- ___= AIAA 2002-4412 MARS SMART LANDER SIMULATIONS FOR ENTRY, DESCENT, AND LANDING S. A. Striepe,* D. W. Way,* A. M. Dwyer,* and J. Balaramt Abstract Two primary simulations have been developed and are being updated for the Mars Smart Lander Entry, Descent, and Landing (EDL). The high fidelity engineering end-to-end EDL simu- lation that is based on NASA Langley's Program to Optimize Simulated Trajectories (POST) and the end-to-end real-time, hardware-in-the-loop simulation testbed, which is based on NASA JPL's Dynamics Simulator for Entry, Descent and Surface landing (DSENDS). This paper presents the status of these Mars Smart Lander EDL end-to-end simulations at this time. Various models, ca- pabilities, as well as validation and verification for these simulations are discussed. NOMENCLATURE 2009 launch, this mission uses an onboard guidance during entry and terminal descent phases to direct the AGL Above Ground Level spacecraft to the desired landing site, while avoiding CFD Computational Fluid Dynamics any surface hazards detected by the onboard systems. DEM Digital Elevation Map The guidance controls the vehicle through lift vector DOF Degree of Freedom modulation during the entry phase and thrust vector DSENDS Dynamics Simulator for Entry, pointing during the terminal descent. Figure 1shows an Descent and Surface landing illustration of the Entry, Descent, and Landing (EDL) for the Mars Smart Lander. This mission will demon- EDL Entry, Descent, and Landing strate some of the precursor steps required for a Mars ETPC Entry Terminal Point Controller GCM Global Circulation Model sample return mission. IMU Inertial Measurement Unit JPL Jet Propulsion Laboratory Current approaches to EDL at Mars have focused Mars-GRAM Mars Global Reference Atmosphere on unguided and uncontrolled ballistic entry with no Model precision landing or hazard avoidance capabilities. The MOLA Mars Observer Laser Altimetry new generation of "Smart" Landers with their lifting MSL Mars Smart Lander body designs, aerodynamic steering, active terrain POST Program to Optimize Simulated sensing, and powered-descent/precision-landing capa- Trajectories bilities requires a new generation of simulation tech- nology to support their development. The EDL systems INTRODUCTION are required to successfully deliver the surface systems through the entry, descent, and landing phases, which The Mars Smart Lander (MSL) mission will de- begin at lander separation from the cruise stage and end at touchdown on the surface. For MSL, these EDL sys- liver an advanced rover to a specified Martian site with an accuracy never before achieved by using a guided tems will be designed, developed, tested, and evaluated aeromaneuvering lander. [Ref. 1] Currently slated for a using end-to-end high fidelity computer flight simula- tions developed by a team of engineers from NASA- JPL, NASA-JSC, NASA-Langley, and the University *Aerospace Engineer, Vehicle Analysis Branch, of Texas at Austin. NASA Langley Research Center. tPrincipal Member, Avionics Systems Division, NASA Jet Propulsion Laboratory. Two primary EDL simulations have been devel- oped from existing tools and are being updated specifi- Copyright © 2002 by the American Institute of Aeronautics and Astronautics, Inc. No copyright is cally for the MSL project. The first of these simulations asserted in the United States under Title 17,U.S. code. is the high fidelity engineering end-to-end EDL simu- The U.S. Government has a royalty-free license to lation, which is based on NASA-Langley's Program to exercise all rights under the copyright clamed herein for Optimize Simulated Trajectories (POST) [Refs. 2 and Governmental Purposes. All other rights are reserved 3]. The second program is the end-to-end real-time, by the copyright owner. 1 American Institute of Aeronautics and Astronautics hardware-in-the-sloimopulatiotnestbedw,hichisbased to support the objective above is a real-time, hardware- onNASA-JPL'DsynamicSsimulatofrorEntry,De- in-the-loop end-to-end EDL simulation to allow sys- scenatndSurfacleanding(DSENDS[R) ef4. ].Models tems level tests of flight hardware and software compo- ofvariousMSLEDLsystemasreprovidebdygroups nents, flight software checkout prior to upload, and responsibfloertheirtechnologoyrflightdevelopment.systems level troubleshooting during operations. Over- Initially,thesemodelsareincorporateindtothehigh- all Lander system risk is reduced by fully integrating fidelityengineering(POST-basedsi)mulationfor not only detailed engineering models of the subsystems, evaluatioannddevelopmeEnvt.entualltyh,esemodels, but the actual hardware for testing in the simulated mis- aswellasflight-readaylgorithmasndhardwarwe,illbe sion environment from cruise stage separation through incorporatiendtoareal-timseimulatiocnapabloefsup- lander touchdown. These end-to-end system level portingdetaileddynamicsd,evicemodelsandhard- simulations also provide inputs as well as independent ware-in-the-loospimulations(Mars-DSENDSfo)r validation and verification of subphase simulations used flightoperationevaluationte,stinga,ndrefinement. for subsystem design, algorithm development, and de- DuringtheMarsSmarLtandemr issionb,othsimula- tailed component level analyses. Additionally, the tionswillsupporfltightoperationasndanomalryesolu- simulations receive input conditions from interplanetary tion.Thispapedrescribethsesetwoend-to-ensdimula- trajectory and cruise stage system simulations while tionsaswellasthemodelsa,lgorithmas,ndhardware providing inputs to Rover egress simulations. These theyincorporattoetestandevaluatteheLandedruring simulations enable designers and mission managers to simulateMdarsentries. evaluate specific and overall lander systems perform- ance in the expected Mars environment. These objec- BACKGROUND tives and simulation requirements are met for MSL by the high-fidelity engineering (POST-based) and real- Traditionally, disparate simulation tools have been time, hardware-in-the-loop testbed (DSENDS-based) exercised in an ad-hoc manner on the separate portions end-to-end EDL simulations. of a typical EDL profile. This approach has resulted in The high-fidelity engineering (POST-based) end- a patchwork of tools, with complex, hand-crafted inter- faces requiring manual transfer of data products across to-end EDL simulation provides mission level feasibil- simulation systems. Such an approach was difficult ity, design trades, guidance and control algorithm de- even for the relatively simple requirements of the Mars velopment and analysis capability, as well as success Pathfinder style missions. This approach will be totally criteria evaluation of flight systems using rapid Monte Carlo analyses. The POST tool developed at NASA inadequate for the Smart Lander missions where exten- sive, closed-loop actions by the spacecraft requires in- Langley has been successfully used for mission design, timate integration of all the supporting simulation ele- operations, trajectory determination/optimization and ments. Monte-Carlo analyses in many recent Mars missions. These missions include the Mars Pathfinder, Polar The objective of the MSL EDL simulations is to Lander, Odyssey Orbiter, proposed 2001 Lander, and MER 2003 Lander missions. POST has also been used support Lander systems design, trade studies, develop- for a number of non-Mars missions involving atmos- ment, testing, and operations by establishing end-to-end simulations that include high fidelity models of the pheric entry (e.g., Stardust, Genesis, etc.). [Refs. 5-19] Mars Smart Lander systems and the Mars environment. In turn, the real-time, hardware-in-the-loop To support this objective, a capability to provide end- to-end engineering simulation of all phases of EDL (DSENDS-based) end-to-end EDL simulation allows flight throughout the MSL project lifecycle is required. detailed system integration analyses and testing for the Mars environment. The DSENDS (Dynamics Simulator This engineering simulation is needed to enable rapid initial screening to define critical mission and lander for Entry Descent and Surface landing) real-time simu- lation tool is an extension of the Darts/Dshell multi- parameters and sensitivities for conceptual design and mission spacecraft simulation tool developed at NASA- system level trades. Not only is this high fidelity simu- JPL. This tool has been used on the Cassini mission, lation necessary for detailed lander design, mission statistics, and operations support, but also to verify in- which will study Saturn and send an entry probe into Titan's atmosphere. DSENDS is capable of providing tegration of EDL subphases performance (such as the control system during entry or the parachute during detailed spacecraft system/device simulation (sensor, actuators, communication devices, flex/rigid multi-body descent) through evaluation of lander systems and con- dynamics, aerodynamics, etc). Additionally, flight soft- figurations in an end-to-end environment. Also needed 2 American Institute of Aeronautics and Astronautics ware can be embedded in the simulation creating a opment, testing, and operations of vehicles for particu- system integration testbed to support flight system vali- lar missions. Simulation complexity varies from first- dation, ATLO, and mission operations. order trades (e.g. parachute size and deployment condi- The current MSL EDL simulation strategy has sev- tions, terminal descent engine size, etc.) to all-up eral strong advantages. The use of POST-based and Monte-Carlo simulations to day-of-entry operations. DSENDS-based end-to-end simulations in conjunction with independent subphase simulations enables each The models required for these simulations depend critical segment of EDL to be covered by three simula- on the desired fidelity of the analysis. In the initial tions for synergistic purposes. The independent devel- phases of mission definition and vehicle conceptual opment of these simulations provides a strong basis for design, basic models already available in POST are used without modification to provide a tool for top level independent validation and verification of all simula- tions. End-to-end simulations provide coordination with trades and conceptual level design. By using existing specialized subphase simulations leading to early iden- models, rapid turnaround vehicle assessment and design tification of interface issues as well as subsystem con- simulations are possible. These engineering models can flicts from one phase of EDL to the next. Figure 2 il- be rapidly adapted for performance evaluation and top- lustrates the coverage that the two end-to-end simula- level trades of new designs. As the mission and systems tions provide and their interaction with more specific get better defined and higher fidelity models become subphase simulations. Although these end-to-end available, they are incorporated into the POST simula- simulations are undergoing further improvements, the tion to perform more mission specific trades and analy- design engineers and project managers are receiving ses of the updated systems. Eventually, three and six useful simulation results even as the simulations are degree-of-freedom (3- and 6-DOF) simulations which being updated and their fidelity is increasing. By pro- span an entire phase of a mission (such as entry, de- viding these results in the early phases, the project scent and landing at Mars from the final exoatmos- schedule is less likely to be impacted by system level pheric trajectory correction maneuver to lander touch- EDL design issues late in the design cycle. down) using the latest engineering models of onboard systems are available for detailed mission trades, sys- The following two sections describe the POST- tem analyses, system testing, and mission operations. based and DSENDS-based simulations for Mars Smart This approach has been and is being applied to the Mars Lander Entry, Descent, and Landing in more detail. The Smart Lander mission for the Entry, Descent, and third section includes information about the validation Landing high fidelity engineering simulation using and verification approach for these simulations. POST as the main simulation engine. PART 1.HIGH FIDELITY ENGINEERING The POST-based simulation tools have been used {POST-BASED) END-TO-END EDL to support all elements of the design life cycle for a SIMULATION wide variety of missions. Early conceptual studies have been conducted using models in the basic production The Program to Optimize Simulated Trajectories version of POST. [Refs. 5-9] Higher fidelity simula- tions have included many mission specific models and was initially developed in the 1970's to support Space Shuttle development. [Ref. 2] It has been continually data including aerodynamic parameters from wind tun- upgraded and modified since then to support a large nel testing and Computational Fluid Dynamics (CFD) runs, vehicle mass properties, parachute, control sys- variety of aerospace vehicle development and opera- tems, and onboard propulsion systems as these data and tions through trajectory simulation, analyses, and sys- models became available. [Refs. 10-15] POST-based tem performance assessments. [Ref. 3] POST contains simulations have been exercised for extensive Monte- many basic models (such as atmosphere, gravity, and Carlo runs including those for "stress tests" that deter- navigation system models) that are used to simulate a mine the limits of system capability. [Refs. 13-18] The wide variety of launch, orbital, and entry missions (see technical capabilities of POST have already been vali- Fig. 3). dated against other Mars mission data. [Refs. 16-19] Note that these references focus mainly on Mars mis- However, exploiting the modular nature of the sions, whereas a much larger set of references exists for POST program by adding mission specific models in concert with the existing POST architecture allows for other applications in which POST is used, such as Earth the development of higher fidelity, mission specific launch and entry vehicle development as well as entry simulations. These simulations support design, devel- systems for other planetary missions. 3 American Institute of Aeronautics and Astronautics TheMarsSmarLtandemr issionisillustratedin inertial measurement unit (IMU), vehicle aerodynamics Fig.1.ThecurrenEt DLtimeline is shown in Table 1. (entry, parachute, and terminal descent configurations), As noted in the timeline, two configurations are cur- parachute dynamics, as well as various control system rently included in the POST-based MSL high fidelity and propulsion system models. The environment mod- end-to-end EDL engineering simulation. The simulation els in the simulation include high fidelity gravity mod- starts at cruise stage separation, whereas the actual en- els, Mars-GRAM atmosphere models, and surface to- try begins at atmospheric interface when aerodynamic pology based on MOLA data. Table 2 shows the re- forces (albeit small at altitudes over 60 km) are acting sponsible groups for various vehicle specific models. on the vehicle. During atmospheric entry, the flight path Several of these models are briefly described below, is controlled by the entry guidance until it commands whereas references are provided for details of other models. supersonic parachute deploy (nominally around 9 km above the ground). Next, the backshell and supersonic The POST-based simulation has been used to pro- parachute separate from the lander and deploy the sub- vide a variety of products to the Mars Smart Lander sonic parachute around Mach 0.8. Ten seconds later, the heat shield is jettisoned. The radar begins to get project. These products include: entry and terminal de- usable altitude and velocity data about three seconds scent vehicle designs and trades; entry aeroshell con- later when the heat shield is clear. Upon command of figuration trades; guidance and control algorithm de- the terminal descent guidance, the main engines are velopment; parachute sizing trades; terminal descent started at 20 percent thrust. After two seconds, the sub- engine trades; project Monte Carlo statistics (precision sonic parachute is released and the preliminary touch- landing, touch down velocity, etc.). Current plans are to down target is identified. When hazard avoidance is continue to supply mission and vehicle trade studies, as included, the system will use LIDAR to scan the sur- well as provide inputs to the design process (e.g., heat- face to determine if the preliminary target is suitable. If shield thermal protection system sizing). The POST- based simulation will also be used in the operations not, a new target will be established and the lander will be diverted to it. In the simulation, a single divert ma- phase to support day-of-entry preparation and analyses. neuver of 100 m (horizontal distance) at 300 m above The POST-based simulation is also providing validation ground is used in the Monte Carlo analyses to simulate and verification support to the real-time, hardware-in- the divert capability. During terminal descent, guidance the-loop (DSENDS-based) simulation; further discus- is commanding the thrust vector (magnitude and direc- sion on this support is provided in the third part of this tion) to reach the target point such that a constant ve- paper (POST-DSENDS Validation and Verification). locity, vertical motion only phase is started at 5 m above the ground. The radar stops returning useful data Gravity Model at about 10 m above the surface. All engines are shut off when the lander is one meter above the surface. To The gravity model in the POST-based simulation assess the current configurations under consideration, uses zonal, sectoral, and tesseral harmonic terms to an EDL Challenge Site has been identified (which is determine the acceleration due to gravity. This model is about 2500 m above the MOLA areoid) as the nearly 20 based on the one used in the Artificial Satellite Analysis km wide crater at 41.45 ° S latitude and 286.5 ° E longi- Program (ASAP). [Ref. 20] The 50-by-50 Mars gravity field used in the simulation is the MARS50C model tude. The entry and terminal descent guidances are con- established to support the Mars Polar Lander mission. figured to land inside the Challenge crater. [Ref. 21] This gravity model will be updated as higher The MSL high fidelity engineering simulation of order data (such as the 75 x 75 gravity model used to these EDL events currently includes various engineer- support the Mars Odyssey mission) becomes available. ing models at varied levels of complexity. Both 3 DOF and 6DOF simulations, which start at lander separation Planet Model from the cruise stage and finish at touchdown on the surface, are under development, with the 3DOF simu- An oblate spheroid Mars model is also used in the lation the most mature at this point. As mentioned POST-based simulation. This planet model defines the above, the latest engineering models are incorporated physical dimensions (e.g., equatorial radius, polar ra- into these simulations, or existing models in POST are dius) and characteristics (e.g., rotation rate) of Mars. used until the engineering model is developed. The ve- This model is not only used for latitude and longitude hicle specific models include terminal descent and entry determinations, but is also necessary to determine Mars guidance algorithms, flight navigation filter, sensors, 4 American Institute of Aeronautics and Astronautics relativevelocityusedbytheguidancaelgorithmasnd /3= center-spacecraft-subsurface angle (angle other simulation models. between gcrad and hgtagl) The local altitude is determined using Mars Ob- 0= center-surface-spacecraft angle (angle server Laser Altimetry (MOLA) topography data and a between rcalc and hgtagl) reference areoid model. The recent availability of elec- tronic topographic data provided by the MOLA project This calculated radius is compared to the sum of [Ref. 22] has enhanced POST simulations by allowing the radius to the areoid plus the topographic altitude, the calculation of vehicle height above local features at determined from the MOLA dataset at the geocentric Mars. There are three primary surface references for latitude equal to the guessed declination. A bisection measuring altitude: the areoid, the reference ellipsoid, root finding method is used to drive the error between and the surface. The areoid (or Mars geoid) is a gravi- the calculated and actual radii to zero. Once the surface tational equipotentiai surface, analogous to the theoreti- declination is known, the altitude of the spacecraft, cal mean sea-level surface on Earth. A "plum bob" measured geodetically, is calculated from the law of would hang perpendicular to its surface at every point. cosines The angle between this vertical and the equator defines the astronomical latitude. The MOLA areoid is defined hgtagl 2 = gcrad 2+ rcalc 2-2gcrad" rcalc cos(a) to be a surface with the same gravitational potential as the mean equatorial radius (3396 kin) and is determined from an 80 by 80 coefficient representation of the where gravitational field. The geocentric radius to the areoid is provided as a dataset at 1/16 degree resolution. The hgtagl = height above ground level (AGL) reference ellipsoid, used within MarsGRAM and POST, is an engineering approximation of the areoid. The sur- gcrad = geocentric radius to spacecraft face of the ellipsoid is completely defined by an equato- rial radius of 3396 km and a polar radius of 3378.32 rcalc = calculated radius to surface km. The normal vector to the ellipsoid is the direction that a plumb bob would hang if it where not for local a =spacecraft-center-surface angle (angle be- gravitational anomalies. The angle between this normal tween gcrad and rcalc) vector and the equator defines the geodetic latitude, which is the basis for most maps and charts. The Atmosphere Model MOLA dataset provides a planet-wide model of the Mars surface topography at 1/32 degree resolution, ex- The Mars Global Reference Atmosphere Model pressed relative to the areoid. [Ref. 23] version 2001 (Mars-GRAM 2001) has been included in the simulations (as FORTRAN subroutines The problem of determining the vehicle's altitude in POST). Mars-GRAM provides all of the atmospheric above the surface, in the geodetic sense, requires an data (temperature, density, pressure, and wind velocity) iterative solution. A declination to the surface (the an- as well as random perturbations to certain atmospheric gle between the radius vector and equatorial plane, quantities (e.g., density) while including seasonal, diur- declnstar) is first guessed which defines a point be- nal, latitudinal, and longitudinal variations. The atmos- neath the spacecraft, measured in the geodetic sense. pheric data is a function of the spacecraft location (See Fig. 4.) The length of the vector measured geo- (latitude, longitude, and altitude) as well as other user centrically to this point from the center of the planet is supplied inputs. [Ref. 24] These inputs include the date calculated geometrically using the law of sines: of Mars arrival, the minimum update distance for dis- persion calculations, a scale factor on the atmospheric rcalc =gcrad sin(/3) dispersions, interpolation option for the upper atmos- sin(0) phere, and the fl0.7-cm solar flux value. In addition, the capability to model the effect of dust opacity or dust where storms is included. The Mars arrival date and fl0.7-cm solar flux values reflect the period during the solar cy- rcalc = calculated radius to surface cle in which the entry occurs. gcrad = geocentric radius to spacecraft 5 American Institute of Aeronautics and Astronautics The Mars-GRAM subroutine is a parameterization continuous interpolations between a database of dis- of the atmospheric properties, so that the model runs crete solutions. This interpolation scheme is applied to free molecular solutions for the rarefied region of the relatively quickly and the overall simulation speed is not hampered by the atmospheric subroutine. Recent atmosphere, and computational fluid dynamic (CFD) versions of Mars-GRAM include density profile data solutions combined with wind tunnel test results for the from more detailed simulations using global circulation continuum regime. A modified Lockheed bridging models (GCM) being developed at the NASA Ames function [Ref. 26] is used in the transitional region be- Research Center (by Robert Haberle and James Mur- tween rarefied and continuum regimes. The various phy) and at the University of Arizona (by Steve flow regimes are delineated according to Knudsen number. Bougher); that is, Mars-GRAM can reproduce the more realistic densities from the GCM for a specific entry profile in the simulation but in a fraction of the time. Entry capsules for robotic missions tend to spend a These recent versions also include atmospheric wave significant amount of time in rarefied and transitional models which incorporate wave effects on atmospheric flow regimes. Therefore, free molecular values are in- density. The latest Mars-GRAM version includes the cluded in the aerodynamic databases. The aerodynamic 1/2 degree resolution topographic data for Mars from data in the rarified regime are a function of vehicle at- the Mars Observer Laser Altimetry (MOLA) instrument titude. In the transitional regime, the aerodynamic data onboard the Mars Global Surveyor spacecraft in orbit are a function of both vehicle attitude and Knudsen about Mars. The dust opacity parameter is used to de- number. fine the amount of airborne dust particles so that Mars- GRAM can simulate their affect on the atmospheric For the continuum region, static aerodynamic data were obtained from CFD solutions using the Langley properties. Aerothermodynamic Upwind Relaxation Algorithm Two significant adjustments have been made to (LAURA) [Ref. 27-29] and tests conducted in the support Mars-GRAM inclusion in the high fidelity en- NASA Langley's Unitary-Plan Wind Tunnel (UPWT) gineering MSL EDL end-to-end simulation. First, a [Ref. 30]. LAURA was used to generate aerodynamic databases for the Mars Pathfinder [Ref. 31], Mars Mi- wrapper subroutine was developed to provide a soft- ware interface between the Mars-GRAM program, de- croprobe [Ref. 26], and Stardust [Ref. 32] entry cap- sules. Confidence in the LAURA solutions comes from veloped by Jere Justus (through the NASA Marshall validations with Viking data, wind tunnel data, and Space Flight Center) and the POST-based simulation. The wrapper converts between the double precision Mars Pathfinder mission results. [Ref. 33] Dynamic variables used in the flight simulation and the single aerodynamic quantities were included from the data precision variables used by Mars-GRAM, and it pro- generated for the Viking missions. hibits Mars-GRAM from being called too frequently while dispersed density atmospheres are generated Parachute aerodynamic data is taken from Viking during Monte Carlo analyses. Second, a higher resolu- and Mars Excursion Rover (MER) mission data. Super- tion MOLA topography data (1/16 degree resolution) sonic parachute data is taken from existing disk-gap- was added to the Mars-GRAM 2001 software. Jere band parachutes from the Mars Pathfinder mission [Ref. Justus suggested the necessary subroutine adjustments 34] and planned for the MER mission in 2003. The sub- that were implemented to include this newer data. The sonic parachute is a ringsail parachute of the type used 1/16 degree resolution MOLA data includes most of the in Apollo, Gemini, and Mercury [Ref. 35] with para- surface features (e.g., craters) found in higher resolution chute area scaled for the mass of the MSL entry system. data (such as 1/32 degree resolution), while requiring a Terminal descent phase aerodynamics were taken manageable amount of computer memory. from Viking mission data. Prior to heatshield separa- Aerodynamic Model tion, the entry phase aerodynamics is used, After heat- shield separation, the data from wind tunnel tests con- A FORTRAN subroutine supplies aerodynamics ducted on a Viking lander inside a backshell was incor- data to the POST-based simulation. [Ref. 25] Whereas porated into the aforementioned aerodynamics subrou- different subroutines are supplied depending on the tine format. [Ref. 36] configuration to be simulated, the basic difference be- tween routines is the data included for the specific con- figuration. The routine uses first derivative, or C(1), 6 American Institute of Aeronautics and Astronautics and moments of inertia. In the second mode, the con- Control System troller first determines the necessary angular accelera- Reference 37 describes the 6 DOF entry control tion and then solves for the actual terminal descent en- system under development, whereas reference 38 dis- gine throttle settings that would provide the correct cusses the 6 DOF terminal descent controller being angular acceleration and commanded thrust, within designed. While higher fidelity 6-Degree-of-Freedom prescribed minimum and maximum throttle limits. (DOF) simulations with entry and terminal descent These throttle settings are used to determine the actual control systems are being developed, a lower fidelity 3- angular acceleration, which is then used to update the DOF simulation that executes faster while providing vehicle's attitude. This mode requires detailed knowl- similar results is desired. That is, a model that simulates edge of the magnitude and direction of thrust and mo- the behavior of a 6-DOF controller in a 3-DOF simula- ments provided by each engine that is being manipu- tion of the EDL phase of a planetary entry vehicle is lated by the controller. wanted. For the entry phase, a pseudo-controller of In either acceleration control or throttle control, the bank angle is employed. Whereas, the terminal descent controller in 3-DOF simulations controls attitude of all terminal descent controller must first determine the three axes. Both of these controller models were devel- desired angular velocity vector necessary to achieve the commanded attitude. The direction of the angular ac- oped and integrated within POST to respond to vehicle celeration vector is chosen such that the resultant an- guidance commands in a 3-DOF POST simulation. gular velocity vector lies along the single axis-of- During the entry phase, vehicle attitude in the 3- rotation between the current attitude and the com- DOF simulation is determined by balancing the aerody- manded attitude. The single axis-of-rotation is found namic moments acting on the vehicle (i.e., flying in an from the vector component of the quaternion that, when aerodynamic trim attitude) for angle of attack and multiplied by the current attitude quaternion, produces the commanded attitude quaternion. The magnitude of sideslip angle while a pseudo-controller is employed for the commanded bank angle. Work with previous 6DOF the acceleration vector is determined from the angular simulations has shown that aerodynamic trim condi- error between the commanded and current attitudes, a tions generally occur at about the mean attitude when controller gain, and the maximum allowable angular rotational dynamics are included. Therefore, results velocity. The strategy employed is to complete a certain from 3 DOF simulations using aerodynamic trim trans- percentage of the desired angular rotation, controlled by late better to the 6 DOF simulation than constant atti- the gain, within the current time step. However, if the tude 3 DOF runs. The single axis controller determines maximum angular velocity would be exceeded, the an- the appropriate bank angle and bank rate change for the gular rotation is limited to the product of the maximum input maximum acceleration and bank rate (5 deg/s 2 angular velocity and the time step. This acceleration and 20 deg/s for MSL) using an Euler integration vector is finally scaled such that the maximum compo- scheme. The maximum acceleration is assumed until nent of acceleration along each axis is not exceeded. either maximum rate is achieved or the controller de- termines that maximum deceleration must begin to Guidance Al2orithms reach the commanded bank angle. This pseudo- controller model was also used in the MSP'01 Lander The guidance algorithm for the entry phase (known simulation. [Ref. 14] as the Entry Terminal Point Controller, or ETPC) de- termines if modifications to the current atmospheric A terminal descent controller was developed and flight path are required and directs the control system to integrated within POST that models the 6-DOF rota- make attitude adjustments based on the navigation sys- tional dynamics of a vehicle in a 3-DOF simulation. tem input and the desired target location. As illustrated This terminal descent controller may operate in either in Fig. 5, this system modulates the vehicle bank angle of two modes: acceleration control or throttle control. (direction of the lift vector, q5)such that the vehicle ad- In the first mode, the controller solves for only the an- justs its atmospheric trajectory. In this manner, the ve- hicle can accommodate off-nominal entry-state or at- gular acceleration vector needed to obtain the com- manded attitude, within prescribed angular velocity and mospheric-flight conditions and achieve a significant acceleration limits. This acceleration vector isthen used reduction in landed footprint over non-lifting (ballistic) or constant bank angle (Viking-type) entries. Maximum to update the vehicle's attitude. This mode is advanta- control authority occurs when the vehicle is traveling at geous when only limited information is known about the vehicle's propulsion system, attitude control system, hypersonic speeds through the peak dynamic pressure 7 American Institute of Aeronautics and Astronautics (andpeakdeceleratiopno)rtionoftheatmospheerinc- Monte Carlo Dispersions try.TheETPCalgorithmisderivedfromthefinalphase logicoftheApollocommanmdoduleentryguidance. A Monte Carlo dispersion analysis is used to quan- Bankanglecommanfdosrterminaplointrangecontrol tify the acceptability and robustness of a given vehicle arederivedwithalinearperturbatioanlgorithmusing configuration, as well as determine areas of risk associ- influencceoefficientosfdragacceleratioanndaltitude ated with certain designs and mission phases. These rateerrorswithrespecttoafixednominarleference dispersion analyses are obtained by randomly varying trajectoraysafunctionofrelativevelocityC. rossrangekey parameters and characteristics of the environment controilsaccomplishwedithbankreversaalsttarget as well as the vehicle assuming a normal or uniform out-of-planceorridorlimits;howevear,finalheading distribution of these quantities. The engineers responsi- alignmenpthaseis usedtonullterminaclrossrangeble for the subsystem models identify the 3-sigma or errorsA. dditionallyth, eguidancienitiatesthesuper- maximum/minimum values of the uncertainties for sonicparachudteeploymetnotachievmeinimumtarget these key parameters. These inputs are then used in the rangewithinsupersonpicarachudteeployconstraints MSL end-to-end EDL engineering simulation to deter- (Machnumbearnddynamicpressurceonstraintasre mine various outputs of the trajectory. The outputs are implementeimdplicitlyasrelativevelocityanddrag compared with given metrics for each; thus, the suit- acceleratiocnorridors)F.urthedretaiol ftheETPCis ability of the vehicle and mission can be assessed. A giveninreferenc3e9. similar approach has been applied to the entry phase of several previous missions. [Refs. 5-8, 11, 14] Fortheterminadlescenpthaseth,eguidancaelgo- rithmisnotonlyusedtoensureasuccessftuoluch- Table 4 indicates the parameters currently varied in down,butalsoprovidesacapabilitytodivertaway the POST-based simulation during the Monte Carlo fromdetectehdazardTs.heguidancceommanadnsac- analyses. This table also shows the nominal value, type celeratiopnrofilebaseodnnavigatioenstimateosfpo- and limits of variation (either minimum/maximum or 3- sitionandvelocityT.hisdesireadcceleratioisnimple- sigma) for each. These quantities are varied randomly mentevdiaappropriattherottlesettingosnthesixmain over 2000 simulation runs. Various mission and vehicle terminadlescenetnginesT.hecontrolsystemisas- parameters are recorded at certain events throughout the sumetdoalignthethrustvectotrothecommandaecd- simulations. These quantities are evaluated relative to celeratiodnirectionviatheappropriavteehicleattitude. MSL project metrics to assess vehicle performance, Acommanddeidverot rchangienthedescepnrtofileto mission risk, and system robustness. Characteristics of avoidahazardisreflecteidntheacceleratiopnrofile Monte Carlo cases that consistently fail are identified commandbeydtheguidancIen.thefinalfivemetersa, for further investigation by system, vehicle, and mis- constanvtelocitydescenistcommandeudntiltheen- sion designers. During mission operations as day-of- ginesareshutoffatonemeterF. urthedretailonthe entry approaches and occurs, the POST-based Monte terminadlescegnut idancisegiveninreferenc3e8. Carlo capability can be used to rapidly assess many off- nominal conditions to identify several challenging sce- narios to be further analyzed using the real-time, hard- Navigation System ware-in-the-loop (DSENDS-based) EDL testbed simu- A model of the onboard inertial measurement unit lation. This rapid assessment using the POST-based simulation to support detailed subsystem hardware (IMU) and navigation system is included in the simula- tion. This model uses an estimate of the initial states analyses using the DSENDS-based testbed permits that would be determined as the spacecraft approached quick, but very detailed analysis of any anomaly that Mars, whereas the simulation uses an actual or delivery occurs as entry isapproached. state provided by the interplanetary trajectory analysis. Sample Monte Carlo results of 2000 runs for the 70 A model of the IMU provides adjustments to simulation deg trim shelf configuration are shown in figures 6, 7, generated quantities to account for sensor errors. The and 8. The results at supersonic parachute deploy (see output from the IMU model is used by the navigation Fig. 6) indicate that the parachute deploy constraints on system model to produce an estimate of the vehicle Mach and dynamic pressure were met, and the guidance state for use by other onboard system models (such as delivered the entry system right on its target (note that guidance algorithms, control systems, etc.). More de- tailed information about IMU/navigation system mod- the guidance only acts on the NAV or knowledge state, actual states differ due to knowledge error and els can be found in reference 40. IMU/Navigation error buildup). Figure 7 shows the 8 American Institute of Aeronautics and Astronautics actuaflootprinat tvariousEDLevent(snotethatthe provider is responsible for confirming with the imple- Challengecrater'sedgeoccursat about41.45S, menter that results are reasonable for the system that 286.28Eand41.45S2,86.75Ein)dicatintghatthesu- the data is provided. personidceployfootprinbtasicalldyefinesthetouch- downfootprinstize,butnotlocationT.helastfigure Some model developers are using their own spe- providehsistograminformatiofonrthetouchdowcnon- cific subphase simulation for elements such as the entry ditionsofthelanderT.hesehistogram(sseeFig.8)in- phase only for control system development or parachute dicatethatallthecasems ettheprojecmt etricofverti- phase for sizing and dynamics modeling. Validation of caltouchdowvnelocitylessthan4m/sandhorizontal the results from these subphase simulations with the velocitybelow2 m/swhilemaintaininagnearzero POST-based simulation provides a verification of both. orientatiornelativetoverticalT.hesefiguresshowonly Additional verification of the POST-based simulation afewofthekeyoutpuptarametegresneratedduringa with the real-time, hardware-in-the-loop DSENDS- Monte based simulation in discussed in the third part below Carlo run. A much larger set of date is generated with various subsystem design and assessment teams (POST-DSENDS Validation and Verification). interested in different subsets of the data. Using this Future Work information, overall mission and vehicle statistics as well as risk assessments are provided to the MSL pro- Development of the POST-based simulation sup- ject leaders. Further discussion of Monte Carlo results porting the MSL mission is continuing. References 37, can be found in references 1and 39. 38, 40, 41, and 42 describe various models that either have become or are becoming available soon. These Validation and Verification models (which will be implemented in the POST-based simulation in the near future) include a multibody para- Each model or dataset that is included into or used chute model, surface terrain model, hazard avoidance by the POST-based high fidelity MSL EDL end-to-end logic, 6 DOF entry and terminal descent control sys- engineering simulation must complete the validation tems, reaction control system data and firing logic, a and verification process described below. In this proc- navigation filter and associated sensor models, as well ess (summarized in Table 3), both the model devel- as LIDAR and RADAR models. The simulation is also oper/data provider and the model/data implementer updated as newer, higher fidelity models of various must concur before the process is complete. The devel- systems and the environment are developed and vali- oper is responsible for model formulation and verifying dated. that the model and data are correct for the system it is supposed to reflect. The developer also is responsible PART 2.END-TO-END EDL REAL-TIME for providing computer code of the model formulated SIMULATION TESTBED (DSENDS-BASED) and verifying that the code produces expected results when used in a standalone mode. As such, the devel- The Smart Lander system uses extensive sensor- oper is responsible for providing a set of test data and based real-time control and decision making for preci- results from the standalone runs. The developer is also sion landing and hazard avoidance during the entry, required to provide expected ranges of key input pa- descent and landing phases. Testing and validating such rameters associated with their system and model for use a system requires the use of a high-fidelity, real-time in Monte Carlo dispersion analyses. spacecraft simulator. The Jet Propulsion Laboratory (JPL) is in the midst of adapting its EDL simulator, The implementer of the model into the POST- DSENDS (Dynamics Simulator for Entry, Descent and based simulation must properly include the data or Surface landing) [Ref. 4] for use by the Smart Lander. software into the simulation and ensure that all the ap- DSENDS is an EDL specific extension of the JPL Darts/Dshell multi-mission spacecraft dynamics and propriate interface quantities are provided to the model. The model must produce the same output from within devices simulation toolkit [Ref. 43, 44] used by mis- the POST-based simulation as was produced in the sions such as Cassini, Galileo, etc. [Ref. 45]. standalone test case. Both implementer and developer DSENDS provides for the modeling of the dy- provide their expertise to resolve any discrepancies in the output. The implementer then provides sample input namics of tree-topology multi-body systems with flexi- ble modes within a real-time simulation. It also pro- and output from a typical nominal and off-nominal case that can be checked by the developer using their vides the capability to simulate, in real-time, various standalone capability. When only data is provided, the spacecraft devices such as actuators (e.g. thrusters) and 9 American Institute of Aeronautics and Astronautics sensor(se.g.IMU's).A varietyofEDLrelatedenvi- as well as by additional inter-body forces and torques. ronmenmtodels In the EDL simulation context, these forces and torques (e.g. gravity, terrain digital elevation represent the actions of gravity, aerodynamic forces, maps, atmospheric models), are adapted for real-time and non-linear spring elements between the bodies. The use and support modeling of EDL flight system ele- underlying dynamics engine also support the notion of ments such as parachutes, landers, and terrain interact- ing instrument simulations (e.g. altimeter, LIDAR). prescribed motions where forces and torques are de- rived from a kinematic specification of the trajectory. Together these capabilities allow the modeling of the This allows certain simulation elements to be driven by flight-train dynamics and sensor-based control during the Smart Lander EDL sequence. A block diagram of trajectory profiles rather than force/torque applications the DSENDS architecture and associated model librar- and is useful for modeling elements where the trajecto- ries are well known (e.g. from test data) but the ies are shown in Figure 9and 10. The recent capability to include the aerodynamic libraries from POST allows force/torque relations are not. The rigid-body modeling capability allows models for the entry capsule, heat- high-fidelity aerodynamics modeling, especially during shield, lumped approximations to parachutes, and the entry phase of flight leading to parachute deploy- tether/bridle link elements. The flexible-body modeling ment. Planned extensions for landing kinematics and capability allows modeling of lightweight members dynamics will allow the modeling of contact and impact forces associated with touchdown. Nominal as well as such as landing gear and sensor mounts. The prescribed fault behaviors are incorporated into the device models. motion capability is potentially useful for certain EDL A state-machine driven model switching capability parachute reefing and bridle-lowering models. within DSENDS handles spacecraft separations and re- Real-Time Aerodynamics configurations such as the example in Figure 11. Stub guidance/navigation controller modules for hypersonic DSENDS provides a number of aerodynamics steering, parachute activation, hazard avoidance, and models at various levels of fidelity. The highest fidelity powered descent guidance/control allow standalone simulation as shown inFigure 12and 13. models are encapsulated subroutine libraries from the POST program. These libraries are C routines compiled Some of the system engineering issues related to for the Solaris operating system and embed calls to de- termine aero-coefficients (as a function of Mach and the DSENDS system are presented in a companion pa- Knudsen number and aerodynamics angles) as well as per [Ref. 46]. Here, we focus on a system overview as atmospheric models (e.g. MarsGram [Ref. 48]). Other they relate to the real-time architecture of the simula- tion. We also briefly discuss the verification of these lower fidelity models available for use in DSENDS include analytical linearized as well as table-interpo- capabilities, including comparisons with off-line simu- lated models for aerodynamics coefficients, stand-alone lators (e.g. POST), mission data (e.g. Mars Pathfinder), as well as experimental data (e.g. Smart Lander Rocket encapsulations of the MarsGram atmospheric database, and several table-driven models of atmospheric density Sled tests). and temperature profiles. Within the MSL simulation Real-Time Multi-Body Dynamics. project high-fidelity models from POST are the primary models used for the entry phases of the flight. These DSENDS utilizes the Darts multi-body dynamics models preserve the high-fidelity performance of the original aerodynamics databases within POST. During engine developed at JPL. This dynamics engine pro- vides for extremely fast computations of rigid and the parachute and later descent phases either POST derived aerodynamics or the lower fidelity models flexible body dynamics of a tree-topology multi-body within DSENDS may be used, with the choice deter- system. The underlying computational algorithms for Darts are based upon the Spatial Operator Algebra sys- mined by availability and computational burdens. tem [Ref. 47] and result in the numerical complexity of In order to use the libraries obtained from POST the dynamics algorithm growing only linearly with the number of bodies. Such O(n) algorithms allow high- within a real-time simulation testbed, two options are possible. To maintain maximum fidelity it is desirable fidelity modeling of spacecraft dynamics without com- to execute the libraries on the same processor as that promising fidelity to meet real-time constraint. Con- used for POST execution. The other option is to cross- straining forces and torques between each connected body in the multi-body system are transmitted through compile the code to the typical processor and operating joints that can be of a variety of types. Each body in the system environment used in real-time testbeds. The first multi-body system can also be acted upon by external option requires the utilization of a Sparc® processor 10 American Institute of Aeronautics and Astronautics