roll is less so. The actuator set is completed by an additional eight cold-gas (nitrogen) micro-thrusters ($0. This is accomplished by a multi-stage thermal control/filtering scheme. a sun sensor and an Earth sensor. Gradiometer calibration and performance Fig. Angular accelerations in yaw and pitch are well controlled. whereby the gradiometer middle panel is actively controlled to <0. is the lower accelerometer sensitivity along one axis. Josselin. Residual gradiometer errors include those caused by the accelerometer alignment and gain (i. 3. The main cause for this.. The carbon– carbon structure on which the accelerometers are mounted must then be kept extremely stable in temperature. Performance estimates for the GOCE angular acceleration. The attitude actuators are magnetic torquers. the gradiometer baseline stability must be ensured to $10À12 m over 200 s. besides what is in each accelerometer.e. which. 3. An in-orbit calibration . operated in on–off mode for gradiometer calibration as described below. sudden motions of multi-layer thermal insulation (‘‘thermal clanks’’). The gradiometer operation requires high thermo-elastic stability and no moving parts in the entire satellite.01 K over 200 s. in spite of the low control authority. the non-linearities. Though a full 6-DOF DFACS had been initially envisaged.. The gradiometer electronics is on the lower tray. 8.

1. This scheme is an iterative one and has been shown (by simulations) to converge fairly reliably to the actual gradiometer parameters.. support to combined GPS/ Galileo attitude determination. Such modulation and firing are designed to produce perfectly uncorrelated pseudo-random 6-DOF excitations of the satellite dynamics. the goal being to excite evenly and independently all gradiometer imperfections. 2000). 2003). 2004) and features enhancements such as: processing of Galileo signals as well as modernised GPS signals. High-performance Galileo/GPS components The GPS receiver on GOCE is a state-of-art instrument based on the AGGA (Advanced GPS/Glonass ASIC) radiation-tolerant digital integrated circuit. the procedure includes optimal estimations of angular rates and centrifugal accelerations from gradiometer and startracker data. Fischer.. a SOC architecture with onchip processor (LEON-2) and floating-point unit capable of >100 MFLOPS. mainly the quadratic effects. The correlation values are then input to a gradiometer parameter correction algorithm. including on-board Failure Detection. the local closed-loop control of actuators. Errors at high frequency are driven by the accelerometer noise and cannot be reduced improving calibration or DFACS. are then estimated from the average (DC) measured accelerations and reduced adjusting the position set-points of the PM control loops. They in fact invest the whole mission:  at payload level: the 6-DOF digital control of the accelerometer PM. and finally the high-resolution modelling of the gravity field. Isolation and Recovery (FDIR). 36 flexible tracking channels with 10 correlators per channel. based for instance on model predictive control. . & Weigand. the star pattern recognition and tracking in the star-sensor. Performance estimates for the radial gravity gradient tensor component.05 Hz. the GPS signal tracking for sub-mm carrier phase precision. i. the satellite autonomy. Complements on GOCE-related sensors and actuators 4. the fine thermal control. The next version of the chip is under development (Berberich. Fig. the performance is much degraded at f < 20 mHz. For the gradiometer calibrated as described above. For the ideally calibrated gradiometer. For the uncalibrated gradiometer. The method applied to estimate them relies on the pseudo-random modulation (or firing) of the ion thruster (or the eight calibration micro-thrusters.1 Hz for minimum interference from the gravity gradient signals. 2003).  at system level: the attitude estimation from the payload and from other sensors. the precise orbit determination. 4. which are very low above 0.05–0. respectively) during calibration periods. Considerations of system observability and of dependencies among the gradiometer parameters lead to reduce the number of the unknown parameters from 72 (12 per accelerometer) to 38. which is also at the core of atmospheric sounding sensors in the MetOp meteorological satellites (Silvestrin et al. Conclusions Only a part of the control and estimation issues in GOCE has been illustrated here. an improvement at the lowest frequencies (<8 mHz) is being pursued via fine tuning of the attitude control laws. which can only be outlined here and are still being optimised (Lamarre & Floberhagen. the DFACS. A companion radio-frequency chip optimised for precise applications is also under development. The nonlinearities. The GOCE gradiometer calibration problem can be seen as a problem of identification of the parameters of a non-linear system from the knowledge of the output signals and of the statistics of the inputs only.254 P. digital beam-forming and digital downconversion.e. followed by the computation of correlations among retrieved accelerations.4. the performance target is basically met. 10. The calibration involves both in-orbit and onground operations.18 mm CMOS technology. The new chip includes $5 million logic gates in 0. 10 shows a recent performance estimate for the radial (ZZ) gravity gradient tensor component under different conditions (Catastini et al. tracking loop aiding. Since incorrect gradiometer model parameters lead to the imperfect rejection of accelerations in the gravity gradient estimation process. reviving a concept for enhanced attitude determination that could not be realised with GPS/Glonass due to the limited development of Glonass (Silvestrin et al. the gradiometer calibration. Fig. 1996). Griesauer. 3. which are extremely critical for a satellite in such a low orbit. Silvestrin. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 based on the periodic estimation of these gradiometer parameters is then necessary to achieve the final gravity field recovery performance.  at mission level: the a posteriori (on ground) recovery of the gravity gradient tensor and the satellite attitude. The modulation spectrum is bandlimited to 0.. The accelerometer non-linearities are taken care of by generating a train of sinusoidal PM accelerations by means of a specific calibration signal in the accelerometer loops during calibrations. which can be seen as a large-size application of optimal estimation.

For the latter. with a non-contact discharging of the PM. and on needle or capillary emitters with indium propellant.. as for atmospheric sounding by radio occultation (e. e. . Frequency up-conversion in the CDS virtually eliminates the impact of 1/f noise in the CDS electronics. to achieve the best ‘drag-free’ conditions for the PM.e. parasitic capacitances. etc. & Touboul. the specific impulse is very high (>5000 s).g. This approach will be used in fundamental physics missions such as the Laser Interferometer Space Antenna LISA (LISA Study Team. on the mentioned MetOp satellites). Two complementary FEEP technologies have been developed in Europe.3. as in gradiometric missions.g.g. hence the accelerometer MBW. In this way the accelerometric performance is optimised at the expense of PM motion. the electrostatic accelerometers of the GOCE type provide excellent performance with limited accommodation problems. This so-called ‘cold damping’ contributes to preserving the accelerometer sensitivity. Reynaud. In this case the DFACS fulfils the task to ensure the ‘tracking’ of the PM position to the largest possible extent. Electrical noise in the CDS does in fact induce force noise on the PM. are the prime limiting factor for the accelerometer sensitivity. with all components of satellite velocity recovered to <0. The first term provides damping of the PM motion. FEEP (Field Emission Electrical Propulsion) thrusters are close to maturity and very promising: the maximum thrust is $1 mN. with ground GPS/Galileo receiver data availability or not). based on the actual application aims. achieved with much smaller noise than in the case where mechanical damping is applied. the sensitivity could be still improved by several orders of magnitude. The control imperfections then become less relevant and the sensor acts as an inertial sensor rather than an accelerometer (Da Lio. Ion thrusters can provide substantial thrust (tens of mN) and can be well controlled.) that could be still reduced in future accelerometers. and the thrust can be finely controlled in a wide bandwidth (>1 kHz). mainly via the gold wire. 2004). which cannot be eliminated by differential GPS techniques. For applications where data latency up to about one hour is tolerable. To finally achieve this improvement. 1999). constraints and requirements (e. even if this strongly couples spacecraft and PM dynamics. delivery of information in real-time or with some latency. Feedback control of the PM electrostatic suspension is mandatory to stabilize the system and is realised with minimum noise increase. 1992) of extreme sensitivity: $6 pm/Hz1/2 in GOCE. 4. the application of ion thrusters in GOCE has been mentioned. The term proportional to the measured PM displacement defines the closedloop stiffness. A sensitivity improvement of one to two orders of magnitude seems feasible for future GOCE-like missions. parasitic couplings to the PM. 4. 2000) that by eliminating these couplings. It is based on a resonant capacitive bridge concept and designed to obtain quantum non-demolition displacement measurements (Braginsky & Khalili. In this case the motion of the PM is not the primary concern and the PM loop gain can be made large. Both technologies can be scaled up – though in different ways. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 255 The navigation performance achievable by instruments based on these components will depend on the specific orbits. The spectrum of their output noise shows 1/f behaviour at very low frequency because of parasitic damping effects on the PM motion. & Vitale. enhanced (with respect to GRACE) by optical tracking. Both the above cases are of interest for post-GOCE missions.. However. two problems remain to be solved: the output signal dynamics and the noise induced by the CDS noise via back-action. Electrostatic accelerometers and inertial sensors for future gravity field missions Although superconducting gradiometers have also been developed. In this situation the loop gain (in the MBW) must be kept as low as possible to limit the PM disturbance caused by the back-action. Recent measurements confirmed the low noise levels of $10 nN/Hz1/2 above 1 mHz at 20 mN thrust. one proportional to PM velocity and one to PM displacement. for which improved gradiometers and (optical interferometric) tracking systems are being studied (Astrium. in the latter case requiring MEMS technologies – to achieve higher thrusts. The first case is when the goal is to measure differential accelerations. This will allow for a strong reduction of local multipath errors. The former can be solved by placing the accelerometer in the quiet environment provided by a highperformance 6-DOF DFACS. Courty. based on slits with cesium propellant. The second case is when the goal is to minimise the disturbances on the PM. The CDS is the key element of the accelerometer. MicroScope) and qualified for ultra-precise pointing and drag compensation in the low-orbiting EO missions of the next decade. taking advantage of the abundance of measurements with GPS and Galileo. i. Bortoluzzi. 2000) and in low-low satellite-to-satellite tracking systems for gravity field monitoring. for instance because it is placed directly in an inter-satellite optical interferometric measurement path. The electrostatic force applied to the PM is due to two main terms. The stringent requirements for relative position reconstruction to <1 mm that arise in interferometric (X-band) SAR using satellite formations (inter-satellite distances <1 km) will also be met. the orbit can be determined to a few centimeter accuracy. 2004. which is important in assessing the effect of the back-action from the CDS on the PM. with minimum electrostatic actuation in the sensor. Josselin. It has been shown (Grassia.P.2.1 mm/s. though constraints in the regulation of fuel flow and grid voltages limit the maximum slew rate. especially since they operate at ambient temperature. FEEP will soon be used in fundamental physics missions (LISA PathFinder. The CDS performance is then limited only by a few factors (loss in bridge transformer. Electrical thrusters Electrical thrusters are of key importance in missions when very fine pointing control or drag compensation has to be achieved. while for the latter two application cases need to be distinguished.

11. proper sampling is needed in all the dimensions of the phenomena: spatial. temporal. user-selected according to the specific biomes in the visible. Over 100 geophysically representative sites have been identified for repeated. The SPECTRA mission The goal of the SPECTRA mission is to model the role of vegetation in the global carbon cycle and its response to climate variability (ESA.1. considering its $675 km high orbit. 12 shows the sequence of image acquisitions over a site. preceded by cold-sky imaging for calibration. The satellite observes them with a 3-day access. For each site. quick acquisition of optimal attitudes in terms of solar array illumination or thermal stabilisation when not imaging (Fig. which. . adaptive motion compensation and yaw steering. shortwave. 14. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 5. can perform a 258 slew in $10 s. 2004). The four CMG’s are arranged in a pyramidal configuration optimised for the higher agility needs around pitch and roll. 11 illustrates the complexity of the interactions to be modelled. thanks in particular to the use of four Control Momentum Gyroscopes (CMG). shown in Fig. The approach behind SPECTRA is to ‘sample’ the biosphere to build a library of robust models. Two of the OZA’s are not specified a priori: they depend on the specific geometric and sun illumination conditions during passes over a site. Fig. taking into account the past observation history. data downlink (pointing of the antennas for higher transmission rates). The images are in $60 spectral bands. 13). sequences of hyper-spectral images at seven Observation Zenith Angles (OZA) are taken. Fig. the vegetation properties must be measured and models of surface processes (at local scales but suitable for global modelling) developed. spectral and directional. The satellite. The satellite concept An agile satellite is the best answer to the SPECTRA observation requirements. based on optimising the relevant observation parameters. Since the observed electromagnetic radiation is determined by complex interactions at the surface and in the atmosphere. which is used to advantage also for: calibrations (pointing at specific sites and at the Moon and Sun with dedicated apertures).and thermal-infrared spectral regions covered by the spectroradiometer instrument. The pixel line motion is slowed down during acquisition to achieve the radiometric performance needed for precise quantitative modelling. In this way the bi-directional reflectance distribution function (BRDF) of each site is monitored via repeated image sequences. detailed observations. Different biomes are found at these globally distributed sites. Fig. near-. which are high-torque actuators based on single. 5. Their choice is a mission planning task. implies an across-track pointing capability by $358 off nadir. Agility means that the satellite is capable of rapid and precise slewing.or dual-gimballed spinning wheels. Interactions in land surface processes.256 P. To this end.

Manoeuvre parameters are then up-loaded as telecommands. integrated with the guidance profile computation. the computation of guidance profiles for imaging and rallying manoeuvres is done on ground.. Acquisition sequence (seven images at seven OZA’s). Each CMG provides $15 N m s and 45 N m maximum momentum and torque.g. Simulations show good margins for the pointing performance in terms of: line-of-sight (LOS) pointing ($40 m error at nadir). 2001. pointing stability ($1 m error during the pixel integration time). but the CMG-based agility allows for a large flexibility in choosing such angles. ESA. Mission planning and autonomy In the baseline design. two gyroscopes. have been developed and optimised for low computation load in view of a possible on-board implementation to improve satellite autonomy and imagery. On SPECTRA the algorithms for on-board manoeuvre computation. and a GPS receiver. for close-by sites) or conflicts induced by on-board Fig. Example of the operational applications of agility on SPECTRA. which can then be optimised based on science needs.2. 5. respectively. Fig. 13. already tested with the technology satellite PROBA (PRoject for On-Board Autonomy) (De Lafontaine & Vuilleumier. The acquisition of image sequences over the different sites needs the resolution of conflicts. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 257 three magnetic torquers. Slew performance is a constraint in the selection of intermediate OZA’s. The on-board estimation and control are based on extended Kalman filtering of sensor data and on a relatively wide bandwidth ($0.P.5 Hz) controller with feed-forward of slew and guidance profiles. 2003). . Algorithms that accomplish this task. The AOCS also uses: a multi-head high-rate autonomous star-tracker. for instance access conflicts (e. are made more complex by the need to manage the CMG singularities. and co-registration between any image pair in a sequence (<10 m). 12.

The higher bars indicate the total possible number of acquisitions. An OPRQP (Optimisation of Parameters by Recursive Quadratic Programming) optimiser is used to tune the parameters. accounting for operational constraints and yielding the possible acquisition image sequences. based on specified indices. image acquisition and agility). obviously a rather unrealistic situation. Realistic site distribution and calendar of imaging requests are used. The ‘ground’ mission plans are elaborated periodically. scan of the across-track angle range of interest (important for BRDF analysis). 15. for instance one orbit repeat cycle (14 days in SPECTRA). transitions in/out of eclipses. mainly for failure detection and recovery during plan execution. and the path planning optimisation. (c) the analysis of events (accesses to targets and their characteristics with respect to image geometry distorsions and illumination conditions. every 3 days. which is transformed into a telecommand sequence. In addition. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 Fig. e. and is followed by the optimisation of OZA’s and finally by the computation of attitude manoeuvre parameters.) over a specified time horizon. This last step includes: the propagation of constraints (e. a search for the dynamically feasible solutions. and cover a specified time horizon.g. taking into account the constraints from the use of CMG’s. ground contacts. site conflicts are resolved based on quality indexes assigned to each observation event depending on: user requests. (d) the mission plan optimisation proper. Inclusion of cloud information based on weather prediction in such a slow scenario is possible but is of marginal value. 50 sites are observed and 100% utilisation is defined as 25 image sequences per day.g. In general. the weights assigned to across-track angles. largely driven by the agility goal. previous and future angles at closest approach. and other criteria. In this scenario. as needed to compensate for cloud coverage. the benchmarking of the feasible plans. particularly in the context of future EO constellations requiring much more autonomy. The compact configuration of the SPECTRA satellite. The lower bars indicate the scheduled acquisitions that are the minimum to fulfil the specified sampling scenario in cloud-free conditions. resource constraints. re-acquisitions. 15 shows the nominal SPECTRA capacity use from a prototype of the mission planning procedures. This provides the (optimal) baseline sequence of image acquisitions. based on given criteria. In the initial design. iterative process finally yields a mission plan. implemented onboard with autonomous low-level functions. The difference between the two bars for each month illustrates the capability to perform several Fig. e.g. which yields manoeuvre timings. acquisition geometry with respect to specific physical processes. This margin exists mainly because of the satellite agility.258 P. and the computation of corresponding indexes. Use of the SPECTRA observation capacity. As these do not exceed 60% of the satellite utilisation. mission planning involves several steps: (a) the definition of an initial mission scenario. etc. the alternative of on-board (reactive) planning to improve autonomy and mission data return is also being studied. (b) the computation of precise orbits. assuming 50 sites globally distributed. there is still much residual capacity and many more sites could be observed. orbit predictions and possibly manoeuvres for orbit control. Although the initial design of the SPECTRA mission planning assumes that this is carried out on ground. if an unforeseen . in the simulated scenario. The batch implementation of mission planning on ground has some drawbacks that can be too penalising for future imagery missions of an operational nature. 14. This complex. Fig.

but also on estimation/control system concepts. resulting in a sub-optimal plan execution or even in its failure at some point. especially for applications dealing with unplanned events (which are only ‘exceptional observations’ in the SPECTRA context. Permanent features of targets could be recognised and changes of interest detected by on-board processing. Optical head for autonomous high-rate star-sensor. of cloud cover (detected from a ‘first look’ or with a dedicated imager). . eruptions.g. This can be achieved via on-board observation (re)planning. Similarly.. which include recognition of patterns in both spatial and spectral domains. based on robust control concepts in the sensor processing. Obviously any on-board model-based planning capability implying autonomous decision-making must be supported by a robust plan execution capability. More in general. using a model of the satellite that takes into account resource availability. The application of automatic control at the system level already provided substantial payload complexity reduction in the present SPECTRA design (the earlier non-agile designs included complex opto-mechanical payload arrangements). the buildup of water or nitrogen stress in crops.  handling of short-term user requests. Model identification applied to pointing. it may take long until a new plan is available that takes the negative event into account. translating the plan into lowlevel autonomous commands in a safe way. 16. it can be missed.  better imaging performance. will enable:  handling of failed acquisitions or predicted failures because e. the inclusion of on-board image analysis to detect special conditions of the targets (besides cloud coverage) is especially interesting for future monitoring missions with satellite constellations.3. which can then be used for fine-tuned attitude guidance. inter-OH and calibration parameters. is the key to achieve immunity to OH blinding by Moon and Sun and to the graceful degradation when one or two OH fail. fires. though with emphasis on different aspects: satellite robust agility. on-board dynamic (re-)planning and optimisation of observation approaches provide promising avenues for future monitoring systems. The joint ESACNES CMG development has provided both the CMG hardware and the singularity avoidance laws.) but also more subtle changes. particularly in the perspective of future operational missions. for precision agriculture applications. Engineering model of 14 N m s CMG (left). would then be used for generating optimally adapted mission plans on a shortcycle basis. stable coastlines) in order to achieve a fine identification of instrument parameters and an improved LOS modelling. increased read-out flexibility and better anti-blooming capability. Conclusions Like for the GOCE mission. Changes can include catastrophic events (floods. The estimation. given its scientific nature). landslides. Analyses of these scenarios are part of current research. The imaging performance improvement can be achieved for instance realising an automated support to pattern recognition over calibration sites or over natural features (e. Complements on SPECTRA-related sensors and actuators The development of (single-gimbal) 15 N m s CMG’s has been mentioned. automation pervades the design of new multi-/hyper-spectral missions. in addition to robust performance. Beyond mission planning. autonomy at spacecraft level as well as in (on-board) payload exploitation. flexibility in observation planning. The on-going development of autonomous multi-head highrate star-sensors for the AOCS of EO satellites builds on both new detector technology. Data fusion of three Optical Heads (OH) outputs is indeed obtained from adaptive estimation of intra-OH. since the planning process needs to be initiated well before the end of the current time horizon. the predicted state can be rather uncertain.  target-dependant steering. 5. for instance an image acquisition is missed because of a temporary system anomaly or wasted because of cloud coverage.g.P. The maximum CMG gimbal rate (acceleration) of 3 rad/s (6 rad/s2) is enabled by a fast (10 Hz bandwidth) and precise closed-loop control of the gimbal position that preserves the wheel low-noise performance. etc. which allow for an 8 N m s singularity-free momentum sphere. e. if an unforeseen short-lived observation opportunity like a fire or a flood occurs. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 259 negative event occurs. 6. Finally. A shorter response time would clearly be desirable.. namely APS arrays with lower cost. performance constraints and FDIR status. An on-board mission planning capability designed for high responsiveness with dynamic re-planning and supported by onboard event predictions and by real-time on-board guidance profile generation and control. etc. automatically optimised based on geometric and/or radiometric criteria.g. The outputs of such image processing and analyses. Autonomous hierarchical FDIR is realised in the OH and Fig. co-registration and image enhancement could improve performance and decrease development cost.

when the satellite angular rate is 38/s. Hagg. Braginsky. R. E. Parisch. J.. Lorenz. Bortoluzzi. Bagge. M.html. LISA Study Team.. Documents on ‘Interferometry’ at http://earth. Spaceborne GNSS radio occultation instrumentation for operational applications. Quantum theory of fluctuations in a cold damped accelerometer. UK. (2003). Silvestrin. in particular those flying in formations. et al..g.. E. A. S. Kerr. Brodin. Lamarre. illustrating some significant and real examples. 65(1–2). Y.esa. P... S.. Non-linearities. GNSS receiver design for attitude determination. Peixoto. Aardom. Christensen. In Proceedings of the AAS/AIAA Astrodynamics Specialists Conference. From what presented. 118. ESA Internal Report (unpublished). Daly. & Willemenot. After working at the Joint European Torus (JET) research project on nuclear fusion at Abingdon. A. J.. Among the many aspects of such developments. J... reaches a noise equivalent angle of $1. Burgmann. precise guidance. Enabling Observation Techniques for Future Solid Earth Missions... Berger.5 arcsec.. R. J.esa. (1996). The essence of chaos. Multibody model for spacecraft drag-free control of the laser interferometer space antenna technology demonstration mission SMART-2. Pielke.. Silvestrin. Noordwijk.. At ESTEC he worked first as Control System Engineer in the Technical Directorate.... P. & Oort. Claussen.. (1999). 275–285)... De Lafontaine. (2003).. H. which could only be outlined here in broad terms. Cesare. P. Gradiometer In-flight Calibration. & Walsh. at satellite angular rates of 18/s and 10 Hz sampling. Space Technology. European Physical Journal D. ¨ ¨ Silvestrin. American Institute of Physics. (2000). (1992). 45(10). Berberich. M.. (2000). 605– 617. ESASCI(2000)11. F. The Netherlands. Pierluigi Silvestrin received the laurea degree in Electronic Engineering in 1986 from the University of Padova. Cambridge University Press. an absolute pointing error of $5 arcsec and an angular rate noise of $3 arcsec/s. He is currently Head of the Mission and System Studies Section in the same Directorate. & Vitale. Beniston. A.esa. M.. Rial. specialising in Control Systems. S. 24(1). P. Physics of climate.. J. & Font. Final Report of ESA Contract 13570. Climatic Change.  developing observation methods based on highly autonomous satellites. B. P. Cooper. Bonnedal. (2004). P. University of Washington Press.. 28. M. In Proceedings of the ION GPS 2000 13th International Meeting. 16 shows the optical head cross-section. then as Earth Observation Systems Engineer in the Earth Observation Directorate.. feedbacks and critical thresholds within the Earth’s climate system. In Proceedings of the Second International Workshop on GOCE. Reynaud. Paris. Grassia. Reports on SPECTRA and other candidate Earth Explorers at http://www. E. 11–38.. Da Lio. common electronics. 11–14. et al. (1999). (2004).. M. Griesauer. (2000). ESA’s second earth explorer opportunity mission: the soil moisture and ocean salinity mission—SMOS.. & Vuilleumier.. Silvestrin / Annual Reviews in Control 29 (2005) 247–260 Catastini. (2004). Annual Review of Earth and Planetary Sciences. ESA SP-381 (pp. P. respectively. The GOCE end-to-end system simulator. P. D. M. & Khalili. P. E. 7. V. PROBA documents at http://www. Autonomous navigation and guidance on-board earth observation satellites. Conclusions This paper has shown how the application of automatic control technologies enables new approaches in Earth observation from space. Laser Interferometer Space Antenna (LISA): A Cornerstone Mission for the Observation of Gravitational Waves..  extending the performance envelopes of GNC sensors and actuators. More extensive and advanced applications of automation technology in EO missions stem not only from technological progress. et al.260 P. M. AGGA-3: A next generation GNSS baseband ASIC. S. Acta Astronautica. Josselin.. The sensor performance. Quantum measurement (Chapter X). but especially from new developments in EO concepts and requirements. (2001). Architecture mixte pour les accelerometers ultrasensibles dedies aux mission spatiales.. (1992). which grow to 13. J. In Proceedings of the Third International Conference on Spacecraft Guidance. Cox. . F. ESA-ESRIN. References Astrium. (2000). & Fielding. as was the case for a long time. with responsibility for the conceptual and preparatory studies of new Earth observation missions and systems and for related technological developments. B. Courty.. Synthetic aperture radar interferometry to measure Earth’s surface topography and its deformation. P. M. as well as for the ability to operate systems in more flexible ways. Rosen. Electrostatic space accelerometers for present and future missions. ONERA Technical Note NT 1999-8. In Proceedings of the ION GPS 2004 17th International Meeting. Dumontel. 15 and 23 arcsec/s. 8. Italy. 57–71. S. in 1987 he joined the research and technology centre (ESTEC) of the European Space Agency at Noordwijk. ESA. ESA ESTEC. ESA. Touboul. Final Report of ESA Contract 16668.. Saponara. 101–110. R... de Sanctis. Canadell. Carlstrom. (2001). Fig. J. (2001). P. G. V. & Weigand. Foulon... & Floberhagen (2003). A. S. IEEE Geoscience and Remote Sensing Society Newsletter. M. model parameter tuning based on fusion of attitude and imagery data.. Fischer. (2004). (1993). The fully autonomous start-up is achieved in <4 s and the tracking of stars is ensured up to high rates (108/ s).  applying more extensively advanced autonomy and FDIR techniques. J. R. Navigation and Control Systems. & Touboul. J. the key on-going work aims in particular at  further exploiting and embedding control technologies in observation techniques (e. 169–209. M. F. Y. D.html. (2004).. it should also be evident that automation allows for an increased performance and efficiency of EO instruments and satellites. ESA. DFACS). (1997). Astrium. Utilisation of Satellite Formations for the Implementation of EO

Sign up to vote on this title
UsefulNot useful