Attitude Control Challenges and Solutions for Hayabusa Spacecraft

Masashi Uo* , Kenichi Shirakawa** , Tatsuaki Hashimoto*** , Takashi Kubota*** , Jun’ichiro Kawaguchi***

*NEC TOSHIBA Space System, Ltd **NEC Aerospace System, Ltd ***Institute of Space and Astronautical Science(ISAS),JAXA

Hayabusa attempted to touch-down two times in November, 2005 after two years of cruising and two months of proximity observation around ITOKAWA. One of three reaction wheels failed in late July of 2005, and the second wheel failed in early October. While the attitude stabilization logic with two wheels was inherent in the onboard system, the stabilization with one wheel required the use of onboard re-programming function and special ground support operations. This paper describes how the attitude control system and operation were adapted to cope with the loss of reaction wheels, along with the introduction of attitude control system that was designed to realize various attitude functions with minimum onboard resources.

「はやぶさ」における RW 故障対応姿勢制御

卯尾匡史* , 白川健一** , 橋本樹明*** , 久保田孝*** , 川口淳一郎*** *NEC 東芝スペースシステム株式会社, **NEC 航空宇宙システム株式会社 ***宇宙航空研究開発機構宇宙科学研究本部

はやぶさは 2 年間の巡航飛行の後、イトカワ近傍で 2 ヶ月の観測運用を行い、 2005 年の 11 月には 2 回のタッチダウンを実施した。イトカワ到着直前の 2005 年 7 月末に 3 台のうち 1 台目のホイールを失い、観測運用途中の 10 月初めに 2 台目を失った。 1 台故障時の姿勢制御方式はあらかじめバックアップ制御として姿勢制御系に組み込まれていたが、 2 台故障に対応するには、搭載系プログラムの一部変更と新規の地上ソフトウェアによる支援が必要であった。本報告では、RW 故障時に実施した対策を、フライトデータをまじえて述べる。

I. Introduction The Japan Aerospace Exploration Agency (JAXA)/Institute of Space and Astronautical Science (ISAS) launched a sample and return spacecraft MUSES-C, renamed as "HAYABUSA" in orbit, toward a near earth asteroid 1998SF36(ITOKAWA) on May 9th,2003. After about two years travel propelled via ion engines and an Earth swing-by, HAYABUSA arrived at ITOKAWA on September 12th, 2005. Spending two months for remote observations at an altitude range of 3km to 20Km, Hayabusa made numerous scientific and engineering achievements including reconstructed 3-D model of ITOKAWA. The touchdown onto the surface of Itokawa was attempted two times in late November, preceded by several rehearsals of descending operation. Two of three reaction wheels were lost before November, severely restricting the attitude and orbit operations. Onboard parameters and operation strategy were changed rapidly and efficiently to cope with this unexpected situation, so that successful touchdown was achieved finally. Table-1 summarizes the major events and the incident of wheel failures. This paper describes how the attitude control system and operation were adapted to cope with the loss of reaction wheels, along with the introduction of attitude control system that was designed to realize various attitude functions with minimum onboard resources.

Table-1 Major events of HAYABUSA and RW failures July 30, 2005 RW-X failure September 12 Arrival at ITOKAWA (distance 20Km) September 30 Arrival at Home Position (distance 7Km) October 3 RW-Y failure October 5 - 25 Scientific observation (Home Position tour) November 4 First rehearsal descending November 19 Touchdown #1 November 25 Touchdown #2

II. Attitude Control system of HAYABUSA The Guidance Navigation and Control (GNC) system of HAYABUSA spacecraft was designed so that it can cope with various situations within the spacecraft's severe weight and power restrictions. As the description of GNC system and GNC components is covered in Ref-4 and Ref-5, this paper focuses on the description of attitude determination and control function. Table-2 summarizes the components used for attitude determination and control. The GNC logic, including attitude determination and control logic, is implemented in AOCU (Attitude and Orbit Control Unit), where a high performance microprocessor is equipped. TSAS (Two axis Sun Aspect Sensor), STT (Star Tracker) and IRU (Inertial Reference Unit) are combined to determine the spacecraft attitude. RW (Reaction Wheel) and RCS (Reaction Control System) thrusters are used for attitude and position control. Due to severe weight restriction, HAYABUSA has only three wheels which are mounted so that their rotation axes align with X, Y, Z axis of spacecraft body. When all the wheels are operational, each wheel controls the corresponding axis directly. This means that the wheel system has no redundancy. If any of the wheels fail, the normal attitude

1 control becomes impossible. Instead of adding redundant wheels, functional redundancy was realized by implementing DRW (Dual Reaction Wheel) mode logic, a logic to stabilize the spacecraft attitude with two wheels, in AOCU. The thrusters are installed on the spacecraft so that translational and rotational motion can be controlled independently. HAYABUSA has ion engine system (IES) as primary propulsion system to bring the spacecraft to ITOKAWA. IES has four thruster heads. The heads are mounted on the spacecraft so that their thrust vectors almost align with -X axis of the spacecraft and pass through the nominal location of spacecraft center of gravity. Any deviation of the thrust vectors from actual center of gravity produces small torque around Y and Z axis that eventually develops the angular momentum of the spacecraft. To minimize this torque, IES heads are mounted on two axis (Y and Z) gimbals that are controlled by commands from AOCU. The AOCU controls the gimbals so that the spacecraft angular momentum around Y and Z axis is held constant.

Table-2 Components for attitude determination and control

Component Qty. Specification Two axis Sun Aspect Sensor 1 Field of view +/-50 x +/-50 deg (TSAS) Accuracy better than 0.05 deg Inertial Reference Unit 2 Range 1432 deg/s Max -4 (IRU) Resolution 1.093×10 deg Bias stability 3 deg/h（3σ） 1/2 Random walk 0.07 deg/hr STar Tracker (STT) 1 Field of view 30 × 40 deg Star dynamic range 3～-1 (CCD Magnitude) Accuracy (random, bias) 3arcmin (3sigma), 1arcmin (3sigma) Update rate 1 Hz

Reaction Wheel(RW) 3 Maximum momentum 4 Nms (@5000rpm) Torque 12mNm minimum

Reaction Control System 12 Thrust 20N nominal (RCS) jets Minimum impulse 0.2Ns

III. Attitude Control with 2 wheels

A. DRW control design 1. Requirement for DRW control logic The DRW control logic is implemented in the onboard software of AOCU as backup control logic for the case one of three wheels fails. In the normal situation, in which three wheels are all operational, High Gain Antenna rotation of three axes is independently controlled with the wheel that is ＨＧ exactly aligned with each axis. When one of the wheels fails the rotation around the corresponding axis becomes uncontrollable. The straightforward solution is to control the axis with reaction jets. The reaction jet control, however, requires much fuel consumption. The spacecraft carries fuel so that the attitude control during the mission phase may be accomplished with reaction jets, it is impossible to fully use reaction jets for transfer or return phase that is 2 and 1.5 years long respectively. The DRW control was introduced so that the spacecraft attitude can be maintained with two wheels and with minimum use of reaction jets. The ion engines that are mounted on two axis gimbals are also utilized LIDAR beam Field of view of optical for attitude control. camera The attitude motion of the spacecraft is governed by total angular momentum of the spacecraft itself. It is obviously necessary that the Figure-1 Spacecraft configuration wheels should hold all the angular momentum to control the body rate to zero. If the spacecraft has three wheels, and if they are mounted mutually orthogonal, the distribution of the angular momentum on the wheels is arbitrary. It means that the spacecraft can be stabilized in any attitude if the spacecraft has (more than) three wheels. If one of the wheels fails, the angular momentum expressed in body frame is restricted in the plane that is formed by remaining two wheels. This explains the natural fact that only two degrees of freedom are controllable with two wheels, if the spacecraft has residual angular momentum. In the case of HAYABUSA, as is explained later, it is not possible to accurately form the angular momentum vector to desired direction. The DRW logic for HAYABUSA is designed so that the controllable two degrees of freedom are used for Z-axis pointing (i.e. control of X-axis and Y-axis rotation). As shown in the figure-1, High gain antenna for communication is aligned with +Z axis. The LIDAR and optical cameras for asteroid observation are aligned with -Z axis. So the communication with the earth requires the +Z axis to be pointed to the earth. The asteroid observation requires –Z axis to be pointed to the asteroid. The Z axis pointing (control of the attitude around X and Y axis) is thus most important.

2. DRW control design To stabilize the attitude with two wheels and without reaction jets, the only possible solution is to utilize the momentum of remaining wheels. Inherent problem of the momentum stabilization scheme is the nutation. In the DRW logic, one of the remaining wheels is assigned to work as momentum wheel, and the other wheel is assigned to damp the nutation. To implement the logic, the momentum axis is denoted by i-axis, and the nutation damping wheel axis is denoted by j-axis and another axis, the axis of failed wheel, is denoted by k-axis. The i-, j-, k- axes are assigned to X, Y, Z axis depending on the failed wheel. The adopted assignment is summarized in figure-2. The principle of the assignment is to make Z- axis pointing possible as is described earlier. Torque command to i-axis wheel and j-axis wheel is a linear combination of three control logics: PCL, NDMP, and HPC. PCL is the normal attitude control utilizing PD control logic. NDMP is the nutation damping logic with simple gain selection logic to enhance the nutation damping performance. HPC controls the distribution of momentum between i- and j- axes, thus realizes control of the rotation around k-axis. The control of ion engine gimbals is added both for reaction wheel momentum maintenance and for attitude control. Utilizing the ion engine control logic, three axis control becomes possible for RW-Y fail case and RW-Z fail case. In the case of RW-Z failure, without ion engine control, rotation around Y-axis is uncontrollable. This means that Z-axis pointing is impossible for this case. To solve this problem, special control mode is prepared for RW-Z failure case. In this special mode, X and Y axis is controlled with

2 normal wheel control logic, and Z-axis is controlled with reaction jets. As the attitude requirement around Z-axis is not severe, the angle width of RCS control logic can be widened to save fuel consumption. The details of the control logics are as follows. Some explanation how the logic works will be described in the chapter of flight experience. Table-2 summarizes the control logics that are applied to each axis.

Z Z Z

Y Y Y

k j j i j

k k i i

X X X RW-X fail RW-Y fail RW-Z fail

Figure-2 Assignment of X,Y,Z to i,j,k axis

Table-2 DRW control law assignment

RW-X fail i j k Corresponding Axis Y Z X Control law PCL NDMP HPC SIESUL , SIESAC 1 , 0 1 , 0 N/A

RW-Y fail i j k Corresponding Axis X Z Y Control law PCL NDMP + PCL HPC SIESUL , SIESAC N/A 1 , 0 0 , 1

RW-Z fail i j k Corresponding Axis X Y Z Control law PCL NDMP + PCL HPC SIESUL , SIESAC N/A 1 , 0 0 , 1

RW-Z fail(*) i j k Corresponding Axis X Y Z Control law PCL PCL RCS (*)Z-axis pointing without IES

According to the assignment of table-2, torque command to RW-i and RW-j is calculated as follows.

TRWi＝TPCLi + THPCi TRWj＝TPCLj + THPCj + TNDMPj

Axis Control law i - KPDRWj * θERRj -KDDRWj * ωSCj TPCL j KPN * ( - KPDRWj * θERRj -KDDRWj * ωSCj ) i N/A TNDMP j -KNDRWj * ωSCj T i -( i ・(k ×Unit(H ij)) ) * KHDRWk * θERRk THPC T j -( j ・(k ×Unit(H ij)) ) * KHDRWk * θERRk

i , j , k : unit vector representing i-, j-, k- axis e.g. if i=X then i=(1 0 0)T H ij : angular momentum vector in i-j plane. =Hi・i ＋Hj・j KPDRW : Gain for Angular error KDDRW : Gain for Angular rate error KHDRW : Gain for Momentum exchange rate KPN : Gain for j-axis attitude control adopted only when SIESAC=1 KNDRW : Gain for NDMP control θERR# : Attitude control error angle ω# : Attitude rate

Ion engine gimbals are also utilized for momentum control of wheels and attitude control. Ion engines can produce the torque around Y and Z axis by offsetting the gimbals angles from the position that the ion engine force pass through the spacecraft center of gravity. The gimbals are controlled according to the following torque requirement.

3 TIESY ＝ SIESULY*TIESULY＋SIESACY*TIESACY TIESZ ＝ SIESULZ*TIESULZ＋SIESACZ*TIESACZ

TIESUL#＝-KPIESUL*dΩRW＃-KIIESUL*ΩDI# TIESAC#＝-KPIESAC#*θERR# -KIIESAC#*integ(θERR#) - KDIESAC#*ω# #= Y or Z

KPIESUL : Proportional gain for IES unloading KIIESUL : Integral gain for IES unloading KPIESAC# : Proportional gain for IES attitude control KIIESAC# : Integral gain for IES attitude control KDIESAC# : Derivative gain for IES attitude control SIESUL# : IES unloading execution flag defined in table-2 SIESAC# : IES attitude control execution flag defined in table-2

dΩRW＃ : Difference of current wheel rate from nominal rate

ΩDI# : Integrated dΩRW＃

3. Operation strategy of DRW control The DRW logic is essentially a bias momentum scheme that is effective for steady-state pointing. The logic can deal with small angle (less than a few degrees) maneuvers and low-rate target tracking such as asteroid tracking and earth tracking. Large angle attitude maneuver and orbit- maneuver (delta-V operation) shall, however, be performed utilizing reaction jets. The use of reaction jets inevitably perturbs the distribution of angular momentum in the spacecraft. The balance of the angular momentum is also disturbed by external torque such as solar-pressure torque and torque coupling from ion engine thrusting. Even when the angular momentum around Y and Z axis is maintained by gimbals control, small but uncontrollable torque around X axis is generated due to small differences of thrust vectors of ion engine heads. The implementation of DRW logic on the spacecraft takes into account such actual operation requirement. The mode transfer chart of the onboard software is shown in the figure-3. To accomplish the steady state DRW stabilization, the following steps are executed. (1) Initial attitude acquisition (ACQ) The spacecraft attitude is oriented to the target attitude within the accuracy of a few degrees using reaction jets. (2) Momentum initialization (INZ) The RW rotation rate is controlled to desired rate. Nominal rate is 2500rpm for i-axis and 1500rpm for j-axis. The spacecraft attitude rate shall be minimized here so that the angular momentum of the spacecraft body is within the range that the DRW logic can deal with. (3) DRW damping (DMP) The DRW logic is activated after the angular momentum condition is settled by the INZ step. It usually takes 30 to 60 minutes because the nutation damping performance is limited by the spacecraft nutation motion whose period is around 15 to 20 minutes. (4) DRW momentum hold (MHLD) When the attitude error for i- and k- axis becomes small, the DRW mode transfers into MHLD mode in which the j-axis attitude and i-, and j- wheel rate is watched. The control logic is the same with DMP mode. When the attitude error of j-axis exceeds the predetermined limit or when the wheel rotation rate of i- and j- axis differs with nominal rate by the predetermined tolerance, the onboard software commands the mode to repeat from ACQ mode. The DRW control is also initiated from ACQ mode when the large maneuver and delta-V is required.

Control with reaction jets

Initial Attitude Acquisition (ACQ)

Attitude Acquired

(DMP)

Attitude Error < Limit

DRW momentum hold (MHLD)

Figure-3 DRW mode control

B. Flight performance of DRW control Figure-4 shows a typical example of DRW performance in flight. Figure-5 shows an example of automatic DRW reset operation. The torque coupling of ion engine produces small disturbance torque around X axis (k-axis). There is no direct actuator, except for reaction jets, to cancel the X-axis angular momentum that gradually develops by the torque coupling. The DRW logic works so that the angular momentum of i-wheel(Y- axis for RW-X failure case) covers the momentum of k-axis by controlling the j-axis (Z-axis) rotation as is illustrated in figure-6. Thus, the Z-axis gradually rotates with X-axis angular momentum develops while the X-axis and Y-axis rotation is maintained. When the j-axis attitude error exceeds a limit (10.5degrees for the case of the figure-5), the DRW logic is automatically initialized.

Delta-V for approach 2 X-axis (k-axis) Attitude Error(deg) 1.5 1 0.5 0 -0.5 -1 -1.5 -2 2005/9/1 2:20 2005/9/1 2:30 2005/9/1 2:40 2005/9/1 2:50 2005/9/1 3:00 2005/9/1 3:10 2005/9/1 3:20 0.4 0.3 Y-axis (i-axis) Attitude Error(deg) 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 2005/9/1 2:20 2005/9/1 2:30 2005/9/1 2:40 2005/9/1 2:50 2005/9/1 3:00 2005/9/1 3:10 2005/9/1 3:20 4 Z-axis (j-axis) Attitude Error(deg) 3 2

1 0

-1 9/1 2:20 9/1 2:30 9/1 2:40 9/1 2:50 9/1 3:00 9/1 3:10 9/1 3:20 RW&RCS DRW (ACQ->INZ->DMP) DRW (MHLD)

Figure-4 Typical operation of DRW

Automatic DRW initialization 4 Attitude Control Error (deg) X-axis(k-axis) 2

0 Y-axis(i-axis) -2

-4 Z-axis(j-axis) -6 -8

-10 Gradual increase of j-axis control error -12 due to torque coupling from ion engine

-14 05/8/13 22:00 05/8/14 0:00 05/8/14 2:00 05/8/14 4:00 05/8/14 6:00 05/8/14 8:00

Figure-5 Automatic DRW initialization

k k k

j-wheel momentum i-wheel momentum j j j

i Spacecraft total momentum i i

Steady state pointing When k-axis momentum develops, As a consequence of nutation damping(*),

Total momentum is maintained by the spacecraft shall exhibit j-axis rotates so that total momentum is held

i- and j-axis wheels. nutational motion. by i- and j- wheel momentum again.

(*)with i- and k-axis rotation constrained

Figure-6 Explanation of j-axis rotation due to development of k-axis angular momentum

5 IV. Attitude Control with 1 wheel On October 3rd, the second wheel: RW-Y stopped. The only possible attitude control scheme is to use reaction jets for the axes of failed wheels. To minimize fuel consumption, the following operational consideration was applied. (1) The attitude control width was expanded to +/-3 degrees or 5 degrees except for the time of earth communication (HGA pointing) and asteroid observation. For earth and asteroid pointing, the attitude width was reduced to within +/-0.7 degrees. The duration for earth / asteroid pointing was decreased as short as possible (2) Angular momentum trimming operation was introduced to keep the spacecraft in free nutation mode realized by the angular momentum of remaining wheel (RW-Z). To maximize the free motion period, the initial condition of the motion is essential. The fine control of the initial condition was undertook using two method: FNI(Free Nutation Initiation) and AMT(Angular Momentum Trimming). (3) Performance of reaction jets at small pulse width was tested and evaluated on the ground. As the result of the test, it was found that pulse of 5~6msec can be used though the repeatability of impulse is degraded than the nominal pulse width of 15msec. After the several on-orbit confirmations, the short pulse width was actually employed for initial angular momentum control in FNI and AMT.

A. FNI (Free Nutation Initiation) The FNI control utilizes the GSP(Guidance Sequence Program) function that was implemented on AOCU to handle the complicated touchdown sequence. The GSP function uses a set of predetermined event flags and tables of actions corresponding to the events. The tables and actions are reprogrammable from the ground. The event includes the "convergence of attitude control". And the command that can be set in the action table includes the command to change attitude control parameter. Using the event and command, the control flow described in figure-7 was programmed in GSP.

B. AMT (Angular Momentum Trimming) The FNI was effective for reducing the fuel consumption. The performance of angular momentum trimming was, however, not precise enough to realize earth and asteroid pointing. That is because the onboard function for attitude convergence judgment has limitation about parameter tuning range. The AMT control was introduced to improve the pointing accuracy than the FNI control. The accuracy target was initial nutation of less than 0.7degrees. The attitude error angle and attitude rate is monitored on the ground. The attitude motion is propagated to determine the proper firing timing to reduce the nutation and to reorient the angular momentum vector to desired direction. To improve the accuracy of propagation, the output rate of attitude packet was increased by manipulating the packet control table in onboard data handling unit. For thruster firing, one of the commands that was defined for AOCU testing was utilized. This command allows activating one of twelve thrusters with very short pulse width: selectable within 8msec. The command is sent to onboard data handling unit with time-tag to execute the command.

Set attitude control parameter “fine mode”

NO Converged?

YES

Set attitude control parameter “wide mode”

NO Attitude YES error>Limit?

Figure-7 FNI control flow

V. Conclusion This paper described the outline of attitude control logic adopted after the failure of reaction wheels. The DRW logic, employed after the first wheel failure, has worked as designed. Ion engine guidance was also carried out based on DRW attitude stabilization. Effect of the first wheel failure was almost negligible by utilizing the DRW logic. Thus, HAYABUSA could arrive at ITOKAWA and could perform many fruitful science observations. The second wheel failure was out of sight of system design. By quick development of logic and ground tool to support the single wheel operation, the effect of wheel failure was minimized.

References 1 http://www.hayabusa.isas.jaxa.jp/ [cited 1 August 2006]. 2 J.Kawaguchi, T. Satoh, T. Kominato, M. Kimura, M. Uo, N. Muranaka, “An Optical Guidance and Navigation in Approach to Small Celestial Bodies” 53rd International Astronautical Congress, IAC, Houston, Texas, 2002 3 Jun'ichiro Kawaguchi, Masatoshi Matsuoka, Takashi Kominato, "Ion Engines Cruise of Hayabusa to Itokawa - Trajectory Synthesis and Results", AAS06-210 4 Masashi Uo, Ken'ichi Shirakawa, Tatsuaki Hasimoto, Takashi Kubota, Jun'ichiro Kawaguchi, " Hayabusa's Touching-down to ITOKAWA - Autonomous Guidance and Navigation", AAS06-215 5 Tatsuaki Hasimoto , Takashi Kubota, "Use of Laser Range Finders in Terrain Alignment and Touchdowns" This conference AAS06-216 6 H. Morita, K. Shirakawa, T. Kubota, T. Hashimoto, J. Kawaguchi, “Hayabusa, Detailed Guidance and Navigation Operations During Descents and Touchdowns” AIAA/AAS Astrodynamics Specialist Conference, AIAA/AAS, Keystone, Colorado, 2006, AIAA-2006-6536 7 T. Hashimoto, T. Kubota, S. Sawai, M. Uo, J. Kawaguchi, “Final Autonomous Descent based on Target Marker Tracking” AIAA/AAS Astrodynamics Specialist Conference, AIAA/AAS, Keystone, Colorado, 2006, AIAA-2006-6538