1. Introduction Design of Satellite Constellations for Optimal Continuous Coverage D.C. BESTE, Member, IEEE General Research Corporation Abstract A satellite-borne sensor can view a region at or above the Earth's surface. The size of this region depends on the satellite's altitude, the maximum range and scan angle of the sensor, the minimum above-the-horizon viewing angle required, the extent in altitude of the region to be viewed, and the maximum altitude of sensor obscuration by the atmosphere. Except for geosynchronous satellites this region moves relative to the Earth, so that constellations of satellites are generally necessary for continuous coverage. Satellite constellations which minimize the number of satellites required for continuous coverage are derived as a function of the angle subtended at the Earth's center by the coverage of a single satellite. This is done for single and triple continuous coverage of the entire Earth and of the polar regions extending to arbitrary latitude. Simple, cogent approximations for the configurations and numbers of satellites are found. Expressions which relate sensor capabilities and surveillance requirements to are presented. Examples are given to illustrate the use and accuracy of the results. Manuscript received September 6, 1977; revised November 28, 1977. Author's address: General Research Corporation, P.O. Box 6770, Santa Barbara, CA 93111. 0018-9251/78/0500-0466 $00.75 466 1978 IEEE In the first twenty years of the space age we have witnessed a dramatic increase in the number of satellite-based systems. We can expect this trend to continue with the advent of the Space Shuttle [ 1,2] Many of these systems will require that the entire Earth, or certain regions of the Earth, be continuously within sight of one or more satellites. For example, the NAVSTAR global positioning system (GPS) [3] requires that every user be able to observe four satellites simultaneouslv. It is of interest to know the configuration of satellites which requires the fewest satellites with sensors of given range and scan angle. Emara and Leondes [4] addressed this problem for a mission which requires quadruple coverage (i.e., GPS). Liuders [5] addressed the problem of continuous single coverage. His results are widely used [6, 7, 8]. This paper is directed toward designing optimum satellite configurations for continuous coverage of the entire Earth or of polar regions extending to arbitrary latitude. It is shown that the method developed here requires about 15 percent fewer satellites than the classical Luders configurations. It is also showin that orbital planes which have a common intersection (e.g., polar orbits) are preferred, and that the number of satellites required for continuous coverage between latitude X and the pole can be approximated by N - 4 cos X/(1- cos4') where 4iis the Earth-centered half-cone-angle of coverage for each of the N satellites (see Fig. 1). A comparison is also made with the number of satellites required if the satellites could remain "stationary." This unrealizable situation is useful for establishing an absolute lower bound on the number of satellites. Surveillance systems which employ angle-only or timedifference-of-arrival techniques require simultaneous viewing by three satellites. For this reason triple coverage is also addressed. Although the analysis is approached differently, the results are similar to those for single coverage except that approximately 24 times as many satellites are required. Triple coverage with somewhat fewer than 3 times as many satellites is in good agreement with intuition since, ideally, the additional satellites should be placed to take advantage of the overlap which occurs with single coverage.. In this paper the optimum corifiguration of satellites is first determined for continuous single and triple coverage of two polar regions extending to the latitudes ±X. Given that the sensor on each satellite can cover a circular region defined by an Earth-centered cone of half-angle 4, an expression is derived for the number of satellite orbital planes (n) and the number of satellites per plane (m) which yield the smallest total number of satellites (N = nm). After establishing the relationship between q4 and N for single and triple coverage, the relationship between 4 and the required sensor range R and maximum scan angle a is derived under a number of different constraints of interest. IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS AES-14, NO. 3 MAY 1978 SATELLITE IL. Single Coverage Two approaches to satellite constellation design were considered. The first approach was to place the satellites in orbital planes which have a common intersection (e.g., polar orbits) and to adjust the plane separation and satellite spacing so as to minimize the total number required. This approach can be considered to be an extension of the Liiders method, and leads to some very simple expressions which relate the number of satellites to the sensor require- ments. Polar orbits result in higher satellite densities at the poles than at the equator. It seerns intuitive that orbital configurations which result in a more uniform distribution of satellites over the Earth would lead to more efficient covering. Therefore the second approach was to select orbital planes which result in as uniform a distribution of satellites as possible (e.g., three mutually orthogonal obital planes). A. Single Coverage With Polar Orbits Fig. 1. Region observed by a satellite. Fig. 2. Coverage geometry. [n = orbital planes, m satellites per plane, 4, = angular coverage, 4, + A = plane separation between codirectional orbits (separation is 2A between opposite-moving = The geometry is shown in Fig. 2. The angle 4 is the Earth-centered half-cone-angle corresponding to the coverage of one satellite, or equivalently, the radius of the coverage circle (measured in angular units) on the surface of the Earth. This region will be referred to as the coverage circle of radius 4. Satellites in adjacent orbital planes move in the same direction, and the satellites of one plane are shifted relative to those of the adjacent plane by one-half of the intraorbit satellite spacing (i.e., r/rm, where m is the number of satellites per orbit). This configuration clusters the satellites in an optimal manner at the equator. At the two boundaries where adjacent orbital chains move in opposite directions, the relative geometry is not constant, so that the angular separation between orbital planes must be smaller than the angular separation between orbital planes of satellites moving in the same direction. The most demanding requirement on 4 occurs at the equator, where the following equation must be satisfied: (n-1)o + (n + )A =Tr CIRCLE OF chains).] (1) Note that in all cases the ratio is nearly equal to 2. The implication is that the coverage averaged over the entire where n is the number of orbital planes, 4 is the angular sphere is double. The result, which holds for ratios of m radius of the coverage circle, and A = cos-m [cos 4/ to n in the approximate range of 1.3 to 2.2, leads to a (cos 7r/m)] . The n orbital planes are separated in angle by simple formulation for specifying the relationship between 4 where the numbers of satellites and their orbital configurations (2) and coverage requirements. ¢,= , + A. Since the solid angle Q corresponding to the circle 4, is = 27r (1 - cos 4,), the preceding result suggests the followQ This results in an orbital-plane separation of 2A at the ing approximation for the number of satellites for a boundaries where satellites in adjacent orbital planes move 4,: given in opposite directions. Equation (1) was evaluated for a wide range of combi1.3n < m < 2.2n. (3) nations of n orbital planes and m satellites per orbital plane. N nm 4/(l Cos IP), The results are summarized in Table I, which also shows the This expression is shown as a dashed line in Fig. 3 along with orbital-plane separation 4,. The last column gives the ratio data points corresponding to the solution of (1). of the total solid angle covered by all satellites to 47r sr. = - BESTE: DESIGN OF SATELLITE CONSTELLATIONS FOR OPTIMAL CONTINUOUS COVERAGE 467 TABLE Requirements for Single Coverage of the Entire Earth n m 2 2 3 3 3 4 4 4 5 5 5 3 4 4 5 6 6 7 8 8 9 10 j (deg) 66.7 57.6 48.6 42.3 38.7 33.6 (deg) 104.5 98.4 69.3 66.1 64.3 49.4 48.3 47.6 38.6 38.1 37.7 nmE./4ir 1.81 1.86 2.03 1.95 1.97 2.00 1.97 1.99 1.98 1.97 1.99 0 30.8 28.9 25.7 24.2 23.0 NUMBER OF ORBITAL PLANES (n) 50 40 30 Fig. 4. Coverage geom,etry at latitude X. TABLE 11 Requirements for Single Coverage of Polar Regions Extending to Latitude X 20 22 N = - 4o Cos n m i (deg) p (deg) nm2'4 ir 3 4 5 64.1 53.4 48.1 39.9 35.8 33.3 28.9 26.8 25.3 22.6 111.8 103.1 98.7 68.4 66.0 64.5 38.8 1.69 1.62 1.66 1.75 1.70 1.72 1.74 1.72 1.73 1.73 49.6 43.2 39.4 30.0 111.2 104.4 100.4 66.9 65.4 64.4 49.2 48.4 47.9 47.4 1.41 1.35 1.37 1.41 1.40 1.42 1.42 1.41 1.41 1.43 X = 300 0o 20 30 40 , degrees 50 60 2 2 2 3 3 3 4 4 4 5 70 Fig. 3. Number of satellites required for single coverage of the Earth. Note that the approximation of (3) degrades gracefully for n and m outside the indicated range. For this reason it is sometimes worthwhile to try combinations of n and m outside this range. For example, if the maximum coverage of one satellite corresponds to 4 = 380, (3) yields N 18.9. The smallest value of N (a product of integers) which exceeds this and satisfies the inequality is N = 24 (n = 4, m = 6). However, it can be seen from Fig. 3 that n = 3, m = 7 (N = 21), which does not satisfy the inequality, also provides total coverage with 4 < 380. It should be concluded that (3) gives a simple, nearly exact relationship between 4' and N when the inequality between n and m is satisfied. In situations where the inequality leads to values of N much larger than that required to satisfy (3) (as in the example above), it may be worthwhile to consider combinations of m and n not satisfying the inequality. This requires a trial-and-error evaluation of (1). X The preceding analysis can be extended to coverage between latitude X and the north pole, and between latitude 468 6 7 7 8 9 9 450 2 2 2 3 3 3 4 4 4 4 - B. Full Coverage Beyond Latitude X 5 4 5 6 7 8 9 9 10 11 12 28.0 26.5 22.9 21.6 20.6 19.9 49.6 48.5 47.8 -X and the south pole. Fig. 4 shows the geometry; it is the same as in Fig. 2 except that coverage must be achieved at X degrees latitude rather than at the equator. In this case the requirement (1) becomes (n -1) + (n+ 1)A =rrcosX (4) where A is defined as in (1). Table II summarizes the evaluation of (4). Fig. 5 shows a plot of 4 versus n m for X= 0, 30, and 45 degrees. These results indicate that (3) may be generalized to IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS AES-14, NO. 3 MAY 1978 N=nm 4 cosX(l -cos4), 1.3 n <m cos X< 2.2n. 450 300 0° -5 5// NUMBER OF ' A 50 (5) The accuracy and implications of this approximation will be discussed later. = ORBITAL PLANES (n) 5 5 40 44 4 C. Nonpolar Orbits 55 4\ -j \ An alternative approach to achieving total coverage is to \ \5\ 30 [ use orbits distributed in some uniform manner. In general L. these are very difficult to analyze for arbitrary numbers of N-1 4CosA~ co \ 4\\\ 44 satellites. However, a few specific cases were individually 3 m: \ 3 \ 3\ examined and it was found that the polar-orbit configura20 [ \33 3 93 tion was superior. Two orbital planes have one intersection and are thus equivalent to the polar case described above. In that case it was 21 -z .32-J) shown that the optimum angle between satellite planes was Iu 20 30 40 50 not 900 but depended on the number of satellites per orbit p, degrees plane. The optimum angle approaches 90° only in the limit Fig. 5. Number of satellites required for single coverage beyond of an infimite number of satellites per plane. latitude X. Three mutually orthogonal orbital planes were investigated for the case of m = 4 (total of 12 satellites). The Fig. 6. Comparison of LUders and Beste results. method used was to adjust the interorbit relative phase, while maintaining the intraorbit spacing of 900 so as to 60r minimize the maximum distance from a satellite to a point in any quadrant. The resulting coverage angle 4 is 50.80. 50F This angle compares with 48.60 required for 12 satellites in 3 polar orbits as given in Table I. 1.1 40 BESTE N = 4 (E + 1) n 1 LUDERS 2 -j D. Comparison With Previous Results Luders' paper in 1961 [5] addressed inclined as well as polar orbits in which orbital planes were uniformly separated. He concluded that polar orbits were optimum. His results were given in terms of minimum satellite altitude (a monotonic function of 4) and expressed in nautical miles. A comparison of his results with those of this paper is shown in Fig. 6. Note that the method described here leads to a substantial reduction in the number of satellites (10-20 percent in most cases). A similar conclusion holds for coverage between the poles and latitudes ±X. The reason for the difference is that Luders' restriction to equally separated orbital planes leads to an unnecessarily large overlap between satellites in adjacent orbital chains which move in the same direction. -j < 30 u 20 iok n 200 400 600 800 1000 II 1200 1400 n mi ALTITUDE, 1600 1800 2000 with two great circles (corresponding to 2 satellites diametrically opposite and at infinity) where the ratio of the solid angle of coverage to 47r sr is 1. Other efficient coverings occur for satellites which are positioned according to the faces and vertices of regular polyhedra. The most efficient of these occur for 4, 6, 12, and 32 circles. In the limit as the number of circles grows to infinity, the optimum pattern is that of a hexagonal grid on a planar surface. Table III summarizes the required 4 and amount of excess coverE. Comparison With Stationary Satellite Coverage age for these configurations. Fig. 7 compares the relationship between 4 and the total Because the satellites are in motion (in their respective number of satellites obtained in this study with that of orbits) a considerable amount of overlap must be experiLuders' paper and with the stationary bound. enced in order to ensure coverage at all times. It is therefore of interest to investigate the numbers of "stationary" satellites required to cover the Earth. This represents an unIll. Triple Coverage achievable (except for 2 satellites) lower bound. Different numbers of circles cover a sphere with different There appears to be no reasonable analytical approach to efficiency. That is, some combinations of circles having a the triple coverage problem. Therefore the method used in common radius 4, cover the sphere with a smaller fraction of this study is to calculate the required 4 by using an iterative overlap. In particular, the most efficient coverage occurs search. BESTE: DESIGN OF SATELLITE CONSTELLATIONS FOR OPTIMAL CONTINUOUS COVERAGE 469 TABLE IIl Sphere Coverage by Circles Number of Circles 2 4 6 12 Radius of Coverage y (deg) 90.0 54.7 37.4 1.0 1.3 1.3 1.2 32 22.7 1.2 lim N-oo cos 1(1- 2.42/N) 1.2 Nfl/4ir 70.5 Configuration diametrically opposite faces of a tetrahedron faces of a cube faces of a dodecahedron faces and vertices of a dodecahedron (or icosahedron) hexagonal pattern on a plane surface 6Or50 V) U.j LUDERS V2 40k cos- Co~q 4lm/(n ±I+ '/4O) -j LLJ .::c Ln 30 F- 4 Li- CD cl. LLJ co M: 20k .2-1 10 0 STATIONARY COVERING BOUND 2.5 i ( l -cos; ) 10 20 30 40 50 60 70 ;, degrees Fig. 7. Comparison of number of satellites required for total Earth coverage. A basic assumption is that, as in the case of single coverage, it is desirable to keep the satellites spaced as uniformly as possible in a fixed relative orientation. The configuration is thus similar to that used for single coverage (Fig.2) except that the circles must overlap to a much greater extent. Adjacent orbital planes have equal angular separation, and satellites in adjacent planes travel in the same direction (except at the two boundaries) and are staggered by onehalf of the intraorbit separation. The required angle of coverage (4) per satellite is determined as follows: m satellites are placed uniformly in each orbital plane; n polar orbital planes are placed uniformly around the equator. The satellites in adjacent orbital planes are shifted in latitude by one-half of the intraorbit spacing (iT/m). A set of points is chosen at the equator. The set extends one-half the interorbital separation in longitude. Each point of the set is successively examined to determine the value of 4 such that three satellites are within a distance 4 of that point. The satellites are then moved along their orbits by a small fraction of their orbital spacing and the procedure is repeated. The largest value of 4 thus obtained is chosen. Satellites having this angle of coverage are necessary and sufficient for triple coverage of the points in the set; that is, points between orbital planes in which satellites move in the same direction. This value of 4i is called 4i. 470 The value of ; required for points falling between orbital planes in which satellites move in opposite directions is called 'o-. 4o is determined by using the same set of points as before. However, in this case, each point is examined repeatedly as the satellites in one adjacent orbital plane are incrementally moved along their orbital path. Qo is of course equal to or larger than 4i. For the equally spaced orbits chosen above, the required angle 4 per satellite must equal 40. This would result in excess coverage between planes in which satellites inove in the same direction. Clearly the value of 4 required for triple coverage can be reduced if the distance between adjacent codirectional orbital planes is increased at the expense of the distance between adjacent opposing orbital chains. Since the ratio of the number of intervals between oppositely moving orbital chains to the number of intervals between codirectional orbital chains is 1 /(n- 1), the first iteration solution to the optimum orbital plane separation 0 and satellite coverage 4 is = (6) [41.(n 1) + 24 o ] /(n + 1). (7) - In principle the procedure can be repeated, leading to better and better approximations to the optimum solution. This was done for a small number of cases and it was found that the results changed negligibly or not at all after the first iteration. Table IV summarizes the triple-coverage requirements in the same form as Table I for single coverage. Note that the ratio of the total solid angle of coverage of all satellites to 4mT sr remains close to 5.5. In other words, a point is in view of 5.5 satellites on the average when this average is taken over the entire Earth. Satellite requirements were also determined for triple coverage between the poles and ±X degrees latitude. The procedure was the same as for total coverage. Fig. 8 shows the number of satellites required as a function of 41 for X = 00, 450, and 600. Also shown in the figure are curves for the approximate relationship N= nm = 11 cos X/(1- cos 1), 1.4n < m cos X < 2.4n. (8) IV. Sensor Requirements In this section, the single-satellite coverage 4 is related to the range and scan angle required for a sensor on-board the satellite. Fig. 9 depicts the geometrical relation for a satellite at altitude H providing coverage of the half-cone-angle 4' at a surveillance altitude A. As can be seen from the figure, the satellite's altitude can be increased or decreased to provide various combinations of required sensor range R and scan angle o. As H is increased the sensor's maximum scan angle decreases at the expense of a longer required sensor range R. Although there is no upper limit to the satellite altitude, IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS AES-14, NO. 3 MAY 1978 TABLE IV Requirements for Triple Coverage of the Entire Earth n 3 3 3 3 m 4 5 6 7 ;p (deg) 80.7 70.3 63.9 61.1 4 4 5 5 5 6 6 7 8 8 9 10 52.2 6 4 9 10 X = 600 450 48.3 43.7 41.1 38.8 37.5 35.8 45.4 36.9 36.6 36.2 5.04 (A a) (1) I 0) 0) -0 5.42 5.36 5.42 5.35 5.54 45.9 d LU -j 30.8 30.5 \6 4\4 c<: 40 3 3 3 \6 -r- m L.) V) C) C) V) 2-1 Li L/) 4 40 Fig. 10. Sensor range and scan angle as a function of satellite altitude. Using the law of cosines and law of sines, the following expressions can be derived for the required sensor range and scan angle as a function of the satellite altitude H: \5 4 \, 3 y, 4 50 60 60 50 40 30 70 70 R-= [(Ref R= (Re+ A)2 [(Re+IHI2 ++(+X) degreess -2(Re + H) (Re + A) cos ]1/2 Fig. 8. Number of satellites required for triple coverage. a Fig. 9. Sensor range and 30 20 10 SATELLITE HEIGHT (H), 1000 km 0 5 \ v, _ scan sin -I {[(R + A)/R ] sin (9) 4} (10) where H> SCAN ANGLE, ,x REQUIRED SENSOR SATELLITE HEIGHT, H = angle requirements. SATELLITE MINIMUM cY- 5\ 3 20 C-1 m .:x Ci NUMBER OF ORBITAL -, PLANES (n) 5\ \4 4 )k CD .u 5.55 5.52 5.59 5.66 \5 50 l 30 46.9 - 8 SATELLITES 15 SATELLITES 32 SATELLITES 5.03 4.98 NN_- 1 1 cos X 1 - cos ) 6 v) E 56.4 -. nmQ2/47r (deg) 64.5 62.3 60.3 60.0 - 0O 60 F L-1 0 -- = [(Re +A min )/cos(4i-cos' e)] Re (11) and RANGE, R = SATELLITE\ HEIGHT, HMIN SURVEILLANCE MINIMUM / PENETRATION ALTITUDE, Hmin ALTITUDE, A AMI N (Re + A min)/(Re +A). The required sensor range and scan angle are shown in Fig. 10 as functions of satellite altitude for single coverage of the entire Earth's surface with 8, 15, and 32 satellites 600, 42.80, and 290, re[from (3) this corresponds to spectively] Equations (9) through (11) provide a means of trading off between sensor range and sensor scan angle. In some applications it is desirable to minimize the sensor range without regard to scan angle; in these situations the required sensor range is given by . EARTH R ~~~~~~SURFACE EARTH CENTER there is a lower limit Hmin below which the satellite loses its line-of-sight to the desired surveillance altitude A. The line-of-sight can be obstructed either by the Earth's surface or by the atmosphere near the Earth's surface. The dashed line in Fig. 9 shows the minimum satellite altitude required to ensure a line-of-sight above Am in. R (Re + A) sin 41, cos >a cos <R. = (Re +Hmin )Sin 4, (12) The corresponding satellite altitude at which the required sensor range is minimized is as follows: BESTE: DESIGN OF SATELLITE CONSTELLATIONS FOR OPTIMAL CONTINUOUS COVERAGE 471 cc ot -z Cu> 6 4 Fig. 11. Sensor range and scan 12 8 10 SATELLITE ALTITUDE, 1000 km angle requirements for global 14 coverage 16 18 20 (example 2). TABLE V System Characteristics for Minimum-Range Single Coverage With 15 Satellites (3). This yields 14.8, which when rounded up to the nearest product of the integers (n = 3, m = 5), results in N = 15. Therefore three orbital planes of five satellites each are reUsing Using quired. Approximate Q, Exact Q Parameter For N = 15 the exact value of according to Table I is 5930 5820 Sensor range (R) in kilometers 42.30 and the approximate value according to (3) is 47.7 47.2 Sensor scan angle (cr) in degrees = 42.80. Using these exact and approximate values of 4, 2330 2250 Satellite altitude (H) in kilometers 66.1 67.8 Orbital plane separation (¢) in degrees the various system characteristics shown in Table V were 42.8 42.3 Circle of coverage (V) in degrees derived using (2), (9), (10), and (13). From the table it can be concluded that for most analysis purposes the approximate value of is adequate. Example 2: The sensor requirements for single coverage (Re+A)cos -Re, cos > GS (13) of an altitude region extending from 20 to 900 km over the H= entire Earth, with 8 and 15 satellites, is determined. The H in. Cos <.R variation of these requirements as satellite altitude is increased is also shown. V. Examples The approximate half-cone-angle of satellite coverage for each satellite is found by inverting (3), using N = 8, 15; To illustrate the preceding results examples are given 600 and 42.8°, which use the approximate relationship (5) between and A = 0;Re = 6400. The results are respectively. These compare with exact values of 57.60 and N. For comparison the equivalent results will also be ob42.30, respectively, from Table I. The sensor range and scan tained from the more complex exact relationship (1). Example 1: Given a maximum sensor range of 6000 km, angles are found by using (9) and (10). These results are the satellite constellation which provides single coverage of shown in Fig. 11 (the curves based on exact and approximate values for are nearly coincident). In this figure the the entire Earth with the minimum number of satellites is sensor scan angles are shown for coverage at 20 km and 900 found. The required number of satellites is obtained by first km; the required sensor range is essentially the same for both using (12) to find for A =Amin = 0, Re = 6400 km, and altitudes. The effect of a constraint corresponding to atR = 6000 km and then substituting the resulting value into mospheric obscuration below 20 km is also indicated. = 472 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS AES-14, NO. 3 MAY 1978 References [1 [2] [3] [4] 1. Bekey, and H. Mayer, "1980-2000 raising our sights for advanced space systems," AIAA Astronaut. Aeronaut., vol. 14, pp. 34-63, July 1977. "Outlook for space," Superintendent of Documents, Government Printing Office, Washington, D.C., NASA Special Publ. SP386, Jan. 1976. D. Smith and W. Criss, "GPS Navstar global positioning system," AIAA Astronaut. Aeronaut., vol. 14, pp. 26-32, Apr. 1976. C.T. Leondes and H.E. Emara, "Minimum number of satellites for three-dimensional continuous worldwide coverage," [51 [61 [7] [8] IEEE Trans. A erosp. Electron. Syst., vol. AES-1 3, pp. 108ll1,Mar. 1977. R.D. Luders, "Satellite networks for continuous zonal coverage," ARS J., vol. 31, pp. 179-184, Feb. 1961. E.D. Harney, Spaee Planners Guide, USAF Systems Command, U.S. Government Printing Office, pubi. 0-774405, 1965. J. Jenson, G. Townsend, J. Kork, and D. Kraft, Design Guide to Orbital Flight. New York: McGraw-Hill, 1962. D.C. Beste, "Design of satellite constellations for optimal continuous coverage," General Res. Corp., Santa Barbara, Calif., Rep. RM-2074, May 1977. David C. Beste (S'60-M'65) received the B.S. and M.S. degrees in physics and mathematics from the University of Michigan, and the Ph.D. degree in electrical engineering from the University of Southern California in 1970. He is presently with General Research Corporation, Santa Barbara, Calif. His professional interests include radar and communication systems, advanced radar concepts, extraordinary space-based systems, detection and tracking, and electronic countermeasures. Dr. Beste is a member of Sigma Xi and Eta Kappa Nu. BESTE: DESIGN OF SATELLITE CONSTELLATIONS FOR OPTIMAL CONTINUOUS COVERAGE 473