Paul B. Brugarolas(1), A. Miguel San Martin (2), Edward C. Wong (3)
Jet Propulsion Laboratory, California Institute of Technology
4800 Oak Grove Dr., Pasadena, CA, 91109 (USA)
(1) Email:
(2) Email:
(3) Email:
This paper describes the RCS 3-axis attitude control
system for the exo-atmospheric and guided entry
phases of the Mars Science Laboratory. The controller
is formulated as three independent channels in the
control frame, which is nominally aligned with the
stability frame. Each channel has a feedforward and a
feedback path. The feedforward path enables fast
response to large bank slews. The feedback path
stabilizes the vehicle angle of attack and sideslip
around its trim position, and tracks bank commands.
The performance of this design is demonstrated via
The Mars Science Laboratory (MSL) is the next NASA
rover mission to Mars. It will launch in 2011 and
deliver a ~900 kg mobile science laboratory, a rover
named “Curiosity”, to the surface of Mars. MSL aims
to deliver this rover within a ~20 km landing circular
region. MSL uses a guided atmospheric entry capsule
to achieve this targeting performance. The entry
capsule guidance, navigation and control system
employs a set of ejectable balance masses, a descent
inertial measurement unit (DIIMU), and a propulsive
reaction control system (RCS). The ejectable balance
masses shift the capsule center of mass enabling
generation of a lift vector during the atmospheric
phase. A guidance law that controls this lift vector
through banking maneuvers corrects for down-track
and cross-track errors. A navigation filter integrates the
DIMU measurements to estimate the position and
attitude of the capsule. The positional information is
used by the guidance law. The attitude information is
used by the attitude controller to track bank commands
and stabilize the angle of attack and sideslip angle. The
attitude controller generates torque commands that get
implemented through the propulsive RCS system. This
paper represents an update to work presented in [1].
The MSL Entry Capsule is about 5 m in diameter and 3
meters in height. Figure 1 shows the elements used for
attitude control. The propulsive RCS system is
composed of 4 thruster pods with two thrusters each.
The cruise balance masses (2), and the entry balance
masses (6). The cruise balance masses are ejected after
the entry capsule has been despun. They serve the
purpose to generate a cg offset such that the capsule
will have an L/D of .2 at mach 24. The entry balance
masses are ejected right before the opening of the
parachute to eliminate the cg offset. Six masses ejected
over a period of few seconds enable a slow correction
of the trim angle of attack.
Fig. 1. The MSL Entry Capsule diagram indicating the
RCS thruster’s directions, the cruise balance masses
(2), and the entry balance masses (6). Not to scale.
The exo-atmospheric phase starts after separation from
the Cruise stage and it last until the capsule enters the
Martian atmosphere. During this phase the attitude
controller will de-spin the capsule, turn to the predicted
desired entry attitude, and hold that attitude. This phase
lasts about 9 minutes. The exo-atmospheric phase
control function diagram is shown in figure 2.
Fig. 2. Exo-atmospheric phase functional diagram.
The Attitude Commander block generates the desired
attitude in two steps: A pitch rotation by the current
predicted trim angle of attack; and a roll rotation by the
desired initial bank angle, denoted as Pre-bank. Then
3-Axis Attitude Profiler generates the profile to take
the entry capsule to that state. In doing so, it first nulls
the angular rates and then it profiles a turn to the
desired attitude. The Attitude Controller takes state
information from the onboard Navigation Filter and the
desired attitude from the 3-axis attitude profiler to
generate the control errors. Then, it calculates the
desired torques to zero the control errors. The Attitude
Controller is a gain scheduled controller. The
parameters are read from a parameter table which is
indexed by the estimated atmospheric relative speed.
The attitude controller will be further described in a
later section. The Thruster Logic block provides a
pulse-width-modulation implementation of the attitude
controller desired torques.
The guided entry phase starts when the sensed
acceleration reaches a given threshold. It finishes at
parachute deploy at about mach 2. This phase lasts 2-3
minutes. The guided entry phase control function
diagram is shown in figure 3.
Fig. 3. Guided entry phase functional diagram.
During the entry phase, the Entry Guidance algorithm,
which is derived from the Apollo command module
final phase guidance algorithm and adapted to Mars
entry, generates bank angle commands to control
range-to-go and cross-range errors by adjusting the
drag acceleration. [3]. When the cross-range errors
exceed a given threshold the Entry Guidance algorithm
will command a bank reversal. These reversals are
large turns and therefore are profiled by the Bank
Profiler, which plans a single axis accelerate-coastdecelerate
attitude maneuver.
The entry capsule approximates a biconic vehicle.
During the entry phase a CG-offset is used to create a
lift to drag (L/D) ratio of approximately 0.24, which
leads to a trim angle of attack of about -15.5 degrees at
entry interface and varies slightly over time. The Entry
Controller calculations are performed in a Control
Frame, which is defined as a non-orthogonal frame
depicted in Figure 4. The yaw (x-axis) and pitch (yaxis)
correspond to eigenvectors of the aerodynamic
oscillatory modes. The roll (z-axis) corresponds to the
bank angle. This definition allows deadband settings
around the dynamic variables of interest. Yaw and
pitch deadbands are sized for rate damping of
oscillatory modes. Bank deadband is sized for guidance
Fig. 4. Control Frame
This Control Frame is time varying since the vehicle’s
trim angle of attack varies over the trajectory. The
predicted trim sideslip angle is nearly zero. These
angles are derived from the MSL aerodynamic
database developed by NASA’s Langley Research
Center [9]. Uncertainty in the predicted values needs to
be accounted in choosing the attitude deadbands. The
Viking project observed a 2 degree discrepancy
between the predicted and actual angles of attack
values [6] for both flights. In addition, the
aerodynamics oscillatory mode frequency (short
period, dutch roll) changes over the trajectory.
The separation of the lateral and longitudinal
aerodynamics, in conjunction with formulating the
control problem in the Control Frame, enables the
parameterization of the Entry Controller as three
independent channels.
The structure of the controller for each channel has a
feedback and a feedforward path. The feedforward path
is used to achieve fast profiled maneuvers for large
turns. The feedback path is used to stabilize the plant.
The feedback path is formulated as a phase plane
controller with attitude and rate deadbands to minimize
fuel usage. It is shown in figure 7. The attitude and rate
errors are computed relative to the deadbands. Crossing
the deadband engages the feedback gradually and
therefore it is expected that the errors will surpass the
deadbands. When the errors cross the attitude
deadbands, a PD controller (dashed blue zone) is
engaged. If the error crosses the rate deadbands, a D
controller (solid green) is engaged. These enable to
shape the frequency response of the controller. If the
errors were to grow further, more the PD and D
controllers will saturate and the behavior will be
equivalent to a bang-bang controller (light shaded
Fig. 5. Feedback path controller
The controller gains and deadbands are tabulated for
different flight regimes and different events. During the
exo-atmospheric phase, the attitude deadbands are set
large in order to minimize fuel during limit cycling.
Just before entry, the attitude deadbands are tightened
up to reduce disturbance effects. During the entry
phase, the attitude deadbands are set large and the rate
deadbands small. This enables the feedback to behave
primarily as a rate control. It provides both energy
damping and robustness against knowledge errors in
the predicted trim angle of attack. Both of the Viking
vehicles trimmed at about 2 degrees higher negative
angles than predicted [6]. Error in the predicted trim
angle of attack is an important factor in the selection of
the attitude deadbands when trying to minimize the
fuel usage. Attitude deadbands need to be large enough
to accommodate for trim prediction errors. Too small
attitude deadbands in face of prediction errors will
cause the controller to fight the actual trim angle of
attack and result in large fuel consumption. For the
same reason, the Viking landers flew a rate damping
controller during the entry phase [6]. For MSL, the
attitude deadbands are tightened up again before
parachute deploy to reduce the initial attitude error and
the subsequent dynamic disturbance during parachute
In addition, a modification was introduced in the bank
channel controller to react to large errors. It was
observed that large persistent roll disturbances during
bank reversals could cause large bank errors, which in
turn could cause the guidance law to fall behind. It was
recognized that in such cases, the controller could
maximize its roll torque capability to fight the
disturbance by giving up coordinated turns and
commanding a pure roll torque.
This section shows a detail simulation example. The
simulation was executed in the Control Analysis
Simulation Testbed (CAST) which is a JPL developed
computer simulation testbed for EDL. This testbed has
models for the environment (gravity, aerodynamics,
etc.) and for the spacecraft dynamics, sensors (IMU)
and actuators (RCS). In addition, it calls the Guidance,
Navigation, and Control (GN&C) algorithms as
implemented in the actual flight software. First, we will
describe the behavior during the exo-atmospheric
phase. Section 6.1 describes the spin-down, turn-toentry,
and the exo-atmospheric attitude hold periods.
Then, we will show the atmospheric phase in section
6.1 Exoatmospheric Phase
Before any GN&C activities start, a train of RCS
thruster firing is fired to warm up the RCS thrusters.
Since the MSL cruise stage is a spinner. Once the entry
capsule separates from the cruise stage, the entry
capsule will remain spining at about 12 deg/s. The first
activity is to spindown the capsule and then turn to the
desired entry attitude. The desired entry attitude
consists of the predicted trim angle of attack and
sideslip angle at contact with the Mars atmosphere, and
a predicted initial bank angle derived to be consistent
with the reference trajectory implemented in the
Guidance algorithm.
Figure 6 shows the GNC mode versus time. This mode
provides information on the on-going GN&C activities.
The simulation was set to start at time equal to 1000 s.
Before any GN&C activities start, a train of RCS
thruster firing is fired to warm up the RCS thrusters (~
20 s). Then, we spin-down (~ 6 s), and perform the
turn-to-entry attitude maneuver (~ 14 s). Then we enter
into the exo-atmospheric attitude hold phase (~ 500 s).
As mentioned earlier, large turns are profiled using the
feedforward path. So, during the spindown and turn to
entry will be profiled by the three axis attitude profiler.
The attitude profiler will generate a profile angular
accelerations. Then the attitude controller will use that
desired angular accelerations to calculate the desired
feedforward torque. In addition, the feedback path of
the controller will evaluate the attitude and rate errors
against their deadbands. It will generate a feedback
desired torque commands to keep these errors with the
deadbands (~ 2 deg). Then, the RCS thruster logic
implements the desired torques through pulse width
modulation of the 8 RCS thrusters. Figure 7 shows the
capsule attitude rates. Figure 8 and 9 show the attitude
and attitude rate errors respectively. Figure 10 shows
the total commanded torque and the feedforward
torque. Figure 11 shows the commanded on-time
durations to the RCS thrusters to implement the desired
torque. Figure 12 shows the cumulative fuel used
during this phase.
Figure 6. GNC Mode
Figure 7. Capsule Rates.
Figure 8. Attitude Errors
Figure 9. Attitude Rate Errors
Figure 10. Commanded (-) and Feedforward (.)
Figure 11. Commanded RCS Thruster durations.
Figure 12. Fuel Used
6.2 Atmospheric Phase
The activities happening during the atmospheric phase
are summarized in the EDL GNC mode, Figure 13. It
starts at the Wait for Guidance start mode, where it
waits until the sensed drag acceleration reaches a given
threshold. At this point the Entry Guidance Range
Control (RC) algorithm was called for the first time
and generated its first bank angle command. The bank
profiler will then profile a turn to take the bank angle
from the pre-bank used during the exo-atmospheric
phase to the new bank command, denoted as RC Slew
to Cmd Bank mode. Then, it goes into RC Track, where
the guidance algorithm tracks the downrange errors.
When the cross-range errors get too large, it commands
a bank reversal and the mode toggles to RC Bank
Reversal. When the heading alignment (HA) phase
starts, the bank profiler plans a turn to the first heading
alignment bank angle, as HA Slew to Cmd Bank, and
goes into HA Track. Before the chute opens a roll slew
is performed to point the radar antennas to the ground,
denoted as SS Slew to Radar Att. (where SS stands for
Straighten-up-fly-right Slew). At the same time the
CG-offset is removed by ejecting 6 balance masses.
Once the slew is complete, it goes into SS Wait for
Chute, where it waits for the right conditions to open
the parachute. When these conditions are met the entry
attitude controller is disabled and the GNC mode goes
into Settle Chute Transients. Figure 14 shows the bank
angle time histories. It illustrates the bank profiles
during the slews and bank reversals. Figure 15 shows
the predicted and actual angles of attack. It shows a
small mismatch between the predicted and the actual.
Figure 16 shows the sideslip angle. Both angle and
attack and sideslip angle shows aerodynamic induced
oscillations. Figure 17 shows the actual spacecraft
rates. Figure 18 and 19 show the attitude and rate
errors. Figure 20 and 21 shows the attitude control
desired torques and the commanded RCS thruster
durations. Figure 22 shows the cumulative fuel
consumption since the start of the exoatmospheric
Figure 13. GNC Mode
Figure 14. Bank angle.
Figure 15. Angle of Attack
Figure 16. Sideslip angle
Figure 17. Capsule Rates
Figure 18. Attitude Errors
Figure 19. Attitude Rate Errors
Figure 20. Commanded (-) and Feedforward (.)
Figure 21. Commanded RCS Thruster durations.
Figure 22. Fuel Used
This paper presented the concept for the Attitude
Controller for the Exoatmospheric and Atmospheric
Entry phase of the Mars Science Laboratory Entry,
Descen and Landing. The controller is parameterized
as 3 independent channels around the predicted trim
position. Each channel controller is composed of a
feedfoward and a feedback path. The feedforward path
enables fast response to large bank commands. The
feedback path stabilizes the plant around attitude and
rate deadbands while minimizing fuel usage.
Feasibility and the satisfactory performance of this
design have been demonstrated by computer
The work was performed at the Jet Propulsion
Laboratory, California Institute of Technology, under
contract with the National Aeronautics and Space
© 2010 California Institute of Technology.
Government sponsorship acknowledged.
1. Brugarolas P.B., San Martin A.M., Wong E.C.,
“Mars Science Laboratory Entry Controller” AIAA
GN&C Conference, Aug. 2008.
2. Mendeck G. F. and Carman G. L.,”Guidance
Design for Mars Smart Landers Using The Entry
Terminal Point Controller” AIAA 2002-4502, August
3. Ingoldby R. N., “Guidance and Control System of
the Viking Planetary Lander”, Journal of Guidance and
Control, Vol. 1, No. 3, 1978, pp. 189-196.
4. Holmberg N. A., Faust R. P., and Holt H. M.
“Viking ’75 Spacecraft Design and Test Summary”
NASA Ref. Publ. 1027. November 1980.
5. “Viking Lander System Primary Mission
Performance Report (Martin Marietta Corp.)”, NASA
CR-145148, April 1977.
6. Thurman S.W. and Flashner H., “Robust Digital
Autopilot Design for Spacecraft Equipped with PulseOperated
Thrusters”. J. Guidance, Control and
Dynamics, Vol. 19, No. 5, 1996, pp. 1047-1055.
7. Calhoun P. C. and Queen E. M., “Entry Vehicle
Control System Design for the Mars Science
Laboratory” Journal of Spacecraft and Rockets 2006
0022-4650 vol.43 no.2 (324-329)
8. Schoenenberger M., Cheatwood F., and Desai P.,
“Static Aerodynamics of the Mars Exploration Rover
Entry Capsule” AIAA 2005-0056, January 2005.
9. Schoenenberger M., and Queen E.M., “Oscillation
Amplitude Growth for a Decelerating Object with
Constant Pitch Damping” AIAA 2006-6148, August