Sie sind auf Seite 1von 8

Automatica, Vol. 9, pp. 243-250. Pergamon Press, 1973. Printed in Great Britain.

Apollo LM Guidance Computer Software for the Final Lunar Descent*


Programmes D'ordinateur Pour le Guidage du Module Lunaire Apollo au Cours de L'alunissage Final Apollo LM-Fiihrungs-Rechner-Programme fiJr das Ende der Mondlandung Cpe)ICTBa nporpaMMuposaaH~ CqeTHO-BbIql4CJIbHO~MatUrtrIbi~.q~ naBe~eHrI~
JIyI-Iaoro o T c e r a ano.rLao laOC.rIeRrIefi cTeilem4 c n y c r a
DONALD EYLESt

A guidance computer can best serve an Astronaut during final descent by calculating spacecraft position less frequently, calculating velocity more frequently, and attaching guidance algorithms to the velocity calculation, as needed, in real time.
Summary--In all manned lunar landings to date, the lunar
module Commander has taken partial manual control of the spacecraft during the final stage of the descent, below roughly 500 ft altitude. Because of the irregularity of the lunar surface and the inevitability of some error in the on-board estimate of the spacecraft's position with respect to the local terrain, fully automatic landings, with men aboard, cannot be anticipated until landing fields are prepared on the moon. This report describes programs developed at the Charles Stark Draper Laboratory, MIT, for use in the LM's guidance computer during the final descent. At this time computational demands on the on-board computer are at a maximum, and particularly close interaction with the crew is necessary. The emphasis is on the design of the computer software rather than on justification of the particular guidance algorithms employed. After the computer and the mission have been introduced, the current configuration of the final landing programs and an advanced version developed experimentally by the author are described. INTRODUCTION THE Lunar Module Guidance C o m p u t e r (LGC) contains 2048 words o f erasable memory, and 36,864 words o f fixed m e m o r y for programs, of 15 usable bits each. It is p r o g r a m m e d in two languages, machine language, with an add time o f 24 /~sec, and a list-processing interpretive language, in which, for example, a vector cross-product, with each c o m p o n e n t in double-precision, consumes 5 msec. The L G C ' s work is done in chunks called jobs, o f varying length, which are assigned priorities so that a higher priority job can break in on a lower * Received 12 July 1971; revised 7 April 1972. The original version of this paper was presented at the 4th IFAC Symposium on Automatic Control in Space which was held in Dubrovnik, Yugoslavia during September 1971. It was recommended for publication in revised form by Associate Editor H. A. Spang III. t Principal Engineer, M.I.T. Charles Stark Draper Laboratory, Cambridge, Massachusetts, U.S.A. 243 priority job. Jobs are requested o f and dispatched by an Executive routine. The L G C also has interrupts usually called tasks. The characteristic software entity for the powered flight phases is the so-called Servicer job. Servicer is requested every 2 sec by the task that reads the L M accelerometers. Servicer computes the vehicle mass, its position and velocity vectors, a matrix relating the attitude o f the spacecraft to the orientation o f the inertial platform, and assorted other parameters, such as altitude during linear descent and ascent. During descent only, Servicer uses data from the landing radar to correct the position and velocity estimates. Servicer typically ends by transferring control, still in the same job, to the guidance program appropriate to the phase in progress: P r o g r a m 63, 64 or 65. Guidance uses the information from Servicer to calculate a throttle setting, if the throttleable descent engine is being used, and a thrust axis direction. After Guidance, selected parameters are displayed on the keyboard and display unit (DSKY). During descent and ascent, a task occurs four times a second to maintain displays o f altitude, altituderate, forward velocity, in descent only, and lateral velocity on special cockpit instruments. Also running in the background is a routine which transmits L G C variables to the ground. Ten times a second, in interrupt, the Digital Autopilot (DAP) is processed. The D A P fires attitude control thrusters and gimbals the descent engine to bring the L M to either the attitude calculated by guidance, if the attitude mode switch is in the A u t o position, or the attitude c o m m a n d e d by the Astronaut through his control stick, if the switch is in Attitude-hold. In the latter

244

DONALD EYLES Program 66, and whenever the throttle is returned to the LGC after being operated manually. From then on it is subject to modification, in signed 1 ft/sec increments, by the Astronaut through his rate-of-descent (ROD) switch, a three position toggle with a spring return to detent. The current rate-of-descent algorithm is as follows

mode, properly called rate-command attitude-hold, a deflection of the control stick produces a rotational rate proportional to the degree of deflection; when the stick is not deflected, the vehicle attitude is held constant. An Apollo lunar landing begins with Program 63, known as the braking phase. Fully automatically, Program 63 guides the LM from the periapsis of a 60 x 9 mile lunar orbit to a set of target conditions called high-gate, usually targeted near 7000 ft altitude, 25,000 ft short of the landing site, with a forward velocity of 400 ft/sec and a downward rate of 160 ft/sec. When these targets have been achieved, Program 64, the visibility phase, takes over and begins with a maneuver which brings the landing site into the view of the Astronaut. During this phase the Astronaut can redesignate the landing site to the area where he wishes to land. Program 64 guides the spacecraft to low-gate, typically targeted at 100-200 ft directly above the site. At low-gate, if not selected sooner by the Astronaut, the final landing programs commence. Programs 63 and 64 have brought the LM to the neighbourhood of the landing site. The task of the final landing program, Program 66, is to give the human pilot as much help as possible in maneuvering the LM to a particular, safe touchdown point. The special software problems confronted in this phase stem from the fact that computational demands on the guidance computer are at a maximum for an Apollo mission, and from the intimate interaction with the crew which is required. CURRENT CONFIGURATION In the current configuration of the final landing program, Program 66, the Astronaut has a choice of four modes of control, differing in the division of duties between Astronaut and guidance computer. When both the attitude and throttle mode switches are in Auto, the computer controls the LM attitude, the thrust axis, to produce zero horizontal velocity relative to the surface, and manipulates the LM throttle to maintain the rate-ofdescent commanded by the Astronaut. When the attitude mode switch is in Attitude-hold, the Astronaut is responsible for controlling the LM attitude, when the throttle mode switch is in Manual, he must control the throttle. When neither switch is in Auto, the Astronaut has to control both attitude and throttle, a job which is difficult for one man, but can be divided between Commander and LM Pilot, where the Commander occupies the left-hand position in the LM and the LM Pilot is on the right. It is convenient to discuss the horizontal and vertical channels of the final landing guidance separately. In the vertical channel, the commanded rate-ofdescent is set to the actual rate at the initiation of

Arc = (((/:/o-/:/+ Lag(G -

A r cos ~))/r) -

G)/cos (1)

where Arc is the commanded thrust acceleration magnitude,/:/o is the desired vertical velocity,/:/is the present vertical velocity, G is the vertical acceleration due to gravity, ~ is the angle between the vertical and the LM thrust axis, and Lag expresses the delay between validity of data and vehicle response in guidance commands. AF is calculated from the measured, average acceleration over the preceeding sample interval and represents the instantaneous, present thrust acceleration magnitude. The throttle change commanded is simply Aec-A r, scaled by the spacecraft mass. To speed its response, the rate-of-descent algorithm is executed at twice the Servicer frequency: once ever)~ second. It is processed once every 2 sec in the conventional spot at the end of Servicer, and again in a special job, of higher priority than the Servicer job, requested to occur 1 sec later. Thus, in steady state, one execution follows Servicer, the other is imbedded in the beginning of the succeeding Servicer pass. The rate-of-descent algorithm includes its own accelerometer reading and velocity computation. In the current configuration, the rateof-descent algorithm is written in the LGC's slower, interpretive language, and takes 0.3 sec to complete. The quantity Lag in the equation above is made up of this computation period plus the throttle response time. Reference [I] may be consulted for justifications of the guidance algorithms employed in (he current configuration.
3500

3000

~_ 2 5 0 0 k2000 15I0 05 v -7:: 0 -05 +,ROD L ' r : ~lr!l]!~: Time,


I

-ROD I :!El !:I~I'ITI[IIlITI!I! s

FIG. 1. Current configuration ROD response.

Apollo LM guidance computer software for the final lunar descent Figure 1, taken from a simulation o f the flight computer program for Apollo 14, shows the response of the current-configuration vertical channel to positive and negative ROD switch inputs. Here the plotting frequency is 0.1 sec; i.e. the plot contains ten points for each hatch mark on the abscissa. The noisiness in thrust, and consequently in altitude-rate, between ROD inputs, is due to the granularity of accelerometers (0.01 meter/see) and clocks (0.01 see), which effect the computed altitude-rate/:/and the thrust acceleration estimate At. Program 66 can be entered at any time in the descent by (1) putting attitude mode in Attitudehold, and (2) deflecting the ROD switch. A ROD input is added to mode in Attitude-hold as a condition for entering Program 66 to allow the Astronaut to get the feel of the LM's handling qualities earlier in the landing by maneuvering in Attitudehold, cued by attitude errors displayed by the digital autopilot, and, in addition, to prevent premature selection of Program 66 in the event of a singlepoint failure. The trajectory flown by Program 64 has always been targeted to have a "flare"--so that around 500 ft altitude the descent-rate is moderate and suitable for partial manual takeover using Program 66: around 15 ft/sec. Apollo 11 entered Program 66 at 410 ft altitude and achieved touchdown 140 sec later; Apollo 12 used Program 66 for 110 sec starting at 368 ft altitude. During Apollo ITs final landing, dust caused more difficulty than expected and, as a result, a horizontal channel was added to the final landing guidance. Dust was first kicked up when the LM was at 300 ft altitude, and later, below 100 ft, it was dense enough to make the view from the window useless as an attitude reference. Attitude fluctuations which resulted interfered with maintaining zero horizontal velocity, although dust did not interfere with the Astronaut's ability to judge his velocity across the surface, since rocks were visible through the dust. What was needed was a mode in which the computer would maneuver the spacecraft to maintain zero horizontal velocity relative to the surface. Accordingly, when the attitude mode switch is in Auto, the guidance computer employs the following algorithm in the horizontal channel Arc = ((VD-- V)/0 - 0.4A FCold G
-

245

vertical, posture, the horizontal component of A~c is limited to ,4U~lm, an acceleration chosen so that Aalim/G is the tangent of 20 . Note that all the vectors mentioned are in the coordinate frame of the inertial platform, which is aligned so that at the time of the landing its x-axis is vertical, facilitating separation of the vertical and horizontal channels. The horizontal control algorithm, which consumes about 0.3 sec is executed every 2 sec at the end of Servicer. When the current configuration in its final form was first tried, it ran with a dutycycle of over 96 per cent--worse than the previous maximumdutycycle case, about 90 per cent in Program 64. Astronaut initiated activities, such as DSKY monitoring of computer parameters, could increase this to over 100 per cent, and there was a further indeterminate load due to increments and decrements of the LGC's interface counters, each of which required one computer memory cycle. It should be noted here that excessive interface activity caused a shortage of computation time during the landing of Apollo 11. To meet this difficulty, logic was provided which omits one cycle of Program 66 guidance, two processings of the vertical algorithm and one of the horizontal, when the time available is insufficient. ADVANCED CONFIGURATION To see if the capabilities of the guidance computer could be utilized more fully in assisting the Astronaut during this phase, an experimental, advanced version of the final landing program was developed. This was an attempt to profit from Apollo experience and was not intended to fly on an actual Apollo mission. However it was Coded for the guidance computer, and its development benifited from use o f all-digital and real-time hybrid simulators at MIT, and the LM Mission Simulator at Cape Kennedy. The program used as a test bed for this work was an offqine version of the LM flight program in which the author developed a solution to difficulties caused b) a shortage of execution time which was more general than the omission strategy, mentioned above, adopted in the current configuration after the horizontal algorithm was introduced. S u c h difficulties arise when more than 2 sec of activity are scheduled every 2 sec by the Servicer job. In the off-line program, the Servicer cycle is allowed to take longer when it is overloaded. All computations which depended on Servicer recurring at exact intervals are modified to use a computed, variable guidance period instead. The computationtime difficulties of the current configuration suggested that further improvements of Program 66 would be easier where execution-time was less critical.

(2)

where AFC is the commanded thrust acceleration, Vo is the velocity of the lunar surface below the LM, V is the present velocity, G is the acceleration due to gravity, and ~ is the time constant. It was found that 5 sec was a suitable value for 3. The vertical component of Arc is set to lunar gravity. To prevent attitude excursions in pitch or roll greater than 20 away from the erect, thrust axis

246

DONALD EYLES Further, an attempt is made to speed up the response to Astronaut inputs, and to achieve greater smoothness in the throttle profile, and thus in altitude-rate, in the absence of Astronaut inputs. The response time of the current rate-of-descent logic is limited (1) by the 1 sec interval at which ROD switch inputs are counted, and (2) by the desirability of using a time constant longer than the sample interval for reasons of stability. Smoothness in the throttle profile is militated against by the granularity of LM accelerometer readings and of the LGC clocks. For shorter sample intervals, preferable in terms of response time, these granularities are relatively exaggerated in the throttle change commanded. A scheme was adopted in which, when the velocity error is greater than a certain limit, the rate-of-descent equation is executed immediately, except that executions are not allowed more frequently than every sec. If the velocity error is less than the limit, the rate-of-descent equation is executed T sec after the preceeding pass, where T is the time constant used the preceeding pass. This arrangement borrows the idea of a deadband from autopilot practice--only the system is not dead within the deadband, only quiet. Additionally, logic was provided so that the time constant used when velocity error is outside of the quietband (TI) can be different from that used inside the quietband (T2). Note that in this arrangement response to ROD switch inputs is faster than in the previous configuration both because (1) ROD inputs are looked for, and in general responded to, four times as frequently as before, and (2) a shorter time constant can be used because the sample frequency, when velocity error is outside the quietband, is twice as high. The noise in commanded thrust is reduced by the use of a sample interval, when within the quietband, longer than the 1 sec sample interval of the current configuration. Noise was also reduced by the use of a finer-grain clock for measuring the sample interval. Optimum choices for the time constants ~l and T2, and for the velocity error limit L have not been sought. Good performance has been obtained in simulations with zl = 1 see, z 2 = 2 sec and L = 0 . 2 ft/sec.

In the advanced configuration the guidance algorithms for Program 66 are all detached from the Servicer job. They are attached to the task processed four times a second to maintain displays of altitude, altitude-rate, and forward and lateral velocity on cockpit instruments. Besides the displayed values, this routine outputs a velocity vector corrected for lunar rotation. None of the guidance equations in use in Program 66 require a position vector as input. Throughout Program 66, a job is requested every sec by the display routine. In this job, a guidance algorithm, or a DSKY display, or nothing, is processed according to decision logic executed at the time. This logic is shown in Fig. 2.
Every I14 ,ec

',:,,I

1 ...............
controller Y
N

d e f l e c t ion~

lost p ~ ?

t-

t~
I HZco., y =.Z ..... , I

t
Flo. 2. Program 66 guidance decision logic.

I I I J

Vertical channel The first part of the logic concerns the vertical channel. This is the current configuration's rate-ofdescent equation, given in equation (1), recoded in machine language so that its execution time is reduced by an order of magnitude, to about 25 msec. Because the quantity Lag is correspondingly reduced, the velocity compensation term is eliminated. As programmed this equation becomes simply

Rote - of - descent algorith

. . . . . .

I
ROD switch input

-- . . . .

f v'ry''c
until n e x t R O D input

-3

-2

-I

I s

Time ,

A FC = (((/:/o--/:/)/Z) -- G)/COS.

(3)

FIc. 3. R O D

algorithm timing.

Apollo LM guidance computer software for the final lunar descent Figure 3 shows the pattern of rate-of-descent algorithm executions following a ROD switch input that occurs at time zero. If the previous execution were at time - 0 . 2 5 the same pattern would occur 0.25 sec later. The dynamics of the advanced configuration of the rate-of-descent system are as follows: the ROD input changes commanded vertical velocity by 1 ft/sec. When next checked, ordinarily within sec, velocity error exceeds 0.2 ft/sec and the ROD algorithm is executed immediately, and again sec later, with z set to I sec. Quick modification of the descent rate is the aim now. It does not matter that the acceleration estimate AF, based on a sample interval of 0.5 sec, is inaccurate. When velocity error becomes less than 0-2 ft/sec, the transition to the quiet mode begins. Deadbeat response is sought by letting the next sample interval equal the current time constant, 1 sec. The acceleration estimate for this sample interval is improved because the interval is lengthened. In the next execution of the rate-of-descent algorithm, the third since the ROD input, z is set to 2 sec, and the fourth execution, which thus occurs 2 sec later, benefits from a still longer sample interval. The rate-ofdescent algorithm continues to be executed every 2 sec until excited by another ROD input. This system takes advantage of its sample data nature by manipulating the time constant and the sample interval together, to achieve quicker response when velocity error is outside a quietband, and smoother control when it is inside.
3500

247

and the reduction of noise during quiet periods. By chance, these plots show the reaction speed of the advanced configuration in the best possible light; response to the ROD inputs begins almost a full second sooner than in the current configuration. Finally, it should be pointed out that a guidance or control system such as this, processed more or less frequently according to need, offers potential computation-time savings over a system processed at a constant frequency intermediate between the fastest and slowest frequencies of the variable system. Whether these savings can be realized depends on the average frequency of excitation, on the time-consumption of the decision logic which must be processed every pass, and on whether the outer loop, the Servicer, can tolerate temporary periods of high activity in the inner loop. As a general statement, it is computation time which is traded for the improved performance of the advanced configuration of the final landing program. However, as a practical fact, this is true only when the i,mer loop, described above and below, is excited by Astronaut inputs. Horizontal channel The remainder of the logic shown in Fig. 2 concerns the operation of the horizontal channel of the Program 66 control system. The horizontal control equation is basically unchanged from the current configuration. In general it is executed every 2 sec. An index, called HZCOUNT, is incremented every sec, and the horizontal control equation is executed when the index equals zero modulo 8. A display computation is executed every second, when the index equals 2 modulo 4, and the entire pattern can be shifted at any time by reinitializing the index register. This mechanism is necessary to the site designation system provided for Program 66 in the advanced configuration as, will be seen. Ref. [2] gives a description of the landing point designation (LPD) system for Program 64, to which this system for Program 66 can be compared. The central idea of the Program 66 designation scheme originated with an Astronaut suggestion. It was proposed that the guidance computer compute and display on the DSKY the angle, along the LPD reticle on the left-hand window, at which the LM Commander can see the spot where he would touch down if at that instant he switches from Attitude-hold to Auto, returning control of the vehicle's attitude to the horizontal velocity pulling algorithm. In effect the Astronaut would enter a "mark", by switching to Auto, when the site he desires passes a spot on his window specified by the computer. Figure 5 shows the arrangement of the LM cockpit, including the LPD reticle on the left-hand window.

3000

I---

2500

20OO 15 I0 0,5
,"r
!

+ROD ::

-ROD
I ! :, s r i p i[',1[',1 i p I11[

Time

FK;. 4. Advanced configuration ROD response. Taken from a simulation of the advanced configuration, Fig. 4 shows thrust and vertical velocity reacting to positive and negative ROD switch inputs, and as in Fig. 1, to which Fig. 3 should be compared, the plotting frequency is 0.1 sec. The varying interval between commanded throttle changes is clearly seen in the thrust plot, as is the modulation of the throttle in moments of excitation,

248

DONALD EYLES

....

LM Controller ~ commander Computer


display

Computer
keyboard

FIG. 5. LM cockpit configuration.

.. So+ z.0~;~

"H

j"

spot ---:
FIG. 6. Program 66 site designation geometry.

See Fig. 6. The angle 0 which is to be displayed to the Astronaut depends on the vehicle's attitude, specifically on the body z-axis, and on the two distances H and D. H is the LM altitude. Since the LGC's idea of the spacecraft's altitude has been continuously corrected by landing radar readings, this altitude is valid for the landing radar antenna, which is mounted on the bottom of the LM descent stage, so the quantity H must be corrected by adding the distance between the antenna and the Astronaut's eye, about 4 m. D is the distance the spacecraft will travel while the horizontal control algorithm nulls its forward velocity to zero. It depends on the horizontal rate, which is known, and the characteristics of the horizontal velocity nulling algorithm. The quantity D is calculated by modeling the acceleration profile flown by the horizontal algorithm when mode is returned to Auto, as follows D = Delay Vu + (V 2 + V ,2 u)/2Amim (4)

where Vn is the LM horizontal velocity. V'n is Vu or zAmlm, whichever is smaller, and Delay is the average time it takes for the attitude maneuver to begin after mode is switched to Auto, plus half the time it takes to reach the braking attitude once the maneuver has started. Ann m is the horizontal acceleration limit which is applied to the output of equation (2). Given D, the line-of-site vector can be easily computed, and from that 0 can be calculated. The time quantity Delay is available for tuning the calculation for the greatest accuracy in the velocity range contemplated for use of the Program 66 site designation capability. Note that if the horizontal control algorithm is executed strictly every 2 see, then there is a 2-sec variability in when, following the "mark", the spacecraft reacts to it. Thus the actual braking distance could vary by 2 Vu. For a horizontal velocity of 50 ft/sec--a typical one for the use of this capability--D could vary through a range of

Apollo LM guidance computer software for the final lunar descent 100 ft. Because this was considered an unacceptable limitation on accuracy, the indexing mechhanism described above was implemented. It allows the pattern of the horizontal calculations to be reinitialized at any time, so an execution of the horizontal algorithm can be commanded immediately when the Astronaut is seen to have switched to Auto. This reduces the variability in time to 0.25 sec--actually 0.35 sec including the uncertainty in how soon the DAP will act on guidance commands--and the variability in distance is reduced correspondingly. Figure 7 shows the timing of horizontal channel executions, including a switch from Attitude-hold to Auto at time zero. Values of H Z C O U N T are in parentheses. It is possible for two executions of the horizontal algorithm, which consumes 0.3 sec, to be requested sec apart. The job priority structure is used to insure that the first finishes before the second starts. Servicer, running in the background, is delayed by this momentary high computational load, but in the test-bed program this has no ill effect. During this phase the outer, Servicer loop contains no guidance; its function is occasionally to provide new base values, incorporating radar data, for the simpler navigation, using only accelerometer data, performed four times a second in the guidance loop.
1 Horizontal algorithm velocity

249

redesignations are also effected by discrete "clicks" of the attitude stick. Velocity increments are input in a spacecraft-fixed coordinate system consistent with the forward and lateral velocity indicators in the LM cockpit. A pitch deflection of the stick modifies velocity in the forward or rearward direction; a roll deflection adds velocity to the left or right. Two ft/sec was found to be a suitable magnitude for each velocity increment. Commanded horizontal velocity can be reset to zero either by putting in opposite stick deflections, or by switching briefly to Attitude-hold and back to Auto. This capability adds no new element to the software organization, except for an interrupt triggered by the stick deflection to count it.

Usage of the advanced configuration


Here is an example of the use of the capabilities of the advanced configuration. At 500 ft altitude, with a descent rate of 15 ft/sec, forward rate of 80 ft/sec, the LM Commander switches mode to Attitude-hold and clicks the ROD switch, selecting Program 66 and reducing the descent rate. He looks forward and, if necessary, manually maneuvers the spacecraft to establish a heading toward the landing spot he picks out. When the Astronaut entered Atttitude-hold the LM was pitched back about 25; when forward velocity has dropped to 40-50 ft/sec, he pitches the vehicle forward so that the thrust axis is vertical and the forward velocity becomes constant. The LM Pilot reads aloud the angle 0 displayed on the DSKY, which is constant if altitude is constant. Making sure that the plane of the LPD reticle and his eye includes the velocity vector, and, if necessary, yawing the LM around its thrust axis so that it does, the Commander watches the surface at the point on the reticle indicated by the computer. He puts his hand on the attitude mode switch, and when the spot where he wants to touch down crosses the indicated point, he switches to Auto. The computer responds by pitching the spacecraft back 20 . As the horizontal velocity is nulled to zero the computer returns the LM to the erect attitude. If the LM is not now precisely over the desired spot, the Commander deflects his stick to establish a small rate in its direction. When the destination is reached, he flips mode briefly to Attitude-hold and back to Auto and horizontal velocity is again nulled to zero. Then, using the ROD switch to command an appropriate descent rate, he lands. CONCLUSION Two systems have been described for use by the Apollo Astronaut in the final stage of the lunar descent. In the advanced configuration rate-ofdescent response is improved, and to the horizontal channel is added a site designation system which

nulling

J~ e computation and DSKY display

Altitude-hold -5 -4
-.7,

to auto
1 1 1 /

-2

-I Time,

O s

FIG. 7. Horizontal algorithm timing. Besides a new rate-of-descent scheme and the site designation system just described, another capability was added in the advanced configuration: the Astronaut is enabled to modify the horizontal velocity maintained by the horizontal algorithm. In the current configuration this velocity is always zero. In the advanced configuration it is zeroed when Program 66 is selected, and whenever attitude mode is switched from Attitude-hold to Auto, but it can be modified by the Astronaut, incrementally, by discrete deflections of the attitude control stick. This capability operates only when mode is Auto, so manual maneuvering with the stick is not affected. It should be noted that Program 64 site

250

DONALD EYLES particuliers de guidage utilis6es. Apr~s I'introduction de l'ordinateur et de la mission, la configuration pr6sente des programmes d'alunissage final et une version avanc6e d6velopp6e exp6rimentalement par I'auteur sont d6crits.

makes it easier to place the LM precisely in a suitable spot. It is suggested that the service rendered the Astronaut can be maximized by providing a high-frequency velocity computation, to which various guidance algorithms may be connected as needed, while allowing the frequency of the other navigational computations, notably calculation of position and incorporation of radar data, to languish during moments of inner loop excitation.
Acknowledgement--This report was prepared under DSR
Project 55-23890, sponsored by the Manned Spacecraft Center of the National Aeronautics and Space Administration through Contract NAS9-4065. The site designation scheme for Program 66 was suggested by Astronaut John Young. The publication of this report does not constitute approval by the National Aeronautics and Space Administration of the findings or the conclusions contained therein. It is published only for the exchange and stimulation of ideas.

Zusammenfassung--Bei allen bisherigen bemannten Mondlandungen flbernahm das Lunar-Modul-Kommandogerfit (LM) eine teilweise manuelle Steuerung der Landef~ihre w~ihrend des Endabschnittes der Landung unter etwa 200 m. Wegen der Unregelm/if~igkeiten der Mondoberflfiche und der Unvermeidbarkeit yon Fehlern bei der bordgebundenen Schatzung der Raumschiffsposition zu dem tirtlichen Gel~inde, kSnnen vollautomatische Landungen mit Menschen an Bord erst vorgenommen werdea, wenn auf dem Mond Langeplhtze vorbereitet sind. Dieser Bericht beschriebt Programme, die am Charles Stark Draper Laboratorium, MIT, zur Benutzung im LMFtihrungsrechner w~ihrend der LandeschluBphase entwickelt wurden. In dieser Ziet erreichen die Rechenanforderungen an den Bordrechner ihr Maximum, wobei eine besonders enge gegenseitige Einwirkung mit der Besatzung niStig ist. Der Nachdruck liegt mehr auf dem Entwurf der Rechnerprogramme als auf der Rechtfertigung der angewandten speziellen FiJhrungsalgorithmen. Nachdem der Rechner und die Aufgabe vorgestellt sind, werden die laufende Konstellation der Programme ffir die Lande-Endphase und eine vom Verfasser experimentell entwickelte verbesserte Version beschrieben. Pe31oMe--,/2o crlx hop BO Bcex nn.qOT~IpyeMblXnoca~Kax Ha .nyHy, KOMMaH,~yloIIIFn~JlyHHbIM OTCeKOMHaCTHttHO ynpaBna,rl BpyqHylO KOCMHqeCKHMKopa6neM BO BpeM~ noc~eaHeli cTynena cnycKa HltXe npll6.qrt3HTe~bHO 500 ~T BblCOTb[. Bcne~cTnBe HepOBHOCTH.rlyHno~lHoBepxHocTH H BcJ]eIICTBHe HeH36eXHOCTI4 I-leKOTOpOi;i HeTOqHOCTrl pacqeTa Ha 6opTy Kopa6na o no.qoxeHnrt KOCMHHeCKOFO~ooa6~Ia OTHOCHTeJlbHO l'pyH/a no MecTy, He.rlb3~l O)KH~aTb UO.rlHOCTblOaBloMairlqecKylo nocaa~y na ~yHy do Tex hop, noKa Ha .qyne He
HO~IFOTOBflT ILrlOIIIaIIH 32Jlfl n o c a a o K .

REFERENCES [I] A. R. KLUMPP: Apollo LM Descent Guidance, Attitude Control and Trajectory Targeting--Fundamental Concepts and Algorithms, Fourth lFAC Symposium on Automatic Control in Space (1971). [2] A. R. KLUMPP: A manually retargeted automatic landing system for the lunar module. J. spacecraft & Rockets 5, 129-138 0968). R6sum6--Dans tousles alunissages humains passes, le commandant du module lunaire a pris le contrSle manuel partial de v6hicule spatial au cours du stage final de la descente audessous d'environ 150 m~tres d'altitude. Du fait de l'irr6gularit6 de la surface lunaire et de l'in6vitabilit6 de certaines erreurs dans l'estimation "3. bord de la position du v6hicule spatial ~. l'6gard du relief, les alunissages enti~rement automatiques avec hommes ~. bord sont impossibles rant que des terrains d'alunissage ne seront pas pr6vus sur la lune. Ce rapport d6crit les programmes d6velopp6s aux Laboratoires Charles Stark Draper, MIT, pour emploi dans l'ordinateur de guidage du module lunaire au cours de la descente finale. A c e moment les demandes en calculs sur I'ordinateur de bord sont au maximum, et une interaction particuli~rement serr6e est n6cessaire avec l'6quipage. On souligne l'importance de la conception du programme d'ordinateur plutft que la justification des algorithmes

B 3Tofi CTaTbe onncblaarorca nporpaMMbl pa3pa6orannble 8 za6opaToprm M I T qapab3a lllxapK ,/Ipenepa a~a
HaBe~teHl,trl OTCeKOB Hocpe~,CTBOM cqeI'HO-BblqHC.~HTe-qbHblX

MaturIH BO BpeMa nocneaae~ cTynenn cnycKa. B HacTomuee BpeMfl nMeeTc~t MaKCHMa.qbnbI!~ cnpoc Ha cpe~craa nporpaMMHpOBaHH~ cqeTnO-abltlnCnaTe.rlbnO~ MamUnbl Ha 6oply, W Tpe6yeTca oco6o TecHa~I roonepattH~ c cy,/1oso~
KOMaH,/I6kl. O c o 6 0 noaqepK~BaeTcrt Heo6xo.B.HMOCTb npoel<-

UHI,I cpe~/CTB nporpaM~lpOBaHH~ CqeTHO-BBIqHCYlHTe.qBHblX MalLIHH, a He a.qFOpHXM HaBellettnfl. B 3TO~ pa6oTe onncb~BaeTca Hacxofltttaa KonqbHrypaurm nporpaMM nocne~nefi clyneHn cnycKa n pa3BrITaa aBTOpOM 3KTOpOM 3KcneprtMenTaJ~bHaa Bepcna.

Das könnte Ihnen auch gefallen