)RUPDWLRQ )OLJKW &RQWURO...Each spacecraft involved in the assembly is both chaser and target at the...

Formation Flight Control

International Journal of Aerospace Engineering


Guest Editors: Yu Gu, Giampiero Campa,and Mario Innocenti

Copyright © 2011 Hindawi Publishing Corporation. All rights reserved.

This is a special issue published in volume 2011 of “International Journal of Aerospace Engineering.” All articles are open access articlesdistributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in anymedium, provided the original work is properly cited.

Editorial Board

Noor Afzal, IndiaBrij N. Agrawal, USASaad A. Ahmed, UAEC. B. Allen, UKN. Ananthkrishnan, KoreaHikmat Asadov, AzerbaijanNicolas Avdelidis, GreeceHyochoong Bang, KoreaWen Bao, SwitzerlandRonald M. Barrett, USADebes Bhattacharyya, New ZealandI. D. Boyd, USAKeh-Chin Chang, TaiwanChristopher J. Damaren, CanadaRoger L. Davis, USAGang Ding, ChinaDany Dionne, CanadaBoris Epstein, IsraelJinghong Fan, ChinaJiancheng Fang, ChinaIkkoh Funaki, JapanMohamed Gad-el-Hak, USAR. Ganguli, IndiaMuhammad R. Hajj, USA

Kathleen C. Howell, USAHui Hu, USAQinglei Hu, ChinaMikhail S. Ivanov, RussiaRatneshwar Jha, USAJacob I. Kleiman, CanadaHamid M. Lankarani, USAYunhua Li, ChinaT. T. Lim, SingaporeRick Lind, USAW. W. Liou, USARichard W. Longman, USAEnrico C. Lorenzini, ItalyT. J. Lu, UKSiao Chung Luo, SingaporeJoseph Majdalani, USAPier Marzocca, USAJosep Masdemont, SpainPhilippe Masson, USAFranco A. Mastroddi, ItalyJames J. McGuirk, UKGiovanni Mengali, ItalyAchille Messac, USAKoorosh Mirfakhraie, USAHong Nie, China

Christian Oliver Paschereit, GermanyChris L. Pettit, USAMark Price, UKN. Qin, UKMarkus Raffel, GermanySrinivasan Raghunathan, UKMahmut Reyhanoglu, USACorin Segal, USAKenneth M. Sobel, USAK. Sudhakar, IndiaMartin Tajmar, AustriaPaolo Tortora, ItalyJean-Yves Trépanier, CanadaSrinivas R. Vadali, USALinda L. Vahala, USAMehdi Vahdati, UKGeorge Vukovich, CanadaShaoping Wang, ChinaPaul Williams, The NetherlandsR. K. Yedavalli, USAYoungbin Yoon, KoreaGecheng Zha, USAYoumin Zhang, CanadaXiaolin Zhong, USAMei Zhuang, USA

Contents

Formation Flight Control, Yu Gu, Giampiero Campa, and Mario InnocentiVolume 2011, Article ID 798981, 2 pages

Guidance Navigation and Control for Autonomous Multiple Spacecraft Assembly: Analysis andExperimentation, Riccardo Bevilacqua, Marcello Romano, Fabio Curti, Andrew P. Caprari,and Veronica PellegriniVolume 2011, Article ID 308245, 18 pages

Cascade-Based Controlled Attitude Synchronization and Tracking of Spacecraft in Leader-FollowerFormation, Rune Schlanbusch, Antonio Lori’a, and Per Johan NicklassonVolume 2011, Article ID 151262, 12 pages

Design of an Extended Interacting Multiple Models Adaptive Estimator for Attitude Determination ofa Stereoimagery Satellite, Hossein Bolandi, Farhad Fani Saberi, and Amir Mehrjardi EslamiVolume 2011, Article ID 890502, 19 pages

Decentralized Model Predictive Control for Cooperative Multiple Vehicles Subject to CommunicationLoss, Hojjat A. Izadi, Brandon W. Gordon, and Youmin ZhangVolume 2011, Article ID 198308, 13 pages

Vision-Based Tracking of Uncooperative Targets, Suresh K. Kannan, Eric N. Johnson, Yoko Watanabe,and Ramachandra SattigeriVolume 2011, Article ID 243268, 17 pages

Hindawi Publishing CorporationInternational Journal of Aerospace EngineeringVolume 2011, Article ID 798981, 2 pagesdoi:10.1155/2011/798981

Editorial


Yu Gu,1 Giampiero Campa,2 and Mario Innocenti3

1 Department of Mechanical and Aerospace Engineering, West Virginia University, Morgantown, WV 26506, USA2 The Mathworks, El Segundo, CA 90245, USA3 Department of Electrical Systems and Automation, University of Pisa, 56126 Pisa, Italy

Correspondence should be addressed to Yu Gu, [email protected]

Received 27 June 2011; Accepted 27 June 2011

Copyright © 2011 Yu Gu et al. This is an open access article distributed under the Creative Commons Attribution License, whichpermits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The formation flight control problem has been extensivelydiscussed in recent years with numerous applications onaircraft and spacecraft systems. For aerial vehicles, advan-tages of performing formation flight have been well doc-umented, including fuel saving, improved efficiency in airtraffic control, and cooperative task allocation. For spacevehicles, precise control of formation flight will enablefuture large-aperture space telescopes, variable baseline spaceinterferometers, autonomous rendezvous and docking, androbotic assembly of space structures.

To facilitate further development in this emerging area ofresearch, this special issue includes several papers discussingthe most recent developments and ideas in the field, withemphasis on both the theoretical and experimental results.Of the five papers selected, three are in the area of spacecraftformation flight control and two are related to autonomousaerial vehicles. Contents of these papers include modelingand estimation, guidance, navigation, and control (GNC),and experimental techniques related to autonomous forma-tion flight.

The first paper of this special issue “Guidance navigationand control for autonomous multiple spacecraft assembly:analysis and experimentation” explores the GNC problemfor the in-plane orbital assembly of autonomous multiplespacecraft. The guidance and control strategies are designedto take into account the evolving shape and mass propertiesof the assembling spacecraft. The proposed approacheswere validated via hardware-in-the-loop experiments, usingfour autonomous 3-degree-of-freedom robotic spacecraftsimulators.

The second paper “Cascades-based controlled attitudesynchronization and tracking of spacecraft in leader-followerformation” presents control strategies for leader-follower

attitude synchronization of spacecraft formations in thepresence of disturbances. In a leader-follower formationconfiguration, the leader spacecraft is controlled to follow agiven reference, while a follower spacecraft is controlled tosynchronize its motion with the leader. A stability analysisis provided in the paper for both known and unknown butbounded disturbances.

The third paper “Design of an extended interactingmultiple models adaptive estimator (EIMMAE) for attitudedetermination of a stereo-imagery satellite” deals with theattitude determination issue for a pair of satellites usedin a stereoimaging scenario. For this type of operations,highly accurate and stable pointing maneuvers are neededto be accomplished in a few seconds, which requires thesatellites to rotate along a relatively large-angle attitudevery quickly. Therefore, different estimation strategies arediscussed for two different modes: “maneuvering motion”mode and “uniform motion” mode.

The fourth paper “Decentralized model predictive controlfor cooperative multiple vehicles subject to communication loss”investigates the control of multiple cooperative vehicles withthe possibility of communication loss/delay. Such commu-nication issues could lead to poor cooperation performanceand unsafe behaviors such as collisions. A decentralizedmodel predictive control (DMPC) architecture is proposedto estimate the tail of neighbor’s trajectory which is notavailable due to the large communication delays. The conceptof the tube MPC is also employed to improve the safety of thefleet against collisions, in the presence of large intervehiclecommunication delays.

The final paper of this special issue “Vision-based track-ing of uncooperative targets” presents both the flight testand simulation results of a follower aircraft tracking an

2 International Journal of Aerospace Engineering

uncooperative leader, using only monocular vision informa-tion. The situations with and without the subtended angleinformation for range estimation are both discussed withdifferent approaches presented.

Yu GuGiampiero Campa

Mario Innocenti


Research Article

Guidance Navigation and Control for Autonomous MultipleSpacecraft Assembly: Analysis and Experimentation

Riccardo Bevilacqua,1 Marcello Romano,2 Fabio Curti,3 Andrew P. Caprari,4

and Veronica Pellegrini5

1 Department of Mechanical, Aerospace & Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, USA2 Department of Mechanical and Aerospace Engineering and Space Systems Academic Group, Naval Postgraduate School,Monterey, CA 93943-5100, USA

3 Dipartimento di Ingegneria Aerospaziale e Astronautica, Scuola di Ingegneria Aerospaziale, Universitá di Roma “La Sapienza”,00138 Roma, Italy

4 Department of Mechanical and Aerospace Engineering, Naval Postgraduate School, Monterey, CA 93943-5100, USA5 Department of Applied Mathematics and Statistics, University of California, Santa Cruz, CA 95064, USA

Correspondence should be addressed to Riccardo Bevilacqua, [email protected]

Received 31 August 2010; Revised 22 October 2010; Accepted 23 December 2010

Academic Editor: Giampiero Campa

Copyright © 2011 Riccardo Bevilacqua et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

This work introduces theoretical developments and experimental verification for Guidance, Navigation, and Control ofautonomous multiple spacecraft assembly. We here address the in-plane orbital assembly case, where two translational and onerotational degrees of freedom are considered. Each spacecraft involved in the assembly is both chaser and target at the same time.The guidance and control strategies are LQR-based, designed to take into account the evolving shape and mass properties of theassembling spacecraft. Each spacecraft runs symmetric algorithms. The relative navigation is based on augmenting the target’s statevector by introducing, as extra state components, the target’s control inputs. By using the proposed navigation method, a chaserspacecraft can estimate the relative position, the attitude and the control inputs of a target spacecraft, flying in its proximity. Theproposed approaches are successfully validated via hardware-in-the-loop experimentation, using four autonomous three-degree-of-freedom robotic spacecraft simulators, floating on a flat floor.

1. Introduction

The technical difficulties presented by the autonomousmultiple spacecraft assembly problem relate to the devel-opment of robust and reliable guidance, navigation, andcontrol techniques for on-orbit evolving systems. The mainopen challenges are: (1) propellant-efficient control of anassembling (also known as evolving system), the evolutionoccurring both in its mass and inertia properties, as well asin its sensors and actuators configuration and (2) accuraterelative navigation among the spacecraft, especially in theevent of low frequency measurements update and interrup-tions of measurements due, for example, to relative sensors’view’s obstruction by other spacecraft. The works of [1–4]

address specifically the problem of a system’s evolution andits control. In [5], more emphasis is given to a potentialsolution for the wireless connectivity of different partsintended for the assembly of a bigger spacecraft, where aWi-Fi bridge acts as the only real “assembly.” Furthermore,wireless capability is becoming a more relevant option forexchanging data amongst close proximity spacecraft whicheventually dock to each other (see [6]). Also, the high-risksituation of an assembly maneuver in space does not leaveroom for computationally intensive logics, such as optimalcontrollers (see [7]). Onboard CPUs must allocate most oftheir performance capabilities to platform safety issues.

The use of Commercial Off-The-Shelve (COTS) relativesensors, such as low-cost cameras, justify the need for


robust relative navigation schemes. Many different filters fortracking a maneuvering target have been considered in theliterature.

Approaches based on Kalman filter include the work ofSinger [8], in which the target acceleration is modeled asa random process with known exponential autocorrelation.The input estimation approach for tracking a maneuveringtarget is proposed by Chan and Couture [9]. In thisapproach, the magnitude of the acceleration is identified bythe least-squares estimation when a maneuver is detected.The estimated acceleration is then used in conjunction witha standard Kalman filter to compensate the state estimateof the target. The standard filter alone is used duringperiods when no maneuver takes place. The augmentedfiltering approach is proposed by Bar-Shalom et al. [10,11]. In this approach, the state model for the target ischanged by introducing extra state components, the target’saccelerations. The maneuver, modeled as an acceleration, isestimated recursively along with other states associated withposition and velocity while a target maneuvers. Bogler [12]used this method as an implementation on high maneuvertarget tracking with maneuver detection.

The input estimation filter and the augmented-dime-nsion filter are commonly used in view of their computa-tional efficiency and tracking performance. Among inputestimation techniques, the Augmented State estimationapproach yields reasonable performance without constantacceleration or small sampling time assumptions. Further-more, it not only provides fast initial convergence rate, but itcan also track a maneuvering target with fairly good accuracyas mentioned by Khaloozadeh and Karsaz [13]. Bahari etal. [14] and Bahari and Pariz [15] propose an intelligenterror covariance matrix resetting, by a fuzzy logic approach,necessary for high maneuvering target tracking, to improvethe estimation of the target state.

In space applications, particularly in the spacecraftrelative navigation for the autonomous rendezvous andassembly, each vehicle is both the target and the chaserfor the other spacecraft. Here, an additional challenge isconsidered: the frequent loss of communications for thedata exchange when the application involves more than onespacecraft. Alternatives means to perform relative navigationmay include a vision-based system. These types of sensorsrequire the image processing and may result in low frequencymeasurement updates, especially for small spacecraft withlimited computation capabilities. Such sensors suffer ofproblems such as limitations on the field of view and/or otherspacecraft obstructing the view. Furthermore, each vehicledoes not usually know the other vehicles inputs, that is,it does not possess the information about the maneuversperformed by its fellow spacecraft. This missing informationneeds to be reconstructed in the estimation scheme thatwould otherwise diverge quickly.

We here focus on the utilization of low frequency updateand low-cost sensors, such as COTS devices. In particular,the spacecraft are envisioned to have range and line ofsight measurements, and relative attitude measurements.The navigation algorithm here presented build upon ourpreliminary work of [16].

Figure 1: Multispacecraft testbed at the Spacecraft RoboticsLaboratory of the Naval Postgraduate School.

In this work, we build upon known techniques in order todevelop guidance, navigation, and control approaches to per-form three-degree-of-freedom spacecraft assembly maneu-vers. Furthermore, the suggested methodologies are vali-dated via hardware-in-the-loop testing, using four roboticsspacecraft simulators.

In particular, the guidance and control problems aretackled by continuously linearizing the dynamics about thecurrent relative state vector between two spacecraft, andemploying a Linear Quadratic Regulator to suboptimizepropellant consumption. The LQR weighting matrices arecomputed in real time, depending on the relative state vector,acting as a feedback control. The LQR real-time solverdeveloped for this research is an extension of what usedduring a real on-orbit spacecraft test inside the InternationalSpace Station [17], where a simplified problem-targeted LQRwas executed (a version of the LQR Simulink solver for bothRTAI Linux and xPC Target is available for download) (see[18]). While the system evolves, changing its mass proper-ties and actuators’ configuration, the LQR-based approachremains unaltered, controlling the growing structure by thesimple online modification of a few parameters when a newspacecraft is docked.

As for the relative navigation, we here propose a designbased on the augmented state estimation technique. Robust-ness to frequent signal loss and/or darkening of the sensors isachieved. Furthermore, the suggested approach reconstructsthe information of the other vehicles’ maneuvers. A space-craft is envisioned to run a copy of the augmented state esti-mation technique algorithm for each other spacecraft in thebunch, every vehicle being chaser and target at the same time.

For the experimental part of this work, two dynamicmodels for the relative navigation filter are considered: (1)the classical Kalman filtering technique, [19], in which theunknown input (the maneuver command) is modeled asa random process and (2) the augmented state estimationtechnique, where the maneuver is estimated, using a Kalmanfilter scheme [19], in real time, as an additional variable in anaugmented state vector.

Between the two approaches, the second one proves tobe the most successful. It yields satisfactory performances

International Journal of Aerospace Engineering 3

without constant accelerations or small sampling timeassumptions. Furthermore, it does not only provide fastinitial convergence rate, but it can also track a maneuveringtarget with a good accuracy under unpredictable loss ofthe data link and slow data rate, allowing the spacecraftto perform critical maneuvers such as the docking and themultivehicle assembly.

The successful results of the study here presented pavethe way for further research and implementation of the newGNC techniques for the full six degrees of freedom spacecraftrelative motion.

Main contributions of this work to the state-of-the-artfor multiple spacecraft assembly GNC are as follows.

(1) Development of a guidance and control approachflexible to mass and actuators’ configuration changesduring the assembly. The methodology is based on asuboptimal LQR for propellant-efficient rendezvousand docking maneuvers.

(2) Implementation of a spacecraft relative navigationscheme based on augmented state estimation, robustto low frequency measurements updates. In par-ticular, the spacecraft are envisioned to have theavailability of range and line of sight measurements,and relative attitude measurements. No relativevelocities measurements are available. This is the firsttime, to our knowledge, that augmented state vectorestimation is used for spacecraft relative navigation.

(3) The first (to the best of authors’ knowledge) hard-ware-in-the-loop laboratory experiment involvingfour spacecraft simulators in a completely auto-nomous assembly maneuver.

The paper is organized as follows. Section 2 presents therobotic spacecraft simulators employed for the experiments.Section 3 presents the equations of the three-degree-of-freedom motion for the spacecraft relative maneuvering.Section 4 presents the augmented estimation approach anddemonstrates the observability of the augmented state.Section 5 illustrates the guidance and control. Section 6describes how navigation and control are performed oncemore spacecraft are assembled. Section 7 is dedicated tothe experimental validation of the proposed methodologies.Section 8 concludes the paper.

2. Third Generation Spacecraft Simulators atthe Spacecraft Robotics Laboratory

This section introduces the third generation of spacecraftsimulators developed at the Spacecraft Robotics Laboratoryof the US Naval Postgraduate School. Figure 1 shows thefleet of operational spacecraft simulators. The simulatorsfloat using air bearings over a very smooth epoxy floor,reproducing a nearly frictionless and weightless environmentin two dimensions and three degrees of freedom, that is,two degrees of freedom for the translation and one for therotation. This experimental testbed allows for the partialverification of guidance, navigation, and control algorithmsin a simulated in-plane close proximity flight condition

[20]. For more details on the different families of spacecraftsimulators employed throughout the world, we address thereader to [6, 16, 20–23].

In order to perform docking experimentations, twoseparate custom designed docking interfaces have beendeveloped and each is currently undergoing experimentaltesting (see Figure 2).

The type 1 docking interfaces are designed in orderto passively connect the spacecraft through electromagneticmechanisms, and their design will allow data/power/fluidsexchange (see Figure 3). Conversely, the type 2 designlacks the afore mentioned characteristics but enhances therobustness on the docking concept by correcting residualtranslational and rotational errors developed during the finaldocking phase of the spacecraft assembly for experimenta-tion. This second design hosts two small permanent magnetsto provide a final docking force and to keep the robotsphysically connected.

Other key features of the spacecraft simulators includethe following.

(1) Ad-hoc wireless communication. Continuous dataexchange amongst each simulator and the externalenvironment over the wireless network providesfor in situ communication. This greatly increasesthe robustness of data collection in the event ofcommunication loss with one of the simulators.

(2) Modularity. The simulators are divided into twomodules where the payload can be disconnected fromthe consumables, thus allowing for a wide rangeof applications with virtually any kind of differentpayload (Figure 4).

(3) Small footprint. The .19 m length × .19 m width ofeach simulator allows for the working area (∼ 5 m ×5 m) on the epoxy floor to be optimally exploited.

(4) Light weight. ∼10 kg.(5) Rapid Prototyping. The capability to rapidly repro-

duce further generations of simulators and improveexisting designs via computer aided design (CAD)with the in house STRATASYS 3D printing machine.

Most notably, point 1 of the previous list has providedan invaluable contribution to the success of our ongo-ing experimentation. The ad-hoc wireless communicationsystem, currently employed onboard the simulators, wasexperimentally verified by a distributed computing test,which demonstrated the wireless communication real-timecapability for the SRL (see [6]).

Table 1 illustrates the characteristics of the electron-ics used onboard each spacecraft simulator. The PC104(onboard computer), the sensors, and the actuators aredescribed below (see [6]).

Each robot performs absolute navigation in the labora-tory environment employing indoor pseudo-GPS for posi-tion, and magnetometer and gyroscope for attitude (Table 1).The measurements from these sensors are processed by twoseparate Linear Digital Kalman Filters, estimating positionand velocity of the center of mass with respect to the


(a) Type 1 (b) Type 2

Figure 2: (a) Patent pending docking interface design (electromagnet and fluid transfer capability). (b) Concept (male/female) dockinginterface used for the experiments in this paper.

Table 1: Electronics hardware description.

Part’s name and manufacturer Details Description

PC104 (plus) Motherboard(Advanced Digital Logic)

ProcessorRAM

SmartCoreT3-400, 400 Mhz CPUSDRAM256-PS

Compact Flash(SanDisk Extreme IV)

— 8 Gbyte capacity

20 Relays Board (IR-104-PBF)(Diamond Systems)

— High-density optoisolated input + relay output

8 Serial Ports Board (MSMX104+)(Advanced Digital Logic)

— —

Firewire PC104 board (Embedded Designs Plus) — IEEE1394 card with 16 Bit PC104

Compact Wireless-G USB Adapter (Linksys) —54 Mbps 802.11 b/g wireless USB network interfaceadapter

Wireless Pocket Router/AP DWL-G730AP(D-Link)Solenoid Valves (Predyne)

—2.4 Ghz 802.11 g, ethernet to wireless converter2 way, 24VDC, 2 Watt

Fiber Optic Gyro DSP3000 (KVH) — Single axis rate, 100 Hz, Asynchronous, RS-232

Magnetometer, MicroMag-3Axis (evaluation kitwith RS232 board) (PNI)

—Asynchronous, RS-232 (the evaluation kit is still adevelopment version)

DC/DC converters: EK-05 Battery Controller andRegulator + DC1U-1VR 24V DC/DC Converter(Ocean Server)

—3.3, 5, 12, 24 Volts outputs. The main board isequipped with a batteries’ status controller.

Battery (Inspired Energy) — Lithium Ion Rechargeable battery (95 Whr)

Metris iGPS pseudo-GPS indoor system — —

laboratory reference frame, and heading and heading rate ofthe robot with respect to the laboratory frame. The detailson the robots’ absolute navigation are beyond the scope ofthis work, and they will not be discussed here; for additionalinformation, the reader can refer to [20].

The maximum computational power of 400 Mhz listedin Table 1 is not required for real-time recomputation ofthe LQR solution. In the SPHERES satellites [17], TexasInstruments C6701 Digital Signal Processor is employed tosolve a very comparable problem.

Figure 5 depicts the main concept of the testbed atthe SRL. The main components and their interfaces areillustrated onboard the robot at the bottom of the sketch.Furthermore, the figure emphasizes the fact that the configu-ration is scalable to an arbitrary number of robots dependingon the application or mission.

The Wi-Fi capability of each robot is not only used tocommunicate with other robots, but it is also necessary forreceiving its own absolute position within the laboratory, assensed by the pseudo-GPS indoor system.


Connectors fordata and power

exchange

Hollow maleconnector formechanical

matching andfluid transfer

Hollow femaleconnector formechanical

matching andfluid transfer

Magnet forgenerating

docking pullingforce

Figure 3: Main components of the patent pending docking interface.

Magnetometer

Polycarbonaterapid prototypedupper module:

8 Fixed thrusters:solenoid valves andsupersonic nozzles

2 paintball 4500 PSIcompressed air tanks

(supplies thrusters andair bearings)

Polycarbonaterapid prototyped

lower module:

3 linear airbearings

Indoor pseudo-GPSsystem: receiver

Wireless serverfor exchanging

pseudo-GPS data

Fiber opticGyro

PC-104 stack:motherboard, relayboard, RS232 ports

board

2 lithium ionrechargeable

batteries

Wireless pocketrouter for ad-hoccommunication

Figure 4: Detailed collocation of the hardware on the spacecraft simulators.

The onboard real-time operating system is RTAI patchedLinux (see [24]), in the Debian 2.6.19 flavor. The classical useof xPC Target by MathWorks as a real-time operating system(OS) is common in academic research (see [25]). A keyadvantage of xPC Target is its seamless integration between

Simulink via Real-Time Workshop which allows for rapidprototyping of navigation and control algorithms for real-time requirements. Real-Time Workshop automatically gen-erates C code from a Simulink model and the correspondingexecutable file for an xPC Target-based computer. On the


Spacecraft 1, IP: 170.160.1.1

PC104

PC104PC104

Solenoidvalve

Solenoidvalve

Relay board

Relay boardRelay board

Supersonic nozzle thruster

Supersonic nozzle thruster


Wi-Fi UDP


Ad-Hoc Wi-Finetwork (UDP)

Wi-Fi UDP

Wi-Fi UDP Wi-Fi UDP

Fiber optic Gyro

Fiber optic GyroFiber optic Gyro

Magnetometer

MagnetometerMagnetometer

RS232

RS232RS232

IGPSwirelessserver

Wi-Fiadapter

Wi-Fiadapter

Wi-Fiadapter

IGPS IGPS

Spacecraft n, IP: 170.160.1.N

Figure 5: Ad-hoc wireless network at the SRL test-bed.

other hand, xPC Target has some disadvantages that includesupport for a limited number of hardware componentsand no support for USB or Firewire devices. Furthermore,the inaccessibility of its source code, due to its proprietarycommercial nature, makes it challenging to add or modifydrivers for unsupported hardware.

RTAI Linux has been successfully used as an onboardreal-time OS. RTAI is a patch to the Linux kernel that allowsfor the execution of real-time tasks in Linux (see [26, 27]).The RTAI Linux solution is being widely exploited in severalengineering areas (see [28–31]). In this work, we use RTAILinux with a wide variety of hardware interfaces to includewireless ad-hoc radio communication using UDP, RS232interface with the sensor suite and power system and aPC/104 relay board for actuating compressed air nozzles.RTAI Linux also allows for automatic generation of C codefrom Simulink models through Real-Time Workshop withthe executable file for the onboard computers being createdoutside MATLAB by simple compilation of the C code.

The details on the ad-hoc wireless network andhardware-software interfaces developed for the SpacecraftSimulators are available in [6].

3. S/C Relative Motion Dynamics andProblem Statement

In this section, we provide the dynamics of spacecraft relativemotion in the three degrees of freedom case. The dynamics

xinertial

yinertial

zinertial

LVLH frame

rL

rS

ρ

Orbital path

Spacecraft bodyprincipal axes

frame

xy

z

Figure 6: Local vertical local horizontal and inertial frames.

encompasses both the relative translation (two degrees offreedom) and rotation (one degree of freedom). We willrefer in the following to a Local Vertical Local Horizontal(LVLH) reference frame (Figure 6) that rotates with theorbital angular velocity ωLVLH. The origin of LVLH moves ona virtual orbit, conveniently chosen to remain in the vicinityof the maneuvering spacecraft. This point can also be chosenas coincident with one of the spacecraft. The x-axis pointsfrom the center of the Earth to the center of LVLH, while they-axis is in the orbital plane in the direction of the motionalong the orbit and perpendicular to the x-axis. The z-axiscompletes the right-handed LVLH frame.

The dynamics of such motion can be represented in thecompact form as

Ẋrel = f (Xrel) + B(Xrel)u. (1)From now on, we will consider the specific application

of hardware-in-the-loop testing using the three-degree-of-freedom spacecraft simulators at the Naval PostgraduateSchool. For the experimental setup, the state vector becomes

Xrel

=[x y θ ẋ ẏ ω

]T=[xT−xC yT−yC θT−θC ẋT−ẋC ẏT− ẏC ωzT−ωzC

]Tu =

[uxT − uxC uyT − uyC MT −MC

]T.

(2)

Being uij , i = x, y, j = C,T the control forcescomponents of chaser and target, and Mj , j = C,T thecontrol torque of chaser and target about the z axis.

It is common use in the literature to linearize the relativemotion dynamics and use the Clohessy and Wiltshire linearequation [32]

ẍ − 2ωLVLH ẏ − 3ω2LVLHx =1m

(uxT − uxC

),

ÿ − 2ωLVLHẋ = 1m

(uyT − uyC

),

(3)


with the assumption that the spacecraft have the same massm.

For maneuvers confined in the vicinity of the LVLHorigin, elapsing a short time in comparison to the orbitalperiod (3) can be further simplified into a double integratorfor both x and y. A double integrator dynamics alsorepresents the dynamics of the spacecraft simulators in thelaboratory inertial reference frame. For the above-mentionedreasons, (4) will be used for the remaining of the paper;

ẍ = 1m

(uxT − uxC

),

ÿ = 1m

(uyT − uyC

).

(4)

Assuming the spacecraft having the same moment ofinertia about the z axis, the attitude dynamics is alsorepresented by a double integrator

θ̈ = θ̈T − θ̈C = ω̇zT − ω̇zC = 1Jz (MT −MC). (5)

The goal of this work is to develop a GNC approach fordriving the state Xrel to perform assembly maneuvers. Thisrequires accurate guidance, especially in the last phases ofdocking, optimized or suboptimized control, to minimizepropellant consumption, and a robust relative navigationscheme. These requirements are addressed in the followingsections.

4. Relative Navigation: The AugmentedState Estimation Method

In this section, the theory for the three-degree-of-freedomaugmented state relative navigation is presented. The con-trols of another vehicle (target) are treated as additionalterms in the corresponding state equation, so that themodel provides an augmented state vector. The measure-ments available on each spacecraft are relative positions(from range and line of sight) and relative attitude, andwe assume the knowledge of the controls of the chaser,onboard the chaser itself. No relative velocities measure-ments are available. Observability proof of the vector[x y θ ẋ ẏ ω uxT uyT MzT ]

T from the measurements

[x y θ]T is provided for the proposed estimation technique,demonstrating how the augmented state technique canreconstruct relative velocities and controls of the target.

In the following developments, the estimated target’scontrols are considered constants within every sample timeinterval. It is worth underlying that the control variables uxT ,uyT , and MzT do not represent the actual actuators’ controlvariables onboard the spacecraft simulator. The way uxT ,uyT , and MzT are obtained from the target does not matterfrom the augmented state filter’s point of view. These controlvariables are estimated in order to add robustness to thefiltering technique; they represent the target’s maneuvers,but not the specific way they are performed by the target’sactuation subsystem.

The same assumption will be used for the observabilitydemonstration. The navigation algorithm is developed usingthe Kalman filter approach.

The augmented state estimation approach presentsnumerical efficiency comparable to the standard KalmanFilter applied on the state only. In fact, the augmented stateapproach introduces a few more variables in the Kalman Fil-ter, without a significant increase on the numerical burden.Additional references with regards to the implementation ofExtended Kalman Filters onboard real space missions can befound in [33, 34].

4.1. Relative Motion Estimation. The assumption is made ofindependent estimation and control for the attitude and theposition, so that we can proceed as follows. For the relativeposition, the state vector can be written as (see (2))

CjTi X =

CjTi

[x y ẋ ẏ

]T. (6)

The discrete dynamics for the problem is the following:

CjTi X(k + 1) =

CjTi Ψ(k)

CjTi X(k)

+CjTi B(k)(uT(k)− uC(k)) + GW(k),

Z(k) = HCjTi X(k) + V(k).

(7)

The expressions of the matrices: G, H ,CjTi B(k), and Ψ as

functions of the measurement update time Ts for this planarcase can be written as

G = I4×4, H =(

I2×2 02×2)

,

CjTi B(k) =

∫ Ts0

(CjTi Φ(t) ·

CjTi B

)dt =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

T2s2

0

0T2s2

Ts 0

0 Ts

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠/m,

CjTi Ψ(k) =

CjTi Φ(Ts) = e

CjTi

FTS

= I4×4 + CjTi FTS =

⎛⎜⎜⎜⎜⎜⎜⎝1 0 TS 0

0 1 0 TS

0 0 1 0

0 0 0 1

⎞⎟⎟⎟⎟⎟⎟⎠,(8)

being

CjTi F =

⎛⎜⎜⎜⎜⎜⎜⎝0 0 1 0

0 0 0 1

0 0 0 0

0 0 0 0

⎞⎟⎟⎟⎟⎟⎟⎠,CjTi B =

⎛⎝02×2I2×2

⎞⎠/m. (9)


The augmented dynamics adds the estimation of uxT anduyT , assuming knowledge of the chaser’s controls uxC and uyC .The related state equation matrices can be written as

CjTi XA =

⎡⎣ CjTi X(K + 1)uT(K + 1)

⎤⎦

= CjTi ΨA(k)⎡⎣ CjTi X(K)

uT(K)

⎤⎦ + GAW(k) + CjTi BC(k)uC(k),

ZA(K) = HA⎡⎣ CjTi X(K)

uT(K)

⎤⎦ + V(K),

GA =⎛⎝ G

02×4

⎞⎠, HA = (I2×2 02×4), uC =⎡⎣uCxuCy

⎤⎦,

CjTi ΨA(k) =

⎛⎝ CjTi Ψ(k) CjTi B(k)02×4 I2×2

⎞⎠ =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

1 0 Ts 0T2s2m

0

0 1 0 Ts 0T2s2m

0 0 1 0Tsm

0

0 0 0 0 0Tsm

0 0 0 0 1 0

0 0 0 0 0 1

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠

,

CjTi BC(k) =

⎛⎜⎝−CjTi B(k)

02×2

⎞⎟⎠ =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

−T2s

20

0 −T2s

2

−Ts 0

0 −Ts0 0

0 0

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠

/m.

(10)

4.2. Relative Attitude Estimation. The same algorithm isimplemented for target’s attitude and control torque estima-tion. Assuming the target rotating only about the vertical axis(z-axis), the ith Target attitude state vector, with respect tothe jth Chaser spacecraft, is chosen to be

CjTi Θ =

Cj

Ti

[θ θ̇

]T. (11)

The discrete dynamics for the attitude problem is

CjTi Θ(k + 1) =

CjTi Ψ(k)

CjTi Θ(k)

+CjTi B(k)(MT(k)−MC(k)) + GW(k),

Z(k) = H CjTi Θ(k) + V(k),

(12)

and the principal dynamics matrices, as function of the timesampling Ts, are

G = I2×2, H =(

1 0)

,

CjTi B(k) =

∫ Ts0

(CjTi Φ(t)·

CjTi B

)dt =

⎛⎜⎜⎜⎝T2s2JzTsJz

⎞⎟⎟⎟⎠,

CjTi Ψ(k) =

CjTi Φ(Ts) = e

CjTi

FTs = I2×2 + CjTi FTs =⎛⎝1 Ts

0 1

⎞⎠,(13)

being

CjTi F =

⎛⎝0 10 0

⎞⎠, CjTi B =⎛⎝0

1

⎞⎠/Jz. (14)The formulation of the augmentation of the state dynam-

ics adds the estimation of MT , assuming the knowledge ofthe chaser’s control torque MC . The related state equationmatrices can be written as

CjTi ΘA(k) =

⎡⎣CjTi Θ(k + 1)MT(k + 1)

⎤⎦

= CjTi ΨA(k)⎡⎣CjTi Θ(k)MT(k)

⎤⎦ + CjTi BC(k)Mc +GAW(k),ZA(k) = HA

⎡⎣CjTi Θ(k)MT(k)

⎤⎦ + V(k),GA =

⎛⎝ G02×2

⎞⎠, HA = (1 0 0), Mc = [Mz],

CjTi BC(k) =

⎛⎝−CjTi B(k)0

⎞⎠ =⎛⎜⎜⎜⎜⎜⎜⎝−T

2s

2

−Ts0

⎞⎟⎟⎟⎟⎟⎟⎠/Jz,

CjTi ΨA(k) =

⎛⎝CjTi Ψ(k) CjTi B(k)01×2 1

⎞⎠ =⎛⎜⎜⎜⎜⎜⎜⎜⎝

1 TsT2s2Jz

0 1T2sJz

0 0 1

⎞⎟⎟⎟⎟⎟⎟⎟⎠.(15)


x y

z

Spacecraft 1

β

yB

zB

xB

Spacecraft 2

Docking coneα

LVLH frame

Spacecraft body frame

rrsw

rrsw⊥

rgoal

r

rrot

Figure 7: Relative vectors used in the alignment and assembly logic.All vectors are in the LVLH xy plane.

4.3. Observability of the Augmented Dynamics. For sake ofsimplicity, considering that the controls are constant in eachsample time, we provide, for the planar case, the proof ofthe observability for the continuous models of the relativedynamics. The observability property holds for the discretemodels [35]. The augmented relative motion dynamics canbe expressed as⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

ẋ

ẏ

ẍ

ÿ

u̇xT

u̇yT

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

=

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

0 0 1 0 0 0

0 0 0 1 0 0

0 0 0 01m

0

0 0 0 0 01m

0 0 0 0 0 0

0 0 0 0 0 0

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

x

y

ẋ

ẏ

uxT

uyT

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦+

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

0 0

0 0

− 1m

0

0 − 1m

0 0

0 0

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

⎡⎣uxCuxC

⎤⎦.

(A) (B)

(16)

The measurements are related to the state as follows:

Z =

⎡⎣1 0 0 0 0 00 1 0 0 0 0

⎤⎦(C)

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

x

y

ẋ

ẏ

uxT

uyT

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦. (17)

It is of immediate demonstration that the followingobservability matrix has full rank:

O =

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

C

CA

.

.

CA5

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦=

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

1 0 0 0 0 0

0 1 0 0 0 0

0 0 1 0 0 0

0 0 0 1 0 0

0 0 0 01m

0

0 0 0 0 01m

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

. (18)

Similar developments lead to observability for the relativeattitude motion. The dynamics can be expressed as

⎡⎢⎢⎢⎣θ̇

θ̈

ṀzT

⎤⎥⎥⎥⎦ =⎡⎢⎢⎢⎢⎣

0 1 0

0 01Jz

0 0 0

⎤⎥⎥⎥⎥⎦⎡⎢⎢⎢⎣

θ

θ̇

MzT

⎤⎥⎥⎥⎦ +⎡⎢⎢⎢⎢⎣

0

− 1Jz

0

⎤⎥⎥⎥⎥⎦MzC.

(A) (B)

(19)

The measurements are related to the state as follows:

Z =[

1 0 0]

(C)

⎡⎢⎢⎢⎣θ

θ̇

MzT

⎤⎥⎥⎥⎦. (20)The observability matrix is as following:

O =

⎡⎢⎢⎢⎣C

CA

CA2

⎤⎥⎥⎥⎦ =⎡⎢⎢⎢⎢⎣

1 0 0

0 1 0

0 01Jz

⎤⎥⎥⎥⎥⎦, (21)

that has full rank.

5. Guidance and Control forthe Assembly Maneuver

This section describes guidance and control for the auto-nomous assembly.

5.1. Guidance. Figure 7 shows the principal vectors used bythe guidance algorithm. It is worth to underline that Figure 7


yB

zB

xB

rport

Figure 8: Body fixed docking port vector.

only represents one possible configuration and that thedocking ports do not have to be aligned with any particularbody axis. rgoal is the vector originating from the center ofa docking interface, terminating at the center of mass ofthe other spacecraft going to dock to it. rrsw is the vectororiginating from a spacecraft’s center of mass, terminatingat the other spacecraft’s center of mass.

The guidance problem is here expressed in terms ofdesired state vector for each spacecraft, defined dynamicallyduring the maneuver. The state vector error to minimize is

xerr

=[x − xdes y − ydes θ − θdes ẋ− ẋdes ẏ − ẏdes ω − ωdes

]T.

(22)

The subscript “des” indicates a desired relative statevector component. The desired state is dynamically changedthroughout the assembly maneuver according to the follow-ing two-phase guidance logic.

The center of mass trajectory is unconstrained, free tobe optimized, unless in the vicinity of the docking phase.As for the attitude, we reproduce a realistic conditionin which the spacecraft has to show one particular side(usually the one with the docking port) towards the currenttarget spacecraft. In other words, the docking port sideis commanded to be perpendicular to either the rrsw orthe rgoal vector (Figure 7), depending on the phase. Eachspacecraft in Figure 7 can be considered either a single agentor an already assembled structure; the following, descriptionapplies to both scenarios. In the following the vectors arealways intended to be parallel to the xy plane. rdock is auser defined distance threshold, specifying when the dockingphase begins.

(1) |rrsw| > rdock, RENDEZVOUS: the spacecraft is at a faraway distance from its target docking port. The state vectorerror is xerr = [rgoalx rgoaly θ − θdes ẋ ẏ ω]T . The desiredattitude θdes is such to align rport to rrsw (Figures 7 and 8).

(2) |rrsw| ≤ rdock, DOCKING APPROACH: the spacecraftis close to its target docking port. The desired state vector tominimize is:

(a) If cos−1((rgoal · rport)/|rgoal||rport|) < α, that is, thespacecraft is within the security docking cone, thereare two subcases.

SUBCASE 1. The distance between the spacecr-aft is greater than the chosen impingement stand-offrange, then xerr = [rgoalx rgoaly θ − θdes ẋ ẏ ω]T .The desired attitude ϑdes is such to align rport to rrsw(Figures 7 and 8).

SUBCASE 2. The distance between the spacecraft isless than the chosen impingement stand-off range,then any thrusters causing plume impingement onthe other spacecraft are shut off, and only used if anemergency brake is needed, in the event of dockingoccurring at high velocity (above a chosen threshold).For the NPS spacecraft simulators, this will meanshutting off two thrusters, as it will be clear lateron. The remaining actuators will compensate forattitude alignment in the last phase of docking andwill provide required forces to push the spacecrafttogether.

(b) If cos−1((rgoal · rport)/|rgoal||rport|) ≥ α, that is, thespacecraft is outside the security docking cone. Inthis case, referring to spacecraft 2 of Figure 7, thevehicle maneuvers orbiting around the one hostingits target docking port, moving along the directionperpendicular to the rrsw vector, towards the waythat is the shortest in order to reach the safetycorridor. The amount of commanded rotation ateach time step, around the target docking port,is a chosen parameter β = const. In terms ofstate vector error to minimize, defining a refer-ence frame which has as a basis the unit vectorsr̂rsw⊥ , r̂rsw, the rrsw can be rotated of an angle βinto rrot and easily expressed as function of the basisrrot = |rrsw|(cosβr̂rsw + sinβr̂rsw⊥). The state errorto minimize is xerr = [rrotx− rgoalx rroty− rgoalxθ−θdes ẋ ẏ ω]T . The desired attitude θdes is suchto show the chosen side to the target docking space-craft, that is, ⊥rgoal (Figure 7). In simple terms, thesatellites circle around each other, in the direction ofshortest angular displacement, to allow the dockinginterfaces to be in the mutual fields of view. Eachspacecraft needs to be in the safety corridor one ofthe other, with the respective docking interfaces’ rportvectors and rgoal vectors aligned. The respective rportvectors of two satellites will need to be at 180 degrees(plus-minus the tolerance); the same applies, as aconsequence, to the rgoal vectors.

5.2. LQR Control. The LQR problem (23) is solved at eachtime step, with dynamically sized weighting matrices Q andR, adapting to the current situation, avoiding high controlvalues when the state vector error is relevant, and vice versa.This choice results in a smoother behavior, in terms ofrequested control actions, with respect to classical fixed gainmatrices LQR;

J =∫∞

0

(xTerrQxerr + u

TRu)dt. (23)


YLVLH

XLVLH

xbody

ybody

ϑC

+ +

+

+

u4 u3

u2

u1

Figure 9: Locations of controls for the planar assembly.

Table 2: Main simulation parameters.

Mass of each simulator 10,5 kg

Inertia of simulator 0.063 kg m2

Inertia of the two simulators assembled 0.18 kg m2

Single thruster estimated force [36] 0.16 N

Docking cone semiaperture 0.75 degrees

Force arms (for torque generation) 5, 10, 21 cm

Limit distance for switching off the thrusters 0.7 m

iGPS accuracy 1 mm

Gyroscope accuracy 0.003 deg/sec

Thrusters minimum actuation time 1.5 · 10−3 sec

The cost function in (23) aims to minimize control effort,while reducing the relative state vector between two satellites.The mutual relevance between state vector error and controleffort is dictated by the relative values of the weightingmatrices Q and R.

The control vector u is chosen as a four-componentvector of forces, expressed in the spacecraft body frame(Figure 9). The choice of u in the spacecraft body frame,removes the need for thruster mapping [22].

For the phases described in the previous section theweighting matrices for the LQR are chosen as

Q =

⎡⎢⎢⎢⎣1∣∣∣rgoal∣∣∣ · I3 03

03∣∣∣rgoal∣∣∣3 ·V · I3

⎤⎥⎥⎥⎦,

R =∣∣∣rgoal∣∣∣a2

I4.

(24)

LQRY

S

K

E

LQRY solution

A

B

C

D

Q

R

Figure 10: LQR Solver Simulink Block [18]. This routine solves thecomplete algebraic Riccati equation accepting the input matrices:A (dynamics matrix), B (control matrix), C (state-output mappingmatrix), D (control-output mapping matrix), and Q and Rweighting matrices. The outputs are the LQR gain matrix, K, whichis the solution to the associated algebraic Riccati equation, thematrix S, and a two-dimensional vector, E, whose first elementindicates an error when it is greater than zero or a somewhatunreliable result when it is negative. The second element of E is thecondition number of the R matrix.

Each time step solution of the LQR generates a gainmatrix KLQR, used to implement the required suboptimalcontrol vector

uLQR = −KLQR · xerr. (25)

The values of the constants a = 3.05 · 10−2 and V =0.06 are chosen as in [23]. In particular, their values arechosen looking at variables with physical meaning, butwe do not assign dimensions to them here, being theirdimensions the appropriate ones for consistency in thecost function (23). The a constant weighs the terms inthe matrix R with respect to the maximum translationalacceleration achievable on the spacecraft simulator. Thisvalue is computed considering two thrusters simultaneouslyactivated on the same side of the vehicle. The thrust valuesand mass of the simulator can be found in Table 2, andthe interested reader can find more details on the thrustersin [36]. The originating idea for scaling the R matrix asin (24) is the desire to maintain the controls required bythe LQR solution below the maximum hardware-achievablecontrols. With regard to the constant V in the Q matrix,it is set to be the maximum translation speed allowed forthe simulator. V weighs the part of the state vector whichis related to linear velocities. The choice to introduce theabove mentioned parameters does not a priori guaranteecontrols below the maximum onboard control authority anda limited translational velocity, but the scaling in (24) hasbeen proven very effective in mitigating high requests on thecontrol and undesired fast maneuvers on the testbed. Thisresult was previously found in numerical simulations, andthen experimentally verified [23].


ybody

+ +

YLVLH

XLVLH

xbody

+

+

ϑC

r

1

2

3

45

6

7

8

u4 u3

u2

u1

Figure 11: Thruster coupling on the spacecraft simulators.

Figure 10 shows the required inputs to the LQR solver,implemented in Simulink.

The LQR solver employed for developing the proposedapproach was downloaded from [18], adapted for automaticgeneration of code through Real-Time-Workshop for RTAILinux (it was originally only compatible with WindowsOperating Systems), and uploaded again on the MathWorksfile exchange website [18].

In specializing the design of Figure 9 to the SRL space-craft simulators, we treat the eight body fixed thrustersin couples, so that symmetric thrusters are reduced toone control variable, which can be either umax, −umax, 0.Figure 11 shows the thruster couplings: 1–4, 2–7, 3–6, and5–8. The control vector is u = [u1 u2 u3 u4]T . The redarrows along the couples in Figure 11 show the positivedirections assumed for the controls. Ultimately, thrusters

coupling allows the LQR to solve a reduced problem in whichthe control vector has four components instead of eight.

Given the choice for the control vector, the controldistribution matrix becomes nonlinear, as in (25). Equation(25) also shows the system dynamics matrices, when theexpression Ẋ = AX + BU, Y = CX + DU is used. Thespacecraft orientation θC is replaced with θ for simplicity

A =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

0 0 0 1 0 0

0 0 0 0 1 0

0 0 0 0 0 1

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠,

B =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

0 0 0 0

0 0 0 0

0 0 0 0

cos(θ)m

cos(θ)m

− sin(θ)m

− sin(θ)m

sin(θ)m

sin(θ)m

cos(θ)m

cos(θ)m

−rJz

r

Jz

r

Jz

−rJz

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠

,

C = I6×6, D = 06×4.

(26)

In order to employ the LQR approach, the controldistribution matrix is linearized at each time step, in thevicinity of the desired attitude.

BLIN =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

0 0 0 0

0 0 0 0

0 0 0 0

cos(θdes)− sin(θdes)(θ − θdes)m

cos(ϑdes)− sin(θdes)(θ − θdes)m

− sin(ϑdes) + (cos(θdes))(θ − θdes)m

− sin(θdes) + (cos(θdes))(θ − θdes)m

sin(θdes) + (cos(θdes))(θ − θdes)m

sin(θdes) + (cos(θdes))(θ − θdes)m

cos(ϑdes)− sin(θdes)(θ − θdes)m

cos(θdes)− sin(θdes)(θ − θdes)m

−rJz

r

Jz

r

Jz

−rJz

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠(27)

By inserting the matrices defined in (26) and (27) andthe weighting matrices described in (24) into the LQR solverof Figure 10, the optimal four components vector of forcesis obtained, at each time step during the maneuver. Theobtained control vector will be a continuous signal. In orderto drive the on/off thrusters from the continuous signal PulseWidth Modulation is used; the PWM collects commandedcontrols over 10 sample times before actuating. Furthermore,a Schmitt Trigger is implemented, to filter out low com-manded controls and reduce the amount of chattering.

6. Navigation and Control ofthe Assembled Structure

Once the S/Cs are assembled, the mass and inertia propertiesalong with the thrusters configuration change. Figure 12shows an example, applied to the SRL testbed, in whichthrusters six and seven of both spacecraft cannot be usedanymore. The assembled new spacecraft has doubled mass,different moment of inertia and four more thrusters, dif-ferently allocated with respect to the single spacecraft. Inassembled configuration, one of the robots acts as master


ybody

xbody

+ +

+ +

+ +

+

+

+

+

+

+

Docking connection(rigid assumption)

Controlreallocation

r0

ri

r

OffOff

OffOff

18

2

3

45

6

7

7

6

54

3

2

1

1 12

2

3

4 5 6 78

91011

8

Combined center ofmass (C.O.M.) shift

u4 u3

u2

u1

u4 u3

u2

u1

u4u3

u2

u1

S/C 1 S/C 2

S/C 1 S/C 2

S/C centerof mass

Clock-wise thrusterrelabeling

Figure 12: Assembled configurations with reallocated thruster coupling and COM shift.

0 0.5 1 1.5 2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

X (m)

Y(m

)

Docking is missed

Accumulated error is above 8 cm

S/C simulator 1 initial condition

S/C simulator 2 initial condition

Figure 13: Experimental Result: bird’s eye view for two spacecraft simulator failed assembly maneuver. The relative navigation isperformed via classical Kalman Filtering, no Augmented State Estimation. The bolded lines are employed to help visualize the simulator’sorientation.


S/C simulator 1initial condition




S/C 1 and 2 assemblyat t = 100.6 s

S/C 3 and 4assembly att = 82 s

−1 −0.5 0 0.5 1 1.5 2X (m)

0

0.5

1

1.5

Y(m

)

Four S/C finalassembly at t = 163.8 s

Figure 14: Experimental Result: bird’s eye view for four spacecraftsimulator assembly maneuver. The relative navigation is performedvia augmented state estimation. The bolded lines are employed tohelp visualize the simulator’s orientation.

and performs both navigation and control of the new biggerrobot. In order to keep using the same logic employed forcontrolling a single simulator, the twelve thrusters of thenew assembled spacecraft are associated according to thefollowing sets:

(1) u1 is generated by firing either thruster 8 (u1 < 0) or3 (u1 > 0),

(2) u2 is generated by firing either thruster 9 (u2 < 0) or2 (u2 > 0),

(3) u3 is generated by firing either thrusters 6 and 7synchronously (u3 < 0) or 11 and 10 synchronously(u3 > 0),

(4) u4 is generated by firing either thrusters 4 and 5synchronously (u4 < 0) or 1 and 12 synchronously(u 4 > 0).

The input matrices to the LQR solver will be changedonce an additional portion of the structure is connected.

Also, the new control vector will have maximum andminimum values reduced, due to the increase of mass. Forinstance, the case represented in Figure 12 leads to the newmatrices

A =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

0 0 0 1 0 0

0 0 0 0 1 0

0 0 0 0 0 1

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠,

B =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

0 0 0 0

0 0 0 0

0 0 0 0

cos(θ)mcomb

cos(θ)mcomb

− sin(θ)mcomb

− sin(θ)mcomb

sin(θ)mcomb

sin(θ)mcomb

cos(θ)mcomb

cos(θ)mcomb

−rJz comb

r

Jz comb

ro + riJz comb

−(ro + ri)Jz comb

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠

,

C = I6×6, D = 06×4,

(28)

where Jz comb is the inertia of the assembled system about thevertical axis and mcomb is the new mass. Linearization of thenew control distribution matrix leads to

BLIN =

⎛⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎝

0 0 0 0

0 0 0 0

0 0 0 0

cos(θdes)− sin(θdes)(θ − θdes)mcomb


− sin(θdes) + cos(θdes)(θ − θdes)mcomb

− sin(θdes) + cos(θdes)(θ − θdes)mcomb

sin(θdes) + cos(θdes)(θ − θdes)mcomb

sin(θdes) + cos(θdes)(θ − θdes)mcomb

cos(θdes)− sin(ϑdes)(θ − θdes)mcomb


−rJz comb

rJz comb

ro + riJz comb

−(ro + ri)Jz comb

⎞⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎠(29)

The thrusters remained available after docking will becommanded by either spacecraft one or two, thanks to the

real-time wireless link (see [6]). Navigation for the assembledstructure is performed onboard the robot acting as the


Table 3: Position, augmented state filter parameters.

Q (1× 10−6) · I6×6 Process covariance matrix, for the adapted augmented state filterR (1× 10−4) · I2×2 Measurements covariance matrix, for the adapted augmented state filterPo Rk · I6×6 Initial covariance matrix, for the adapted augmented state filterTs 0.02 sec Simulation sampling time

Table 4: Attitude, augmented state filter parameters.

Q (1× 10−12) · I3×3 Process covariance matrix, for the adapted augmented state filterR (1× 10−2) Measurements covariance matrix, for the adapted augmented state filterPo Rk · I3×3 Initial covariance matrix, for the adapted augmented state filterTs 0.02 sec Simulation sampling time

master. For the two-robot configuration in Figure 12, itfollows the rigid body equations, being the attitude andattitude rate of the new spacecraft the same of the master, andits center of mass position and velocity deducted by those ofthe master (see also Figure 12).

7. Experimental Results: Four SpacecraftSimulators Assembly

In this section, assembly maneuvers are employed to experi-mentally test the suggested guidance, navigation, and controlschemes. For our experiment, we do not implement anycollision avoidance algorithm, which has been, however,previously successfully tested [23]. At the time of writingthis paper, the simulators do not have hardware dedicated torelative measurements. Relative measurements are assumedto be the range, line of sight, and relative attitude. Thisinformation is obtained via software, by having the robotsexchanging data over the ad-hoc wireless channel [6]. Thisfeature has the benefit of flexibility in imposing the desiredfrequency of measurement update, by simple modificationof the software. Furthermore, for the following experiments,we do not assume any particular noise or bias characteristicsfor the measurements, that is, the filter does not have thatinformation. Noise is present, and it comes from the wirelesscommunication. This assumption does not conflict with thepreviously stated contribution of our work in designing anestimation technique more robust than standard KalmanFiltering in the presence of low frequency updates, asdemonstrated in the following.

Two experimental runs are presented. The first onedemonstrates the unsuccessful relative navigation whenclassical Kalman Filtering is employed, considering theother S/C’s maneuvers as a random process. Only twosimulators are involved. The second experiment involvesthe four vehicles, showing how augmented state estimationcan handle low measurement updates and unpredictableinterruptions of updates, and still perform correct relativenavigation, driving the mission to success. In particular, weare here imposing, via the wireless network, an update of 2seconds. Once the two couples of robots are docked, eachof the assembled structure is considered to be a new vehicle

with new mass and geometry. For this reason, the augmentedstate estimator is reinitialized for the new structure withdifferent mass and inertia as in Table 2. For this part of theexperimentation, the software is running only onboard themaster S/C, that is, one pre-chosen unit for each couple.

The time step, or simulation sampling time, was chosento be: (1) higher than the thrusters minimum actuationtime (Tables 3 and 4), (2) in compatibility with CPUcomputational power, and (3) so to maintain the dynamicswithin the linearity range. The choice was also justifiedby previous experience with the employed hardware andby prior computer numerical simulations. In fact, theexperimental activities at the Spacecraft Robotics Laboratoryare always anticipated by high fidelity numerical simulationsof the test-bed dynamics in Simulink, by visualizing on thecomputer how the experiment will develop. This prototypingapproach reduces the time-to-market and trouble-shootingcosts of newly developed GNC methodologies, by signif-icantly cutting down the need for intermediate hardwareprototypes and the number of experimental tests.

7.1. The Classical Kalman Filter Technique. Figure 13 isthe bird’s eye view of the experiment, demonstrating theunfeasibility of classical Kalman filter for spacecraft relativenavigation, when relative measurements updates occur atlow frequency. Two spacecraft simulators start maneuvering,with the goal of docking, from a short distance. The sidesopposite to the bolded lines are the designated docking sides.After approximately 1 minute of maneuver, the accumulatederror in relative state vector (position) exceeds the toleranceof the docking interfaces (Figure 2), driving the vehicles intoa failed docking maneuver. A video of the experiment can befound online at [37].

7.2. The Augmented State Estimation Technique. Figure 14 isthe bird’s eye view of the experiment, demonstrating thefeasibility of the augmented state estimation for spacecraftrelative navigation. The main data for the filters are presentedin Tables 3 and 4. Four spacecraft simulators start maneuver-ing, with the goal of assembling into a line-shaped structure,from short distances. The sides opposite to the bolded linesare the designated docking sides. After less than 3 minutes of


maneuver, the four vehicles successfully complete the givenmission. The rectangular black and blue vehicles representtwo spacecraft simulators docked and maneuvering as asingle bigger unit. A video of the experiment can be foundonline at [38].

Once the simulators are assembled in couples, theymaneuver as a single bigger unit. In particular, the aug-mented state estimation is reinitialized in order to switchto a new target vehicle in terms of relative navigation. InFigure 14, for the left couple, the cyan-represented unit actsas master of the new assembled cyan-red spacecraft. Likewise,for the right couple, the green vehicle is the master in thegreen-magenta assembly.

8. Conclusion

In this work, we are suggesting a complete solution for guid-ance, navigation, and control of planar multiple spacecraftassembly maneuvers. Guidance is performed by dynamicallydefining a desired state vector, so that the spacecraft canprepare for docking and correctly connect. The control isbased on a real time LQR approach. As for the relativenavigation, augmented state estimation is proposed, allowingfor correct awareness of the other spacecraft configuration,even in the event of low frequency measurements update.The framework adapts itself to the evolving spacecraft,by switching among different values of mass propertiesand sensors and actuators configuration, when a new unitassembles to the aggregate.

Theoretical developments are presented for the three-degree-of-freedom case, considering a planar motion for therelative position and a single axis of rotation.

The experimental validation of the proposed method-ology is presented, via floating spacecraft simulators, usingan assembly maneuver as baseline. Experiments show howthe augmented state estimation can cope with low frequencymeasurement updates, correctly performing the relative nav-igation, driving the mission to success. On the other hand,Classical Kalman Estimation, is not accurate for close dis-tances with low frequency measurement updates as demon-strated in the three-degree-of-freedom experimental section.The dynamic guidance and control demonstrate real-timefeasibility and the capability of performing autonomousassembly.

Nomenclature

Acronyms

COTS: Commercial off the shelveDoF: Degree of freedomGNC: Guidance, navigation, and controlIE: Input estimationLQR: Linear quadratic regulatorLVLH: Local vertical local horizontal

reference frame centered on thechaser spacecraft

NPS: Naval postgraduate school

PWM: Pulse width modulationSRL: Spacecraft robotics laboratory.

Variables and Symbols

α: Docking safety cone semiapertureβ: Commanded orbiting angle

around target docking port indocking phase

Δt: Control system sample timeωLVLH: Chaser orbital angular velocityθ: Target attitude angle in chaser S/C

body frameθ̇: Target angular velocity in chaser

S/C body frameCjTi Φ: Transition Matrix of the dynamics

of the ith target S/C with respect tojth chaser S/C

CjTi ΦA: Augmented transition matrix of

the dynamics of the ith Target S/Cwith respect to jth chaser S/C

CjTi Θ: Attitude state vector of ith target

S/C with respect to jth chaser S/CCjTi ΘA: Augmented attitude state vector of

ith Target S/C with respect to jthchaser S/C

a: Scaling factor in R LQR weightingmatrix

k: Discrete time indexm,mcomb: Single & combined mass of the

spacecraft simulatorr: Torque Arm: Thruster-center of

mass armrrsw: Spacecraft-to-spacecraft vectorrgoal: Docking port-to-corresponding

docking spacecraft vectorrport: Center of mass to docking port

vectorrdock: Spacecraft-to-spacecraft transition

distance between far away phaseand docking phase

t: TimeuC : Chaser’s control vectoruT : Target’s control vectoru: Relative control vectoruLQR: Optimal control vectorumax > 0: Single engine maximum thrustuthr > 0: Threshold value for required

thrust Before using PWMA: State matrixB: Control distribution matrixC: State-output mapping matrixD: Control-output mapping matrixCjTi B: Control matrix referred of ith

target S/C with respect to jthchaser S/C


CjTi B(k): Discretized control matrix referred

of ith target S/C with respect to jthchaser S/C

CjTi BA(k): Discretized augmented control

matrix referred of ith target S/Cwith respect to jth chaser S/C

CjTi F: State matrix of ith target S/C with

respect to jth chaser S/CCjTi FA: Augmented state matrix of ith


G: Input noise matrixH: Measurement matrixGA: Augmented input noise matrixHA: Augmented measurement matrixJ : Cost functionJz, Jz comb: Inertia of a single and combined

Spacecraft simulator about thevertical axis

KLQR: LQR resulting gain matrixMC : Chaser spacecraft torqueMT : Target spacecraft torqueP0: Initial state error covariance

matrixQ: LQR state error weighting matrixQk: Process noise covariance matrixR: LQR control effort weighting

matrixRk: Measurement noise covarianceT: Maneuver total timeTs: Sampling timeV: Measurement noise vector,

assumed to be gaussian white zeromean with covariance Rk

V : Scaling factor in Q LQR weightingmatrix

W: Input noise vector, assumed to beGaussian white zero mean withcovariance Qk

CjTi X: State vector of ith target S/C with

respect to jth chaser S/C,CjTi XA: Augmented state vector of ith


Xrel: Relative state vector between twoS/C

CjTi S: Complete ith target state vector ref

to jth chaser S/CZ: Measurement vectorZA: Augmented measurement vectorIp×p : Identity matrix,0q×s: Zeros matrix[x y z]T : Target Cartesian coordinates in

chaser S/C body frame[ẋ ẏ ż]T : Target linear velocities in chaser

S/C body frame.

Acknowledgments

This research was performed while Dr. Bevilacqua washolding a National Research Council Research AssociateshipAward at the Spacecraft Robotics Laboratory of the US NavalPostgraduate School.

References

[1] S. A. Frost and M. J. Balas, “Adaptive key componentcontrollers for evolving systems,” in Proceedings of the AIAAGuidance, Navigation and Control Conference and Exhibit,Honolulu, Hawaii, USA, August 2008.

[2] M. J. Balas and S. A. Frost, “Evolving systems: a theoreticalfoundation,” in Proceedings of the AIAA Guidance, Navigation,and Control Conference and Exhibit, Keystone, Colo, USA,August 2006.

[3] T. S. VanZwieten and M. J. Balas, “Evolving systems: inheritingstability with decentralized adaptive control,” in Proceedingsof the AIAA Guidance, Navigation and Control Conference andExhibit, Hilton Head, SC, USA, August 2007.

[4] S. A. Frost and M. J. Balas, “Stabilizing controllers for evolvingsystems with application to flexible space structures,” inProceedings of the AIAA Guidance, Navigation and ControlConference and Exhibit, Hilton Head, SC, USA, August 2007.

[5] S. Dong, K. Allen, P. Bauer et al., “Self-assembling wirelessautonomously reconfigurable module design concept,” ActaAstronautica, vol. 62, no. 2-3, pp. 246–256, 2008.

[6] R. Bevilacqua, J. S. Hall, J. Homing, and M. Romano, “Ad hocwireless networking and shared computation for autonomousmultirobot systems,” Journal of Aerospace Computing, Informa-tion and Communication, vol. 6, no. 5, pp. 328–353, 2009.

[7] R. Bevilacqua and M. Romano, “Fuel-optimal spacecraftrendezvous with hybrid on-off continuous and impulsivethrust,” Journal of Guidance, Control, and Dynamics, vol. 30,no. 4, pp. 1175–1178, 2007.

[8] R. A. Singer, “Estimating optimal tracking filter performancefor manned maneuvering targets,” IEEE Transactions onAerospace and Electronic Systems, vol. 6, no. 4, pp. 473–483,1970.

[9] Y. T. Chan and F. Couture, “Manoeuvre detection and trackcorrection by input estimation,” IEE Proceedings F, vol. 140,no. 1, pp. 21–28, 1993.

[10] Y. Bar-Shalom, X. R. Li, and T. Kirubarajan, Estimation withApplications to Tracking and Navigation, John Wiley & Sons,New York, NY, USA, 2001.

[11] Y. Bar-shalom and K. Birmiwal, “Variable dimension filter-ing for maneuvering target tracking,” IEEE Transactions onAerospace and Electronic Systems, vol. 18, no. 5, pp. 621–629,1982.

[12] P. L. Bogler, “Tracking a maneuvering target using inputestimation,” IEEE Transactions on Aerospace and ElectronicSystems, vol. 23, no. 3, pp. 298–310, 1987.

[13] H. Khaloozadeh and A. Karsaz, “A new state augmentation formaneuvering targets detection,” in Proceedings of InternationalConference on Signal Processing and Communications (SPCOM’04), pp. 65–69, December 2004.

[14] M. H. Bahari, A. Karsaz, and M. B. Naghibi-S, “Intelligenterror covariance matrix resetting for maneuver target track-ing,” Journal of Applied Sciences, vol. 8, no. 12, pp. 2279–2285,2008.

[15] M. H. Bahari and N. Pariz, “High maneuvering target trackingusing an input estimation technique associated with fuzzy


forgetting factor,” Scientific Research and Essays, vol. 4, no. 10,pp. 936–944, 2009.

[16] V. Pellegrini, R. Bevilacqua, M. Romano, and F. Curti, “Space-craft proximity navigation and autonomous assembly basedon augmented state estimation: analysis and experiments,” inProceedings of the AIAA Guidance, Navigation, and ControlConference, Toronto, Canada, August 2010.

[17] S. B. McCamish, M. Romano, S. Nolet, C. M. Edwards, andD. W. Miller, “Flight testing of multiple-spacecraft controlon SPHERES during close-proximity operations,” Journal ofSpacecraft and Rockets, vol. 46, no. 6, pp. 1202–1213, 2009.

[18] http://mathworks.com/matlabcentral/fileexchange/2651.[19] P. Zarchan and H. Musoff, Fundamental of Kalman Filtering: A

Practical Approach, AIAA, Reston, Va, USA, 2009.[20] M. Romano, D. A. Friedman, and T. J. Shay, “Laboratory

experimentation of autonomous spacecraft approach anddocking to a collaborative target,” Journal of Spacecraft andRockets, vol. 44, no. 1, pp. 164–173, 2007.

[21] J. L. Schwartz, M. A. Peck, and C. D. Hall, “Historical reviewof air-bearing spacecraft simulators,” Journal of Guidance,Control, and Dynamics, vol. 26, no. 4, pp. 513–522, 2003.

[22] F. Curti, M. Romano, and R. Bevilacqua, “Lyapunov-basedthrusters’ selection for spacecraft control: analysis and exper-imentation,” Journal of Guidance, Control, and Dynamics, vol.33, no. 4, pp. 1143–1160, 2010.

[23] R. Bevilacqua, T. Lehmann, and M. Romano, “Developmentand experimentation of LQR/APF guidance and control forautonomous proximity maneuvers of multiple spacecraft,”Acta Astronautica, vol. 68, no. 7-8, pp. 1260–1275, 2011.

[24] https://www.rtai.org/RTAILAB/RTAI-Lab-tutorial.pdf.[25] http://www.mathworks.com/products/xpctarget/.[26] L. Dozio and P. Mantegazza, “Real time distributed control

system using RTAI,” in Proceedings of the 6th IEE InternationalSymposium on Object-Oriented Real-Time Distributed Com-puting, May 2003.

[27] G. Quaranta and P. Mantegazza, “Using MATLAB-simulinkRTW to build real time control applications in user space withRTAI-LXRT,” in Realtime Linux Workshop, Milan, Italy, 2001.

[28] S. Ricci and A. Scotti, “Aeroelastic testing on athree surface airplane,” in Proceedings of the 47thAIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamicsand Materials Conference, pp. 7131–7147, Newport, RI, USA,May 2006.

[29] L. Dozio, A. Toso, W. Corbetta, E. Vigoni, and G. Ghiringhelli,“Comparison of feedback and feedforward strategies for activestructural—acoustic control of broadband sound transmis-sion into a cavity,” in Proceedings of the 13th AIAA/CEASAeroacoustics Conference, 28th AIAA Aeroacoustics Conference,2007.

[30] M. G. Perhinschi, M. R. Napolitano, G. Campa, B. Seanor, S.Gururajan, and G. Yu, “Design and flight testing of intelligentflight control laws for the WVU YF-22 model aircraft,”in Proceedings of AIAA Guidance, Navigation, and ControlConference, pp. 5925–5936, San Francisco, Calif, USA, August2005.

[31] Y. Zheng, B. J. Martin, and N. Villaume, “VERSE—virtualequivalent real-time simulation environment,” in Proceedingsof AIAA Modeling and Simulation Technologies Conference, pp.368–376, San Francisco, Calif, USA, August 2005.

[32] W. H. Clohessy and R. S. Wiltshire, “Terminal guidance systemfor satellite rendezvous,” Journal of the Aerospace Sciences, vol.27, no. 9, pp. 653–658, 1960.

[33] S. D’Amico, E. Gill, M. Garcia, O. Montenbruck, and E. Gill,“GPS-based real-time navigation for the PRISMA formation

flying mission,” in Proceedings of the 3rd ESA Workshop onSatellite Navigation User Equipment Technologies (NAVITEC’06), Noordwijk, The Netherlands, December 2006.

[34] S. D’Amico, J. S. Ardaens, and O. Montenbruck, “Navigationof formation flying spacecraft using GPS: the PRISMAtechnology demonstration,” in Proceedings of the 22nd Interna-tional Technical Meeting of the Satellite Division of the Instituteof Navigation (ION-GNSS ’09), pp. 2250–2264, Savannah, Ga,USA, September 2009.

[35] K. Ogata, Discrete-Time Control Systems, Prentice-Hall, NewYork, NY, USA, 2nd edition, 1995.

[36] C. Lugini and M. Romano, “A ballistic-pendulum test stand tocharacterize small cold-gas thruster nozzles,” Acta Astronau-tica, vol. 64, no. 5-6, pp. 615–625, 2009.

[37] http://aa.nps.edu/∼mromano/Movies4Web/GNC10 Video1.flv.

[38] http://aa.nps.edu/∼mromano/Movies4Web/GNC10 Video2.flv.


Research Article

Cascade-Based Controlled Attitude Synchronization andTracking of Spacecraft in Leader-Follower Formation

Rune Schlanbusch,1 Antonio Lorı́a,2 and Per Johan Nicklasson1

1 Department of Technology, Narvik University College, PB 385, 8505 Narvik, Norway2 CNRS, LSS-SUPELEC, 3 rue Joliot Curie, 91192 Gif-sur-Yvette, France

Correspondence should be addressed to Rune Schlanbusch, [email protected]

Received 30 June 2010; Revised 5 January 2011; Accepted 23 February 2011

Academic Editor: Giampiero Campa

Copyright © 2011 Rune Schlanbusch et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

We propose controllers for leader-follower attitude synchronization of spacecraft formations in the presence of disturbances, thatis, the leader spacecraft is controlled to follow a given reference, while a follower spacecraft is controlled to synchronize its motionwith the leader’s. In the ideal disturbance-free scenario, we show that synchronization takes place asymptotically. Moreover,we show the property of uniform practical asymptotic stability which implies that the synchronization is robust to boundeddisturbances.

1. Introduction

In recent years, formation flying has become an increas-ingly popular subject of study. This is a new method ofperforming space operations, by replacing large and complexspacecraft with an array of simpler microspacecraft, bringingout new possibilities and opportunities of cost reduction,redundancy, and improved resolution aspects of onboardpayload. One of the main challenges is the requirement ofsynchronization between spacecraft; robust and reliable con-trol of relative position and attitude is necessary to make thespacecraft cooperate to gain the possible advantages madefeasible by spacecraft formations. For fully autonomousspacecraft formations, both path- and attitude-planningmust be performed online which introduces challenges likecollision avoidance and restrictions on pointing instrumentstowards required targets, with the lowest possible fuelexpenditure.

Synchronization of dynamical systems was first studiedby Christian Huygens in the XVIIth century. In recent years,the problem has obtained increasing interest in variousresearch areas due to its impact in technology developmentand challenges it imposes; see, for example, [1–4].

Model-based controlled synchronization consists inusing the physics laws and control theory in order toinduce synchronization in dynamical systems. Successful

instances include synchronization of robot manipulators[5, 6], leader-follower spacecraft formations [7–10], andship replenishment operations [11, 12]. Another form ofsynchronization is consensus, in which a group of systemscoordinate their motion without any subsystem having ahigher hierarchy with respect to the others. An instanceof consensus is cooperative control in which a group ofsystems is controlled in a way that they collaborate inorder to achieve a task as a team of agents. Examplesmay be found in the areas of autonomous vehicles [13–15], underactuated marine vessels [16, 17], and rigid bodies[18–20].

In this paper, we address the simultaneous controlproblems of attitude tracking and leader-follower synchro-nization. That is, we propose a tracking controller forthe leader spacecraft which makes it follow a prescribedreference. Independently, we construct a synchronizationcontrol law for the follower spacecraft which makes it trackthe attitude of the leader, thereby synchronizing in theclassical master-slave configuration.

Our controllers are reminiscent of classical trackingcontrollers for robot manipulators passivity-based controlwhich exploits the system’s physical properties; see [21]. Fortracking control, see the passivity-based PD+ of [22] and thewrongly called “sliding-mode” controller of [23] which mayrather be casted in the passivity-based framework.


u2 y2 y1Σ2 Σ1

Figure 1: Cascade interconnection of two dynamical systems.

Although insightful, these popular control approachesfor robot manipulators may not be directly applied inspacecraft tracking control and synchronization. The firstobstacle is the specificity of spacecraft nonlinear models,expressed in quaternion coordinates. We revise the model inthe following section.

Besides the difficulties imposed by the modeling ofspacecraft, simultaneous tracking control and master-slavesynchronization implicitly suggest controlling the leaderspacecraft towards a reference independently of the slavesystem dynamics. Correspondingly, the synchronizationcontroller inevitably couples the follower spacecraft tothe dynamics of the leader. However, the synchronizationcontroller is demanded to achieve the task regardless ofthe master dynamics as well as the reference that systemintends to track. The ability to control two coupled systemsseparately is called separation principle and is known not tohold in general for nonlinear systems (see e.g. [24]). This iswhere cascades theory enters in play.

Cascaded systems theory consists in analyzing complexsystems by dividing them in subsystems simpler to controland to analyze (see [25] and references within). It mustbe emphasized that such representation is purely schematic,for the purpose of analysis only. Generally speaking, thestability analysis problem consists in finding conditions fortwo systems as in Figure 1 so that, considering that bothsubsystems separately are stable, they conserve that propertywhen interconnected.

In the context of the present paper, the block onthe left corresponds to the leader system in closed loopwith a tracking controller, while the block on the rightconsists in the follower spacecraft in closed loop with thesynchronization controller. The blocks are interconnected viathe tracking errors of the leader system. Hence, in the idealcase, when the leader spacecraft is perfectly controlled, thesystems are decoupled.

The topic of cascaded systems have received a great dealof attention and has successfully been applied to a widenumber of applications. In [26], a cascaded adaptive controlscheme for marine vehicles including actuator dynamics wasintroduced, while [27] solved the problem of synchroniza-tion of two pendula through use of cascades. The authorsof [28] studied the problem of global stabilizability of feed-forward systems by a systematic recursive design procedurefor autonomous systems, while time-varying systems wereconsidered in [29] for stabilization of robust control, while[30] established sufficient conditions for uniform globalasymptotical stability (UGAS) of cascaded nonlinear time-varying systems. The aspect of practical and semiglobalstability for nonlinear time-varying systems in cascade waspursued in [31, 32]. A stability analysis of spacecraft forma-tions including both leader and follower using relative coor-dinates is presented in [10], where the controller-observer

scheme is proven input-to-state stable, and backsteppingwas applied in [17] for leader-follower formation controlof multiple underactuated autonomous underwater vehicles(AUVs). For the control problems at hand, we show that theclosed-loop system has the property of uniform asymptoticstability. Significantly, uniform asymptotic stability guar-antees robustness with respect to bounded disturbances.In this regard, we extend our result to the case wherebounded perturbations affect the system (atmospheric drag,gravity gradient, etc.). In this scenario, we guarantee uniformpractical asymptotic stability. This pertains to the case whenthere exists a steady-state tracking and synchronization errorwhich can be arbitrarily diminished via an appropriatetuning of the control parameters.

The contribution of this paper is application of theframework for stability analysis of cascaded systems of rigidbodies in leader-follower formation and synchronizvation ofPD+ and sliding surface control laws adapted for quaternionspace. The equilibrium points of the PD+ controller inclosed loop with the rigid body dynamics are provenuniformly asymptotically stable (UAS) when disturbancesare considered known, while a sliding surface controller isutilized to prove uniform practical asymptotical stability(UPAS) when disturbances are considered unknown butbounded. Simulation results of a leader-follower spacecraftformation are presen

)RUPDWLRQ )OLJKW &RQWURO...Each spacecraft involved in the assembly is both chaser and target at the...

Documents

Transcript of )RUPDWLRQ )OLJKW &RQWURO...Each spacecraft involved in the assembly is both chaser and target at the...