You are on page 1of 6

ICROS-SICE International Joint Conference 2009 August 18-21, 2009, Fukuoka International Congress Center, Japan

Constrained Trajectory Optimization for Lunar Landing during the Powered Descent Phase
Bong-Gyun Park1, Daekyu Sang , and Min-Jea Tahk
1

Department of Aerospace, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea (Tel : +82-42-350-5758; E-mail: bgpark@fdcl.kaist.ac.kr) 2 Department of Aerospace, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea (Tel : +82-42-350-3758; E-mail: dksang@fdcl.kaist.ac.kr) 3 Department of Aerospace, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea (Tel : +82-42-350-3718; E-mail: mjtahk@fdcl.kaist.ac.kr) Abstract: To design the more accurate trajectory of a soft lunar landing, the constraints on the powered descent sub-phase, such as a breaking phase, an approach phase, a terminal descent phase, have to be considered. In this paper, the trajectory optimization of the lunar landing was performed considering constraints on the sub-phase of the powered descent phase. To convert the optimal control problem, the Legendre pseudospetral(PS) method was used and C code for Feasible Sequential Quadratic Programming(CFSQP) was used as the NLP Solver. Keywords: Legendre pseudospectral method, Hohmann transfer, Powered descent phase

1. INTRODUCTION
The Lunar landing consists of three phases: a de-orbit phase, a coast phase and a powered descent phase. The powered descent phase is also subdivided into three phases, such as a breaking phase, an approach phase, and a terminal descent phase. The lunar landers trajectory has to be designed considering constraints on the powered descent phase for the more accurate trajectory because the required constraints, such as a throttle limit, attitude, and velocities, on each sub-phase of the powered descent phase are different. Thus, in this paper, the trajectory optimization for lunar landing was performed considering the constraints on the sub-phases of the powered descent phase. The performance index is to minimize the landers fuel consumption at each sub-phase of the powered descent phase. And the de-orbit phase and the coast phase were assumed as the Hohmann transfer, which allows a transfer between the parking orbit and the powered descent phase with the minimum fuel consumption. To convert the optimal control problem into the parameter optimization problem, the Legendre pseudospectral(PS) method, which has been widely used for the trajectory optimization in recent years, was used and C code for Feasible Sequential Quadratic Programming(CFSQP) was used as the Nonlinear programming(NLP) solver. The following section will define the optimal control problem for lunar landing and the required constraints during the powered descent phase. Next, the optimization method used to solve the optimal control problem will be described. Finally, the constrained trajectory for the soft lunar landing will be compared with the base trajectory without the attitude dynamics and the trajectory with the only final attitude constraint.

2. PROBLEM DEFINITION
2.1 Assumptions The Moon is assumed as perfect sphericity and has no atmosphere. The rotation of the Moon is negligible because the landing site is not considered. The lunar equatorial radius req is 1737.4 km and the lunar gravitational parameter is 4902.78 km/s. 2.2 Equations of motion In Fig. 1, the lunar lander is assumed as a mass point and the motion is planar motion. The dynamic equations considering the landers attitude can be derived as follows in polar coordinates [1].

r=v

u=

uv Tmax k + cos r m u2 T k v= 2 + max sin r r m Tmax k m= I SP g o

u r

(1)

= =

and the control radial distance, horizontal velocity, landers mass,

The states of the motion are r , , u , v , m , , inputs are , k where r is the is the center angle, u is the v is the vertical velocity, m is the is the thrust angle, is the

angular rate, and is the angular acceleration, Tmax

- 4226 -

PR0002/09/0000-4226 400 2009 SICE

maximum thrust, k is the throttle command, I sp is the specific impulse, and g 0 is the Earths gravity.

Thus, in this paper, the constraints of the terminal descent phase are considered for the soft landing and the knots are applied to the between braking/approach phase and terminal descent phase.

3. OPTIMIZATION METHOD
For solving optimal control proems, there are two major categories: indirect and direct methods. In the indirect method, necessary conditions are used to obtain the optimal trajectory. But the direct method obtains the optimal trajectory minimizing directly the performance index with the control input or the state and the control input simultaneously. In this paper, to convert the optimal control problem for the constrained lunar landing trajectory into a parameter optimization problem, the Legendre PS method as a direct method was used because the direct method is more useful to apply to the optimal control problem with the bounded control inputs and the complex system than the indirect method. The nonlinear programming problem was solved by CFSQP. 3.1 Legendre pseudospetral method Polynomials used in the PS Method are globally interpolating Lagrange polynomials which are obtained from the orthogonal Legendre polynomials. And the optimal control problem using the Legendre PS method is transformed to an NLP problem for the values of the states and the controls at the Legendre-Gauss-Lobatto(LGL) points. The LGL points are the zeros of the derivative of the Legendre polynomial. The optimal control problem can be approximated by the following NLP problem [3].

Fig. 1 Polar coordinate of a lunar landing At each sub-phase of the powered descent phase, the performance index is defined to minimize the landers fuel consumption. And the angular acceleration command with the weight W is included in the performance index because the one is seen to fluctuate, which causes the landers oscillation and the fuel consumption increase for the real system.

min J = m ( t0 ) m t f +

( )

tf t0

W (t )

dt

(2)

2.3 Constraints during the powered descent phase There are different constraints on three separate phases of the powered descent phase. First, the breaking phase needs to reduce the landers velocity maintaining the full throttle. Second, the landers attitude needs to be constant to scan landing area for hazards and to allow time for interpreting sensor scan information during the approach phase. In the final phase, the landers attitude is vertical on the lunar surface and the vertical velocity is about -1 m/s and the horizontal velocity is almost zero. Especially, the terminal descent phase is the most important for the successful lunar exploration missions.

J N ( a, b ) = ( aN , f ) +
subject to

f 0
2

N l =0

L ( ak ,bk ) wk (3)

f 0
2

f ( ak , bk )

N l =0

Dkl al = 0, l, k = 0,..., N
k = 0,..., N

(4) (5) (6) (7)

g ( bk ) 0,

0 ( a0 , 0 ) = 0

f ( a f , f ) = 0

where, al , bl , ck , weight wk , and differentiation matrix Dkl are defined as follows.

al := x ( tt ) , bl := u ( tl )

(8) (9)

Fig. 2 Lunar landing sub-phase during the powered descent phase

ck = xN ( tk ) =

N l =0

Dkl al

- 4227 -

wk :=

2 N ( N +1)

1 LN ( tk )
LN ( tk ) 1 LN ( tl ) tk tl
2

, k = 0,1,..., N (10)
k l k =l =0 k =l = N otherwise
(11)

V =

2 r1 r2 ( r1 + r2 ) r2 V gI sp

(12)

m = m0 1 exp
p = a3

(13)

D := [ Dkl ] =

N ( N +1)

4 N ( N +1) 4 0

(14)

3.2 PS knotting method When the solution in the optimal control problem is piecewise smooth with nonsmooth junction point such as the launch vehicles mass change at each stage, PS knotting method is useful. Knots can be defined at the double LGL points when the time interval is divided into subinterval at the discontinuous controls or state values. At the knots, the continuous state and control input have event equality constraints [4, 5].

phase, r2 is that of the parking orbit, m0 is the initial mass, and a is the semimajor axis. Using the Eq. (12)-(14), V = 19.40 m/s , m = 3.943 kg , and p = 3412 sec are computed respectively. Thus, by above the value, the trajectory to minimize the fuel consumption of the transfer phase can be designed as follows.

where, r1 is the altitude which the powered descent

100

Apolune(Parking Orbit)

80 Altitude (km)

Transfer Orbit

60

40

Perilune(15.24 km)

20

1000

2000 3000 4000 Downrange (km)

5000

6000

Fig. 4 Transfer phase trajectory


Fig. 3 Knot Definition 4.2 Powered descent phase The powered descent phase starts from the altitude of 15.24 km. If the altitude of the powered descent phase starting is above 15.24km, the fuel consumption is increased. And if the altitude is below 15.24 km, considering the lunar terrain and the guidance errors is difficult [6]. At the braking/approach phase, the landers initial mass was defined by the difference of the mass using to move the transfer orbit and the initial horizontal velocity was determined by the velocity of the perilune of the Hohmann Transfer. At each phase, the boundary conditions are as follows [7,8,9]. At the terminal descent phase, the horizontal velocity has to be almost zero and the vertical velocity has to be about -1 m/s for a soft landing. Because the thrust angle is the same as the

4. CONSTRAINED TRAJETORY OPTIMIZATION


4.1 De-orbit and transfer phase The trajectory of the de-orbit and transfer phase from the parking orbit was assumed as the Hohmann Transfer which minimizes the fuel consumption and the landers specification in the parking orbit is as follows. Table 1 Landers specification in the parking orbit

Initial Mass(m) Thrust(N) Specific Impulse(Isp)

600 kg 1,700 N 316 sec

Velocity increment V , fuel consumption m , and transfer time p which the lander needs to move from the parking orbit to the powered descent phase were computed by using the following equation [2].

landers attitude, the one has to be 90 for the vertical landing. At the final phase, the angular rate has to be zero to not tip. The throttle was limited from 0.3 to 1.0 to protect engine off.

- 4228 -

Table 2 Initial constraints at each phase Braking & Approach Terminal descent

Table 4 Fuel Consumption Fuel Consumption Constrained Attitude No Constrained Attitude No Attitude dynamics 277.45 263.90 261.07 kg kg kg

r ( ti )

( ti )
u ( ti ) v ( ti ) m ( ti )

1752.64 km (15.24 km) 0 deg 1692.2 m/s 0 m/s 596.26 kg 180 deg 0 deg/sec

1737.446 km (0.03 km) free 0 m/sec -1 m/sec Free 90 deg free

( ti ) ( ti )

Table 3 Final constraints at each phase Braking & Approach Terminal descent 1737.4 km (0 km) free 0 m/sec -1 m/sec free 90 deg 0 deg/sec

) (t ) u (t ) v (t ) m (t ) (t ) (t )
f f f f f f

r (t f

1737.43 km (0.03 km) free 0 m/sec -1 m/sec free 90 deg free

Fig. 6 shows trajectories as time goes by and the landing time took about 586 sec. The trajectory of the constrained attitude took more about 88sec than the other trajectories because the vertical velocity needed almost -1 m/s and the horizontal velocity needed almost zero during the terminal vertical landing. Fig. 7 shows the trajectories as the downrange computed by the center angle and the lander moves about 439.56 km along the lunar surface. Fig. 8 and 9 show the horizontal velocity and the vertical velocity and the final constraints were satisfied at each phase. At the terminal descent phase, the landers attitude maintained about 90s and the angular rate was zero to not tip when touchdown. In fig. 8-11, the result that the lander landed softly to the ground can be seen. And fig. 12-13 show the control input profiles respectively.
16 14 12 Altitude(km) 10 8 6 4 2 0 0 No Attitude Dynamics No Constrained Attitude Constrained Attitude 100 200 300 Time(sec) 400 500 600

10 LGL points were used at each phase when the constraints on the attitude were considered. And other results were computed using 20 LGL points.

Fig. 6 Altitude vs. Time


16 14

Fig. 5 Nodes and knot allocation The simulations were performed under three different conditions. In first case, the attitude dynamics was not considered. In second case, the landers attitude was only considered at the final condition. In third case, the constraints on the landers attitude for the soft vertical landing at each sub-phase of the powered descent phase were considered. The fuel consumption computed by CFSQP is as follows. The more constraints are considered, the more fuel consumption is increased.
Altitude(km)

12 10 8 6 4 2 0 0 No Attitude Dynamics No Constrained Attitude Constrained Attitude 100 200 300 Downrange(km) 400 500

Fig. 7 Altitude vs. Downrange

- 4229 -

1800 1600 Horizontal Velocity(m/s) 1400 1200 1000 800 600 400 200 0 0 100 200

No Attitude Dynamics No Constrained Attitude Constrained Attitude Angular Rate(deg/s)

0.5 0 -0.5 -1 -1.5 -2 -2.5

No Constrained Attitude Constrained Attitude 0 100 200 300 Time(sec) 400 500 600

300 Time(sec)

400

500

600

Fig. 8 Horizontal Velocity vs. Time


10 0 -10 -20 -30 -40 -50 -60 -70 0 100 200 300 Time(sec) 400 500 600 No Attitude Dynamics No Constrained Attitude Constrained Attitude Angular Acceleration(deg/s )
2

Fig. 11 Angular Rate vs. Time


0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 0 No Constrained Attitude Constrained Attitude 100 200 300 Time(sec) 400 500 600

Vertical Velocity(m/s)

Fig. 9 Vertical Velocity vs. Time


200 180 160 140 120 0.4 100 0.2 80 0 100 200 300 Time(sec) 400 500 600 0 Throttle 0.8 0.6

Fig. 12 Angular Acceleration vs. Time

Thrust Direction Angle(deg)

No Attitude Dynamics No Constrained Attitude Constrained Attitude

1.2 1

No Attitude Dynamics No Constrained Attitude Constrained Attitude

100

200

Fig. 10 Thrust Direction Angle vs. Time

300 Time(sec)

400

500

600

Fig. 13 Throttle vs. Time

- 4230 -

5. CONCLUSION AND FUTURE WORK


In this paper, the lunar landing trajectory optimization was performed with the constraints on the sub-phace of the powered descent phase. And the de-orbit and transfer phase were assumed as the Hohmman transfer. The simulation results considering the constraints during the powered descent phase were compared with the trajectories not including specific constraints. When the constraints on the landers attitude were considered at each sub-phase, the fuel consumption was largest and the landing took the most time. The analysis of the landers fuel consumption will provide information to develop the more efficient lunar lander. And the trajectory and control histories obtained in this paper will be used to design the lunar landing trajectory for the specific missions, such as a search for the landing site, hazard avoidance etc., and the more accurate trajectory.

Moen, Lunar Module Descent Mission Design, AIAA/AAS Astrodynamics Specialist Conference and Exhibi, 2008.

ACKNOWLEDGEMENT
This research was supported by NSL(National Space Lab) program through the Korea Science and Engineering Foundation funded by the Ministry of Education, Science and Technology (S10801000123-08A0100-12310 )

REFERENCES
[1] A. M. Hawkins, Constrained Trajectory Optimization of a Soft Lunar Landing from a Parking Orbit, Masters thesis, Massachustts Institute of Technology, 2005. [2] M. J. Sidi, Spacecraft Dynamics & Control-A Practical Engineering Approach, Cambrige University Press, 1997. [3] F. Fahroo and I. M. Ross, Costate Estimation by a Legendre Pseudospectral Method, Journal of Guidance, Control and Dynamics, AIAA, Vol. 24, No. 2, 2001, pp. 270-277. [4] I. M. Ross and F. Fahroo, Pseudospectral Knotting Methods for Solving Optimal Control Problems, Journal of Guidance, Control and Dynamics, AIAA, Vol. 27, No. 2, 2004, pp. 397-405. [5] Q. Gong, F. Fahroo, and I. M Ross, Spectral Algorithm for Pseudospectal Methods in Optimal Control, Journal of Guidance, Control and Dynamics, AIAA, Vol. 31, No. 3, 2008, pp. 460-471. [6] F. V. Bennet, and T. G. Price, Study of Powered Descent Trajectories for Manned Lunar Landing, Project Apollo Working Paper 108, NASA Manned Spacecraft Center, Houston, TX, 1963. [7] S. H. Jr, Parametric Study of Lunar Landing Techniques Using Predetermined Thrust Orientation, NASA CR-61075, 1965. [8] F. Bennett, Apollo Experience report Mission Planning for the Lunar Module Descent and Ascent. NASA TN D-6846, 1972. [9] A. W. Wilhite, J. Wagner, R. Tolson, and M. M.

- 4231 -

You might also like