A multiple-impulse function for orbital transfer and its derivatives NASA Technical Memorandum Langky Research Center NATIONAL AERONAUTICS AND SPACE ADMINISTRATION , 0. C * 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No. NASA TM X-2492 4. Title and Subtitle 5. Report Date A MULTIPLE -IMPULSE FUNCTION FOR ORBITAL March 1972 6. Performing Organization Code TRANSFER AND ITS DERIVATIVES 7. Author(s) 8. Performing Organization Report No. James F. Kibler and Lawrence H. Hoffman L-8106 10. Work Unit No. 9. Performing Organization Name and Address 815-00-00-00 NASA Langley Research Center 11. Contract or Grant No. Hampton, Va. 23365 13. Type of Report and Period Covered 12. Sponsoring Agency Name and Address Technical Memorandum National Aeronautics and Space Administration 14. Sponsoring Agency Code Washington, D.C. 20546 15. Supplementary Notes 16. Abstract A multiple-impulse function is represented as a sequence of single-impulse func- tions. The single -impulse transfer which yields the velocity change required to transfer from a specified initial orbit to a partially specified final orbit is developed. Analytic derivatives of the function are obtained for use in optimization techniques. A four-impulse transfer is outlined. The analytic derivatives may allow more efficient optimization than numerical derivatives do. 17. Key Words (Suggested by Author(s)) 18. Distribution Statement Impulsive transfers Unclassified — Unlimited SYMBOLS a '• semimajor axis, km e j eccentricity f true anomaly, deg i inclination, deg n number of impulses P unit radius vector lying in orbit plane and pointing toward periapsis Q unit radius vector lying in orbit plane and pointing toward 90° true anomaly R unit vector along line of intersection between two noncoplanar orbits r magnitude of radius at a given true anomaly, km r,h,n radial, horizontal, and normal components of velocity V velocity magnitude, km/sec AV magnitude of change in velocity required to transfer between two orbits, km/sec W unit vector normal to orbit plane and completing the PQW right-handed triad |3 angle between two velocity vectors, deg y flight-path angle, deg 9 angle between two orbit planes, deg M gravitational constant, fl right ascension of ascending node, deg w argument of periapsis, deg Subscripts: 0,1 refers to the first or second orbit, respectively T total Bars over quantities denote vectors. ANALYSIS Derivation of the Impulse Function for a Single-Impulse Transfer The single-impulse transfer between two noncoplanar intersecting Keplerian orbits can be determined analytically if five elements are known on the first orbit and four elements are specified on the second orbit. It is arbitrarily chosen to specify the classi- cal elements (ao,eQ,io,wo>^o) on the initial orbit; and the second orbit is specified by the orientation angles (ii,wi,flj_^ and one of the shape parameters (for example, ajY Of course, other choices are possible and would result in a different formulation. The first requirement is to find the line of intersection of the two orbits which is uniquely determined by IQ, OQ, ij, and fij. It is convenient to perform a coordinate transformation from the XYZ set of axes to the PQW system as shown in sketch (a). Periapsis Sketch (a) - Transformation of XYZ to PQW. In the PQW system, P is a unit vector pointing in the direction of periapsis, Q is a unit vector lying in the plane of the orbit and pointing toward a true anomaly of 90°, and W is a unit vector, normal to the orbit plane, which completes the right-handed triad. By rotating through the angles J2, i, and u>, the resulting transformation (ref. 4) is P = (cos co cos J2 - sin co sin S2 cos i) i + (cos co sin fi + sin co cos n cos i) j + (sin co sin i)k Q = (-sin co cos J2 - cos co sin J2 cos i) i + (-sin co sin fi + cos co cos £2 cos i) j (1) + (cos co sin i)k W = (sin fl sin i) i + (-cos n sin i) j + (cos i)k where i, j, and k are the corresponding unit vectors in the XYZ coordinate system. The unit vector R, along the line of intersection of the two orbits, is determined by W Q XWj R= (2) where WQ = (sin IQ sin OQ) i + (-sin IQ cos OQ) j + (cos ig (normal to first orbit) and Wj = (sin ij sin fli) i + (-sin ij cos flj) j + (cos ij)k (normal to second orbit) are known from equation (1) . Now, ». •*• i . j W 0 X W 1 = sin ig sin -sin ig cos cos ig sin i sin -sin ij cos cos i Expanding this expression results in -sin ig cos ij cos fig + sin ij cos ig cos fij -sin ig cos ij sin fig + sin i^ cos ig sin fij (3) sin ig sin ij sin Afi where Afi = fij - fig. The magnitude of Wg x Wj is given by I Wg x Wj = (sin2ig cos2^ cos2fig - 2 sin ig sin ij cos ig cos ij cos fig cbs fij + sin2ij cos^ig cos^fij + sin^ig cos^ij sin2fig -2 sin ig sin ij cos ig cos ij sin fig sin fij + sin2ij cos2ig sin2fij \l/2 Simplifying this expression results in * i = r1 - (cos Wg x Wj / ig cos ij + sin ig sin ij cos N?~| Afi) (4) Substituting equations (3) and (4) into equation (2) yields the expression for a unit vector in the XYZ system along the line of intersection between the two orbits: -sin ig cos ij cos fJQ + sin *! cos '0 cos R =• -sin IQ cos ij sin OQ + sin 4 cos '0 sin (5) r1 - /(cos 10 cos ii + sin IQ sin ij cos ASJJx2i sin ig sin ij sin AO There are two points where an impulsive maneuver can take place, that is, in the direc- tion of either positive or negative R. Consider first the case of the transfer point lying in the positive R direction. The transfer -point true anomaly on each orbit must be determined. For the first orbit cosf o or COS fg = Pg • R (6) Substituting the expression for PQ from equation (1), and equation (5) into equation (6), and simplifying yields -cos uv)(sin in cos ii - sin i-, cos in cos Afi) + sin wn sin ii sin J cos f0 = ['-(• cos I cos i + sin I sin i cos A \9~1 °)J 1/2 In order to resolve the ambiguity involved with the cosine, an expression for sin is ,i required. By definition, P0 X R = W( PnllR sin f o or WQ sin f0 = PQ x R From equation (2), _P0x(w0xW1) W0 sin f0 = P0 x WQ X WQ X Wj Using the identity A x (B x c) = (A • C)B - (A • B)C where A, B, and C are any vectors yields (P0 • W^WQ - (P0 • WQ V W0 sin f0 = ^- V ° |W0 x Wj But P0 • W0 = 0; therefore, iWr W0 sin fQ = or sin (8) WgXWj Substituting equation (4) and the expressions for PQ and Wj from equation (1) into equation (8) and simplifying yields sin u>g(sin ig cos ij - sin ij cos ig cos Afi) + cos u>g sin ij sin sin 1/2 1 - (cos ig cos ij + sin ig sin ij cos Afi) In like manner, the equations defining the true anomaly of intersection on the second orbit are cos u;i(cos ig sin ij - sin ig cos ij cos AflJ + sin u>i sin ig sin cos fj = - —- (10) 1/2 1 - (cos ig cos i^ + sin ig sin ij cos A Or and -sin wi (cos in sin ii - sin ift cos ii cos Afi) + cos cui sin in sin Afi 1/2 1 - (cos ig cos ij - sin ig sin ij cos Afi) The second point where an impulsive maneuver can take place is in the direction of nega- tive R. The true anomaly of intersection on the first orbit is given by cos fg = Pg • (-R) and ftg sin fg = Pg X (-R) Substituting and simplifying as in equations (7) and (9) yields cos O)g(cos ij sin ig - sin ij cos ig cos Afi) - sin o>g sin ij sin cos fg = (12) 1/2 1 - (cos ig cos ij + sin ig sin ij cos Afi) and slnf -sin u>g(cos ij sin ig - sin ij cos ig cos Afij - cos u>g sin ij sin °" , - (cos ig cos ij + sin ig sin ij cos AO) Similarly, the equations defining the true anomaly of intersection on the second orbit are -cos wjfcos ig sin ij - sin ig cos ij cos An) - sin ojj sin ig sin cos fi = — (14) 1 - (cos ig cos ij + sin ig sin ij cos An) and sin cojfcos ig sin ij - sin ig cos ij cos An) - cos a)j sin ig sin sin fi = 1 - (cos ig cos ij + sin ig sin ij cos An In other words, and For an impulsive maneuver to take place at the point of intersection, the magnitudes of the radii on the first and second orbit must be equal. Therefore, a constraint equation at the point of intersection is - eg2) ai(l - e^) 1 + 6g COS fg 1 + 6j COS fj The elements ag, eg, fg, fj, and a^ or e^ are known. Therefore, equation (16) relates the unknown element in the second orbit (e\ or a^ to known parameters. Next, an expression for the magnitude of the vector difference in velocity (AV) between the two .orbits is required. From sketch (b), using the law of cosines, (AV)2 = V02 + Vj2 - 2V0V! cos 0 (17) where 0 is the angle between the two velocity vectors. Orbit 0 Intersection point Orbit 1 Sketch (b) - Geometry of AV. At the maneuver point, by using the equations 1/2 V-Rr-£ and r= 1 + e cos f the expressions for the magnitude of the velocity on the first and second orbit are + 2e cos f + >2) 1/2 (18) e 2 - 0) and cos \U/2 (19) From sketch (c), by using the law of cosines for spherical triangles, the angle between the orbit planes 6 may be found from cos 8 = -cos i0 cos(l80° - ij) + sin iQ sin(l80° - ij) cos or cos 9 = cos ig cos ij + sin IQ sin ij cos Afl (20) Sketch (c) - Computation geometry for From sketch (d), the velocity vectors may be written as Siny 0 V n = V , cos YQ 0 and sinyj COS yj COS cos y sin ( where YQ and y., are the flight-path angles on the first and second orbits. The angle between the velocity vectors /3 may now be found from Vn • Vi cos ]8 = = cos 0 cos yQ cos y^ + sin yQ Sin (21) ^- . 10 n Sketch (d) - Velocity vectors in the radial, horizontal, and normal coordinate system. Substituting the flight-path angle relations from reference 4 sin y = e sin f (l + 2e cos f + e2) and 1 + e cos f cos y = 1/2 (l + 2e cos f + e2) and equation (20) into equation (21) yields (l + eg cos fg)(l + ej cos fjVcos IQ cos ij + sin ig sin ij cos Afi) + (eg sin fg)(ej sln *: CQS £ = (22) [(1 + 2e0 cos fg + eg^j^l + 2ex cos ft + e^jj Substituting equations (18), (19), and (22) into equation (17) results in the required expression for AV: AV=, r (l + eg cos f Q )(l + ej cos f i) (cos ig cos i^ + sin IQ sin i^ cos infl ]j fU/2 sin T sin (23) 11 The set of equations (7), (9) to (16), and (23) are those required for an analytic solution to the problem of a single-impulse transfer between two noncoplanar orbits. Two solu- tions result from the positive and negative cases for R. In addition, if ej is a function of aj in the constraint equation (16), two solutions arise from the quadratic in ej. All the multiple solutions may not be physically possible. For example, an arbitrarily spec- ified aj may result in a maximum radius on the second orbit which is less than the smallest radius on the initial orbit. That is, the constraint equation (16) cannot be sat- isfied. In addition, if the central body has a finite radius, a region of mathematically possible orbits must be rejected since they intercept the surface of the planet. In the event no practical solutions for the transfer exist, a more realistic set of variables (a.i,ii,an,fli) on the second orbit must be chosen. In the event of multiple practical solutions for the transfer, a choice may be based on mission considerations, such as choosing the transfer which results in the lowest AV. Partial Derivatives of the Single-Impulse Function In orbital transfer problems, it is usually desirable to minimize AV with respect to the control parameters. Many of the minimization algorithms require partial deriva- tives of AV with respect to the independent variables. Finite-difference approxima- tions to the partial derivatives are commonly employed. However, in complex multiple- impulse problems, it may take relatively large amounts of computer time to develop the approximations. If exact partial derivatives are available, the minimization algorithms are generally more efficient and less computer time is used. Therefore, analytic partial derivatives of equation (23) have been developed. It is assumed that a, e, i, w, and O are known on the initial orbit and that a, i, o>, and n are known on the final orbit. Eccentricity of the final orbit is a dependent variable defined in the constraint equation (16). For convenience, define the following intermediate quantities as AO = n[ao(l - eg2)]"1 Aj = M[ai(l - e^)]'1 BQ = 1 + 2eg cos fg + eg2 BI = 1 + 2ej cos fj + ej2 C = -2 M [a 0 (l-e 0 2 )a 1 (l-e 1 2)]" 1/2 D = (l + eg cos fg)(l + el cos fjYcos io cos ij + sin ig sin ij cos Aft) + egej sin fg sin fj 12 Thus, equation (23) can be written as AV = (AgB0 + AjBj + CD)1/2 Or (AV)2 = AflBo + AjBj + CD The required derivatives can be expressed in the form 9AV 1 8(AV)2 da 2AV dot 2AVI da da da da da dal (24) ^ ' where a represents any one of the variables aQ, CQ, IQ, O>Q, OQ, a-^, ij, Wj, and Oj. It is necessary now to develop the partial derivatives of the following six interme- diate parameters: : Let -1 -e 0 2 )] then Similarly, 9AQ = 2AQ6Q 8eQ 1 - 602 and 8Ap = 8Ap = 3AQ = 8Ap _ where 01^ represents aj, i^, wi, and 13 (2) : Let then 9Ai i . 612 9a0 Similarly, 9e0 ~ i _ 612 2e1A1 2e1A1 8ei A! SAj _ 2ej Aj " a 9i i _ 6J2 9 2e1A1 9ei The —— remain to be determined. act <3> T ^ BQ = 1 + 2eg COS fQ + 6Q Then 9aQ since 9 cos =0 14 by inspection of equation (7). Similarly, 3B 9 cos f 3BQ 3 COS fg 3BQ 3 COS fg _ = 2eQ 9ajQ g^ = 2e0 — — ~ = 2G 3 cos fp 3BQ ~ 3 COS fn The —3- remain to be determined. 3d! Bj= 1 + 2ej cos fj + ej then 3Bi 3 cos fi v 3ei N 3ei = 2e + 2 COS f + e 2 cos f + e 3a7 l -^T- ( l l) 3a7 = ( l l) 3^7 since 8 cos by inspection of equation (10). Similarly, . . ei 3Bii 3 cos cos fii , . 3ei ei = 2(cos ft +ei)—i —^ = 261 —-9 — i + 2(cos fx + ej) y—9l i 9l 0 v 1 1; 9eeQ 3ii0 iiQ ^ 3Bi . 3ei 9Bi 3 cos 15 9B 11 rr \\ 9eell 8B 11 cos 9 cos . — =nn2(cosfffl + e,)— _ =2ei -_i t + 2(ccos 9Bi 9 COS fi 9Bi 3 COS fi £ = 2eiL _ - _ =• = 2ei _ =• 9 i 2(cos fi + ej) + 2(cos 9 COS f i The — remain to be determine1!. da (5, : Le, then e J l 9el Similarly, e 8e l l [ 9e0. Ce 9e 9C l l Ce 9e 9C l l Ce 9e 9C _ l l e 9e 9C = J l l 1 Ce 9e 9C ~ l l 16 Ce 80) 1 Ce 9C l 9el - ej2 9nl (6) |jj: Let D = fl + eg cos fgVl + ej cos fjVcos ig cos ij + sin ig sin ij cos AflJ + egej sin fg sin f j Define the intermediate terms as Dg = 1 + eg cos fg = 1 + ej cos fj = cos ig cos ij + sin ig sin ij cos then e 3D / \ l — = p>0D2 cos f ! + e0 sin fQ sin fXJ — 9e 3D / l \ / ^ei = D2 IDg cos f j + Dj cos fg + sin fg sin f j (eg -— + / \ / 3 cos fi SeiN"] Djjcos ig sin ij cos Af2 - sin ig cos iij+ D 2 le 1 — + cos fj —— I 3 COS f[ / 3 sin f i 3 sin D2D1eg ei sin fn —7:-. + sin fi ——9l +. sin fn sin fi iQ V\ ^0 O ' 3 COS f 3 COS f. 3D D 1+ D e 0 el -57^ + cos f l ^7 1 u 3o)0y ' l~O~ 3wg 3 sin fi 3 sin fg\ + en ei sin fn + sin fi — + sin fg sin f^ 17 / 8 COS fl sin in. sin ii sin Afi + ^2\ e l —arT— + cos *1 8 cos f0 8 sin fi1 8 sin fG sin ft • ^ v ) + sin f Q sin f x - = (D0D2 c°s fl + e0 sin f0 sin fj)^i a cos f an / \ = DOu Dilsin A \ in u cos ii A cos AJ2 - cos in u sin ii) + D£ ei v \ A — rj- — + cos 9i a cos f0 8 sin f 8 sin f + D D e l 2O + e O in £ + sin + sin fQ sin f! — 8il cos f 9e 9 cos f 1 O 9a) = Do Dn e, — - 8w - - + cos fi -3- 9a) + Die n —3 l \ l l/ 1 8 sin f i 8 sin fQ\ 8ei~ e0 e s i n £ 0 sin f -— + sin f 0 sin fj - 8D 8 cos fi -Dj sin i0 sin i^ sin Aft + D 2 (ei —— + cos 8^1 9 cos [0 8 sin fiJ. . f 8 sin + DlD2eO + er sin f0 + sin fi + sin fg sin f^ c l 3 Sin f.Q 3 sin The and are to be determined. 3Q! The preceding equations involve derivatives of eccentricity on the second orbit with respect to the elements on the first and second orbit. Rewriting the constraint equa- tion (16) yields + e Ocos fo = e l cos When E is defined as E = agU - CQ j cos fj + + C cos f this equation can be differentiated implicitly to obtain the following derivatives: 18 - el2) GOS fO cos ON 9 COS f 9 COS f -I = ![(!. cos f )J cos COS 9 cos / cos ffl - (l - Finally, it is necessary to obtain derivatives of equations (7), (9), (10), and (11). (This method assumes that the transfer takes place in the direction of positive R. For nega- tive R (eqs. (12) to (15)) simply interchange the initial sign on the derivatives.) Define F = 1 - (cos ig cos ij + sin ig sin ij cos By inspection of equation (7), 9 cos fg 9 cos fg 9 cos fg 9 COS fg =0 9e0 9 GO1 9 cos gfg i !,-r __._ . -, TT: = -•= sin ij sin Afi(cos ig cos ij + sin ig sin ij cos AOJ sin fg 19 9 COS ffl = sin f 9 c sf o O = — sin ij/cos ig sin ij - sin ig cos ij cos Afi) sin fg 9 cos fg i —Q-. = -pfsin ig sin A £2 sin fgj 9 COS f 9 COS f By inspection of equation (9), 9 sin fg 9 sin fg 9 sin £Q 9 sin fp 9a 9e 9a 9a) O O l l 9 sin f Q i r- , , -i ——- = -^ sin ij sin Afi ^cos IQ cos ij + sin IQ sin ij cos Afij cos IQ 9 sin fn = "COS fn ° 8 sin f i r , N -i = — sin ij (sin ij cos ig - cos ij sin IQ cos Aft) cos fQ 9 sin f 1/ sin AS2 cos 9 sin fg 9 sin fg By inspection of equation (10), 9 cos f i 9 cos fj 9 cos f j 9 cos =0 9 cos f i i / v —— - = — (sin ii sin Afi sin fj) 9i F> ' f —i = —i rsin ig (sin / ig•cos '• ij ..•••••.. - cos ig sin ij cos AOj\ sin f nj .9 COS 20 9C os ii Sf i r / N -i JT:° l = — 1C:-, sin ig „sin ; Afi (cos ig cos ij + sin ig sin i^ cos AOj sin fjl 3 COS fi —— = sin I-, 3 cos fj 3 cos fj By inspection of equation (11), 3 sin ft 3 sin fi 3 sin f i 3 sin fi L i = i = i = 0 9 sin f —-—-1 = —(sin 1 / ii sin AO cos fi)\ 3ig F^ l ij 3 sin f i i r , . N —rpr— = — sin in (sin in cos ii - cos ig sin ii cos AO) cos Ouuf\ JC I * 3 sin f i ir , \ -i —r:—=• = -—sin ig sin AO (cos ig cos ij + sin i^ sin ij cos Aflj cos fj 3 sin f - = - cos 3 sin f i 3 sin f j All these relations can be substituted into equation (24) to obtain the desired derivatives of AV. It is prohibitive to write these equations explicitly. However, they are readily programed for the computer. The analytic derivatives have been checked with good agreement against central -difference numerical derivatives. Generalization to Multiple -Impulse Transfer The multiple -impulse transfer is treated as a sequence of single -impulse transfers. The first orbit is specified by a, e, i, w, and £2, and the final orbit is specified by a, i, co, and £2, eccentricity being a dependent variable. On each intermediate conic there are four control variables (a, i, co, and fi) and one dependent parameter (e). By choosing values of the control variables, the total velocity change is the sum of the single impulses. For minimization purposes, the derivatives of total AV with respect to the 21 control variables are obtained by chaining together the derivatives obtained in the pre- vious section for the single-impulse transfer. As an example, consider a four-impulse transfer problem. Let o^ represent any of the four independent parameters of the ith intermediate orbit (i = 1,2,3). The total velocity change is the sum of the four impulses and may be written as AV = AV2 + AV3 + AV4 (25) where AV 1 =AV 1 (o! 1 ) AV = AV3 = AV3(a2,o<3,e2) AV 4 =AV 4 (a 3 ,e 3 ) and e l= e = The partial derivatives of AV-p with respect to the independent elements are 9AVT _ 9AVj + /9AV2 9AV2 9AV 3 /9e 2 9e2 ~ ^"^^^^^^ I T^"^"^~ i ' 9e2 9AV 9e 3 /9e 2 9e (26) 9eQ 9e 9AVT 9AV2 9AV3 9e 2 > 9AV 4 /9e 3 | 9e3 (27) 9d!r 9e2 9or 2 , 9e3 \9o!2 + 9e2 9AVT 9AV3 /9AV 4 9AV4 (28) 9(Xo dctn \ 9Q!o 9eo O O \ O O 22 Thus, to relate any two consecutive orbits, the following'derivatives are necessary: 9AV4,i 1+1 9AVi+1 9AV;,i 9ei.i 364.1 96; Q , n , . 1+i, -jiii, -jiti, and —i-. The derivatives are obtained from Sej 9o(i ' 9o?i+i 3ei ' a^i ' 9^ equation (24) and AV i+ j represents the velocity change required to transfer from the ith to the (i + l)th orbit. The extension of this formulation to n-impulses simply involves n - 1 equations similar to equations (25) to (28).