You are on page 1of 32

SPACE SCIENCE AND GEOSPATIAL INSTITUTE

WITH
ADDIS ABABA SCIENCE AND TECHNOLOGY UNIVERSITY

DEPARTMENT OF AEROSPACE ENGINEERING

Reviews On Optimization of Rocket's Trajectory for Maximum


Payload Delivery to Low Earth Orbit
By:
ZEWDU TEKILU HAILU (GSR0439/15)

Submitted to: Samson Mekbib (Ph.D.)


Submission Date: June 27, 2023

Addis Ababa, Ethiopia


Reviews On Optimization of Rocket's Trajectory for Maximum
Payload Delivery to Low Earth Orbit

Z.T. Hailu

Aerospace Department, Addis Ababa Science and Technology University, Addis


Ababa, Ethiopia

Abstract

Rocket's trajectory optimization for maximum payload delivery is a critical research area in the
field of aerospace engineering. This review paper aims to comprehensively analyze the current
state of research in this domain, focusing on the key issues, methodologies, and findings. The
primary research problem addressed is the optimization of rocket trajectories to maximize the
delivery of payloads to desired orbits while considering various constraints, such as fuel limitations
and atmospheric conditions. The paper reviews different optimization algorithms, including
dynamic programming, particle swarm optimization, and gradient-based methods, employed for
trajectory planning. The main findings highlight the effectiveness of these algorithms in improving
payload delivery efficiency specifically to the low-earth orbit and reducing fuel consumption.
However, a research gap exists in the integration of real-time adaptive optimization techniques to
account for dynamic changes during flight. Generally, this review paper presents an overview of
the current state of research in rocket trajectory optimization and identifies areas for future
investigation, emphasizing the need for advancements in adaptive optimization strategies for real-
time trajectory planning.

Keywords: Trajectory optimization, payload delivery, dynamic programming, low-earth orbit,


adaptive optimization

i
List of Acronyms and Abbreviations

DP: dynamic programming


ELVs: expandable launch vehicles
GBO: gradient-based optimization
LEO: low-earth orbit
MADS: mesh adaptive direct search
NLP: non-linear programming
RLVs: reusable launch vehicles
SQP: sequential quadratic programming
TOP: trajectory optimization problem
TPBVP: two-point boundary value problem

ii
Table of Contents Page
Abstract ............................................................................................................................................ i

List of Acronyms and Abbreviations .............................................................................................. ii

1. Introduction ............................................................................................................................. 1

2. Background on Rockets and Trajectory Optimization ............................................................ 4

3.1 History of Rocket Technology ......................................................................................... 4


3.2 Types of rockets used for payload delivery to low-earth orbit ......................................... 8
3. Trajectory optimization techniques ....................................................................................... 12

3.3 Optimal control theory and Approaches......................................................................... 12


3.1.1 Gradient-based optimization ................................................................................... 14
3.1.2 Convexification-based optimization ....................................................................... 17
3.1.3 Dynamic programming-based optimization ........................................................... 20
3.1.4 Derivative-free (heuristic-based) optimization ....................................................... 21
4. Discussion and Recommendation ......................................................................................... 23

5. Conclusion ............................................................................................................................ 25

References ..................................................................................................................................... 26

iii
1. Introduction

The optimization of the trajectory is crucial for the outcome of any flight vehicle design and has
become the main concern recently[1]. Thanks to computers made it easier the efficient
computation of spacecraft trajectories and as those computers got better, we could solve more
complex problems related to space exploration[2]. An enormous number of studies have been
conducted on the optimization of different space vehicle trajectories to develop an ideal trajectory
that meets the mission objectives and optimizes a particular goal in a reliable and timely manner.
The main focus is frequently to reduce mission cost or fuel consumption, mission time, the number
of asteroids visited, or a combination of these goals. [3]. To optimize the given objectives, different
types of complex optimal control methods like direct and indirect methods, genetic algorithms, or
other kinds of numerical optimization methods have been used by many researchers as depicted
below in the next section in an area of space vehicle trajectory[4]. This challenge of spacecraft
trajectory optimization can also be defined as the identification of a trajectory that meets certain
criteria, such as initial and terminal circumstances[5]. Significant progress has been achieved in
recent years in the advancement of algorithms for determining optimal paths for numerous space
missions. Each phase in spacecraft trajectory design can be classified by the elements that
comprised obtaining an ideal trajectory solution, such as the mathematical model, objective,
technique used, or, more importantly, the methodology, and algorithm[5].
As mentioned by Betts [6], the first significant attempt to categorize approaches that took into
account two well-known ways, named direct and indirect methods, and the primarily associated
techniques in each were summarized. Direct methods aim to find an optimal solution by directly
parameterizing the trajectory and then optimizing the parameters. These methods are also known
as "direct transcription" methods which involve the following main steps:

1. Discretization: The trajectory is divided into a discrete set of segments, such as time
intervals or spatial elements. Each segment is associated with a set of decision variables,
such as position, velocity, and control inputs (thrust, orientation).
2. Parameterization: The continuous trajectory is approximated by representing the state
and control variables at discrete points within each segment. Polynomial approximations,
such as piecewise linear or polynomial splines, are commonly used[7].

1
3. Formulating the Optimization Problem: The problem is formulated as an optimization
program, where the objective function and constraints are defined based on the mission
requirements, such as fuel consumption, time of flight, or reaching a specific target.
4. Solving the Optimization Problem: Numerical optimization techniques, such as
nonlinear programming or direct collocation methods, are employed to solve the
formulated optimization problem and determine the optimal values for the decision
variables.
5. Trajectory Refinement: The obtained solution is often refined by iteratively improving
the discretization and parameterization until a satisfactory solution is achieved.

Direct methods are computationally efficient and can handle complex trajectory optimization
problems. However, they may lack global optimality guarantees and can be sensitive to the initial
guess and discretization choices. Hur et al. [8] developed a direct dynamic simulation trajectory
optimization approach that aims to efficiently solve the Trajectory Optimization Problem (TOP)
by transcribing it into a Nonlinear Programming problem (NLP) without dynamic equality
constraints and intermediate states in its design variables. This is achieved by using the Hermite
spline to interpolate the continuous controls and integrate the system dynamics with corresponding
control parameters in advance[9]. The resulting NLP has a much smaller size than that of the
traditional direct method[6], which typically includes all state and control parameters defined at
each collocation point.

In addition to the direct method, there is an indirect method that focuses on finding the optimal
control inputs that satisfy the necessary conditions for an optimal solution. These methods exploit
the principles of the calculus of variations[10]. The key steps involved in indirect methods are as
follows:

1. Formulating the Optimal Control Problem: The problem is formulated as an optimal


control problem by defining the dynamics of the rocket, its objective function, and any
constraints on the state and control variables.
2. Deriving Necessary Conditions: By applying the calculus of variations, a set of necessary
conditions for an optimal solution is derived. These conditions typically involve the
Hamiltonian function, which combines the system dynamics, the cost function, and the
costate variables. The costate variables are the solution to a differential equation derived

2
from the Lagrangian equation of a system. These differential equations represent the
sensitivity of the objective function to small perturbations in the control inputs.
3. Solving the Boundary Value Problem: The necessary conditions lead to a set of
differential equations, known as the two-point boundary value problem (TPBVP). The
TPBVP involves solving both the state equations and the costate equations simultaneously.
4. Obtaining the Optimal Solution: Numerical techniques, such as shooting methods or
multiple shooting methods, are employed to solve the TPBVP and obtain the optimal
control inputs and corresponding state trajectory.

Indirect methods guarantee optimality under certain assumptions and can provide insights into the
system’s behavior. However, they can be computationally intensive and may face challenges in
handling complex constraints or high-dimensional systems.

According to Stryk and Bulirsch [10], both direct and indirect methods have their strengths and
limitations, and the choice between them depends on the specific requirements of the mission,
computational resources, and desired level of optimality. The article discusses a combination of
direct and indirect methods for trajectory optimization. The direct method involves discretizing
the time interval and approximating the control and state variables at each time step using
piecewise linear and piecewise cubic functions, respectively. The resulting nonlinear program is
then solved using optimization techniques. The indirect method involves using necessary
conditions, such as the Pontryagin maximum principle, to derive a set of differential equations that
describe the optimal trajectory. These equations are then solved using numerical techniques, such
as the multiple-shooting method. The article suggests that a combination of these methods can be
effective for solving nonlinear optimal control problems.

As it is mentioned by Conway[11], another significant addition to the numerical techniques used


in dynamical systems was identified. It gave a good review of several methodologies, akin to Betts'
survey, as well as practical examples. However, the dynamical systems examined in this survey
are general. Additional endeavors are undertaken, however, They are restricted to specific space
missions, such as Earth-Moon trajectories[12], space rendezvous[13], and libration sites[14].
These studies give various categories for spacecraft trajectory optimization problems. This study
proposes such a strategy, which is thought to be complimentary to all previously published survey
studies on the subject. It highlights recent studies done before while also offering a direction for

3
the overall procedure of the optimization of spacecraft trajectory by paying big attention to the
maximum payload delivery.

Shown in Fig.1 is an illustration of the entire process of optimizing a spacecraft trajectory that has
a specific mission goal, mission requirements, anticipated accuracy, intended convergence, and
mission plan may be separated into four parts[5]. This broad process includes mathematical
modeling of system dynamics, identifying acceptable objectives, devising an approach, and finally
realizing the solution.

Identifying developing Realizing


Modeling
objective approach solution

Figure 1: Spacecraft trajectory optimization process

2. Background on Rockets and Trajectory Optimization

3.1 History of Rocket Technology

As mentioned in [15], rocket technology has a long and fascinating history, with roots dating back
to ancient times. However, it wasn't until the 20𝑡ℎ century that rockets began to be developed for
the specific purpose of space exploration. The beginning of rocket technology can be traced back
to China in the 9th century. It was here that the invention of gunpowder was made, and early
versions of rockets were developed for use in warfare. These early rockets were simple and
unguided, but they served as a basis for later developments in rocket technology.

Figure 2: Early Chinese fire arrows

4
In the early 20th century, many inventors and scientists began to experiment with rockets for a
variety of purposes. However, it wasn't until 1926 that the first liquid-fueled rocket was
successfully launched by Dr. Goddard. This breakthrough paved the way for further developments
in rocket technology, and within a few years, rockets were being used in a variety of applications,
including military missiles. During World War II, rocket technology was advanced significantly,
with the development of ballistic missiles by both the Axis powers and the Allies. These missiles
were capable of traveling long distances and could carry heavy payloads, making them a
significant military asset.

Figure 3: Early Rockets - Dr. Goddard's 1926 rocket configuration

Image source:[16]

5
Dr. Robert H. Goddard's rocket design from 1926 marked a significant milestone in the
development of rocketry. While the specifics of the design were not publicly disclosed at the time,
some information about the components used in his rockets has been documented. It's important
to note that the technology and knowledge available in 1926 were significantly different from
today's advancements. As it is shown in the above figure, Dr. Goddard's 1926 rocket configuration
consisted of various components that were crucial to its operation.
The table below shows the function of each component of Dr. Goddard's 1926 rocket.
Table 1: Different components of Dr. Goddard's 1926 rocket and their function

Components Function
Igniter Used to initiate the combustion process.
Needle valves Responsible for regulating the flow of fuel and oxidizer into the
rocket motor.
Rocket motor The main source of propulsion, and was powered by a combination
of liquid oxygen and gasoline.
Liquid oxygen line Deliver the oxidizer to the rocket motor.
Gasoline line Deliver the fuel to the rocket motor.
Exhaust shield Used to protect the rocket from the intense heat generated during
flight.
Pressure relief vent helped to regulate the internal pressure of the rocket.
Pull cord This was used to activate the igniter to start the rocket
Liquid oxygen tank used to store liquid oxygen
Cork float valves used to control the flow of liquid oxygen from the tank to the rocket
engine
Alcohol burner It is the source of heat for the rocket engine which heat up and
vaporize alcohol fuel it would then mix with liquid oxygen before
being ignited.
Oxygen gas pressure This was used to supply oxygen gas to the alcohol burner
line
Check valve used to ensure that the flow of liquid oxygen was only going in one
direction.
Detachable starting used to deliver the fuel and oxidizer to the rocket motor during the
hose start

While Dr. Hoddard's rocket configuration was an important milestone in the development of rocket
technology, it has since been surpassed by more advanced designs. Modern rockets feature more
efficient engines, advanced materials, and sophisticated guidance systems, making them more
reliable and capable than ever before.

6
After World War II, rocket technology continued to develop rapidly. In 1947, the US Air Force
launched a rocket called the X-1, which was the first aircraft to break the sound barrier. This
achievement marked a significant milestone in the development of rocket technology and paved
the way for further advancements.
The US-Soviet space race began in earnest in the 1950s and 1960s. The Soviet Union launched
Sputnik 1, the first artificial satellite, into orbit in 1957 [17]. This achievement was followed by a
series of firsts, including the first human spaceflight by Soviet cosmonaut Yuri Gagarin in 1961
and the first American astronaut, Alan Shepard, to travel into space in 1961.
However, the crowning achievement of the space race was the first human landing on the moon in
1969. The Apollo 11 mission, led by NASA, saw astronauts Neil Armstrong and Edwin "Buzz"
Aldrin become the first humans to set foot on the moon's surface[16]. This event was a watershed
moment in the history of space exploration and demonstrated the incredible power of rocket
technology.

Figure 4: Sputnik-1, Source: Wikipedia


Figure 5: Apollo 11, July 16, 1969,
Image Source: [17] Image source: [16]

Since the Apollo missions, rocket technology has continued to evolve. Private companies such as
SpaceX and Blue Origin have entered the industry, pushing the boundaries of what is possible with
reusable rockets and commercial space travel. Advances in rocket technology have also made it
7
possible to explore further into our solar system, with missions to Mars and beyond planned for
the future.

Figure 6: Falcon Heavy by SpaceX, Image source:[18]

In conclusion, rocket technology has a rich and complex history that dates back centuries. From
its origins in ancient China to the incredible achievements of the Apollo missions, rocket
technology has had a significant impact in shaping the course of human history. As we continue to
explore the vast expanse of space, rocket technology will certainly continue to evolve and push
the boundaries of what is possible.

3.2 Types of Rockets Used for Payload Delivery to Low-Earth Orbit

The delivery of payloads into low-earth orbit (LEO) is an important aspect of space exploration
and commercial activities in space. Rockets are the primary means of launching payloads into
space, and the design of the rocket and its capabilities play a significant role in determining the
payload capacity, launch cost, and reliability of the launch. The primary emphasis for launch
vehicle design and development has shifted significantly from the traditional performance-
optimized design ( maximum payload using the most sophisticated technology with the least
weight) to minimum-cost design and operations.[19]
There are several types of rockets used for payload delivery to LEO, each with its unique
characteristics, advantages, and limitations. The most common types of rockets used for payload

8
delivery to LEO include expendable launch vehicles (ELVs) and reusable launch vehicles
(RLVs)[19].
ELVs are designed for single use, meaning that they are discarded after a single launch. ELVs
consist of several stages that are sequentially fired to produce the required thrust to launch the
payload into orbit. The first stage provides the initial thrust to lift the rocket off the launchpad, and
subsequent stages are fired to achieve the desired orbit. ELVs are the most commonly used rockets
for LEO payload delivery due to their reliability, simplicity, and low cost compared to RLVs[20].
Some examples of the Boeing Delta IV Heavy and the Lockheed Atlas V Rocket Launchers [21].

Figure 7: The Boeing Delta IV Heavy and the Lockheed Atlas V Rocket Launchers

RLVs, on the other hand, are designed for multiple uses, and their various components are
designed to withstand the stress and strains of spaceflight multiple times. RLVs offer several
advantages over ELVs, including reduced launch costs and increased payload capacity due to their
ability to return to Earth and be reused for subsequent launches. RLVs are also more
environmentally friendly as they produce less waste compared to ELVs[22]. However, RLVs

9
require more maintenance and have higher upfront costs than ELVs. Some examples of RLVs
include the Space Shuttle, Falcon Heavy, and New Glenn rockets.

Figure 8: Falcon heavy, SpaceX Figure 9: Space shuttle, NASA

Source:[23]
The selection of a rocket type for payload delivery to LEO depends on several factors, including
payload mass, orbit altitude, mission duration, and launch cost. For instance, ELVs are preferred
for smaller payloads and shorter missions due to their lower cost, while RLVs are preferred for
larger payloads and longer missions due to their increased capacity and reduced launch costs over
multiple launches. Figure 10 below shows an example of a reusable launcher mission. The goal is
for the upper stage to attain the appropriate orbit while the lower and higher stages are safely
recovered for subsequent launches. A multi-branch modeling and a coordinated optimization
strategy are required for this problem[24].

10
Figure 10: Reusable launch vehicle

Another important factor in selecting a rocket for LEO payload delivery is the type of fuel used.
There are solid and liquid types of rocket fuel. Solid fuel rockets use a mixture of chemicals that
are packed into a solid form, whereas liquid fuel rockets use liquid propellants that are stored in
tanks and then mixed and ignited in the rocket's engines[25]. Solid fuel rockets are simpler to build
and have a longer shelf life, but they are less efficient and cannot be throttled or turned off once
ignited. Liquid fuel rockets, on the other hand, are more complex to build and require more
maintenance, but they are more efficient and can be throttled or turned off if necessary[26].
As it is discussed so far, the types of rockets used for payload delivery to LEO vary in design,
capabilities, advantages, and limitations. The selection of a rocket type depends on various factors,
including payload mass, orbit altitude, mission duration, launch cost, and fuel type. As space
exploration and commercial activities in space continue to grow, rocket technology will continue
to evolve, leading to more advanced and efficient rockets for payload delivery to LEO.

11
3. Trajectory Optimization Techniques

3.3 Optimal Control Theory and Approaches

Optimal control theory is an area of mathematics that seeks an optimal strategy to control a
dynamic system and it is a generalization of the calculus of variations [27, 28]. Thus, the theory
applies to many problems that involve systems evolving which also plays a vital role in rocket
trajectory optimization, a fundamental aspect of space exploration and commercial activities in
space. The application of optimal control theory allows engineers to design rockets that follow a
predetermined path while minimizing fuel consumption and maximizing efficiency. It involves the
use of mathematical models and algorithms to determine the optimal control inputs, such as thrust
and trajectory adjustments, to achieve a specific objective, such as reaching a desired orbit. By
optimizing the trajectory of rockets, optimal control theory enables the achievement of more
efficient and cost-effective space missions[28].
As it is discussed in the introductory part, optimum control problems can be solved using either
direct or indirect methods[29]. According to Kelly[30], indirect approaches first create the required
and sufficient criteria for optimality and then solve them numerically. These sufficient conditions
require the existence of a Hamiltonian function, which is a function of the state variables, the
control variables, and the time derivative of the state variables. The Hamiltonian function
represents the total cost of the control problem and it should satisfy the following conditions:
1. The optimal control is obtained when the Hamiltonian function is minimized concerning
the control variables.
2. The costate variables at the terminal time should satisfy certain boundary conditions.
3. The optimal control should satisfy the maximum principle, which means that the optimal
control should be such that it maximizes the Hamiltonian function at each instant of time.
4. The optimal control should satisfy all the system constraints at every instant of time. These
sufficient conditions provide a framework for finding optimal control through an indirect
approach.
The primary disadvantage of indirect approaches is that they require initial guesses for adjoint
variables and are extremely sensitive to the initial guesses' values. The common initial guesses for
adjoint variables in rocket trajectory optimization include:

12
i. The initial guess for the rocket's trajectory variables which includes the initial position,
velocity, and orientation of the rocket.
ii. Initial guess for the control inputs: This includes the initial values for the adjoint variables
like the thrust, angle of attack, and other control parameters that affect the rocket's
trajectory.
iii. Initial guess for the cost function which is comprised of the initial values of the objective
function being minimized or maximized, such as the time of flight, fuel consumption, or
distance traveled and,
iv. Initial guess for the adjoint variables such as the initial values of the Lagrange multipliers
or adjoint variables that are used to compute the gradients of the cost function concerning
the control inputs. Offering these initial guesses for such variables is non-intuitive. Direct
techniques, on the other hand, include parameterizing state or (and) control variables and
converting the problem to a finite-dimensional optimization problem[29].
The drawback of the direct method is that it takes time and requires an accurate initial guess value
to begin iterations. [31]. To improve the low accuracy of direct methods and boost the convergence
areas of indirect methods it is better to use a hybrid approach. Well-developed nonlinear
programming methods can tackle the resulting finite-dimensional optimization problem[32].
A standard nonlinear programming problem (NLP) can be written as [33]:
Find decision variables 𝑥 = [𝑥1 , 𝑥2 , … … , 𝑥𝑛 ]
Minimize objective function: 𝐽(𝑥)

Subject to: 𝑥𝑚𝑖𝑛 ≤ 𝑥 ≤ 𝑥𝑚𝑎𝑥


𝐻𝑖 (𝑥) = 0 (𝑖 = 1,2,3, … … … , 𝐸)
𝐺𝑗 (𝑥) ≤ 0 (𝑗 = 1,2,3, … … … . . 𝐼)

where E and I are the dimensionality of the equality and inequality constraints, respectively. Many
successful optimization approaches can currently be used to handle spacecraft trajectory
optimization challenges. Such as the gradient-based [34], convexification-based, dynamic
programming-based, and derivative-free trajectory optimization techniques.

13
3.1.1 GRADIENT-BASED OPTIMIZATION

This technique involves computing the gradient of the objective function concerning the control
variables and using this information to iteratively update the control variables until a minimum is
reached. Gradient-based methods are widely used for trajectory optimization due to their
computational efficiency and ability to handle complex systems[35, 36]. Recent studies have
shown that gradient-based methods can be applied to a range of trajectory optimization problems,
including lunar landings, earth-moon trajectories, and asteroid rendezvous missions[37]. One of
the gradient-based methods which are intrinsically suitable for dynamic simulation is the direct
multiple shooting method[38]. Even though gradient-based approaches are fast and deterministic,
they are nearly always limited to converging to local minimums, and their performance is highly
dependent on the initial guess of the solution and may require careful initialization to avoid this
issue. According to Chai et al. [39], an Improved gradient-based algorithm is found to be a suitable
way for solving aero-assisted vehicle trajectory optimization problems. This method shows that it
enables the user to control the iteration of the inner loop so that the quadratic programming
subproblem does not need to be exactly solved. By using the iterate solution calculated from the
inner loop, it becomes more accurate to identify the active set, which will have positive influences
in generating the Lagrangian multipliers and next iteration points. Moreover, compared with
standard sequential quadratic programming (SQP), it tends to be more stable and can reduce the
computational time.
In the table-2 below, a different gradient-based optimization technique related to trajectory
optimization is presented along with the authors who have explored them and the key findings
from their works. By analyzing the findings from each work, this compilation aims to offer a
comparative analysis of the different approaches used, highlighting their strengths, limitations, and
potential applications. Each approach used by the researchers below has its unique advantages and
areas of applicability. Researchers have explored these techniques to advance the field of trajectory
optimization, leading to improvements in accuracy, convergence rates, computational efficiency,
and robustness. It is important to note that while this table aims to provide a comprehensive
overview, it may not include every published work on trajectory optimization. The selection
process considered the significance and impact of each study, ensuring the inclusion of influential
works that have contributed to the advancement of optimization problems.

14
Table 2. A summary of a gradient-based optimization technique as used by different authors

Author(s) Title Reference The main findings


Xia et al. Multi-UAV [35] Uses the multi-UAV trajectory planning to solve the challenges in
trajectory planning balancing the various constraints like dynamic effects, collisions,
using gradient-based time, trajectory smoothness, and length by implementing time
sequence minimal segmentation rather than the usual segmentation which makes the
optimization cost function calculation easy. In addition to this, it is shown that
virtual segments are preferable to the trajectory length of different
UAVs to reduce total arrival time. The results show that this
technique performs well in obstacle-rich environments and is
suitable for a high number of UAVs.

Campana et al. Gradient-based path [34] The article discusses a gradient-based path optimization method that
optimization for uses inequality constraints sampled at many points to handle
motion planning collision avoidance and transform the longest path to a shorter one.

Sukhija et al. Gradient-based [40] The document discusses the use of machine learning techniques to
trajectory learn a differentiable dynamics model of a robotic system from data.
optimization with The authors demonstrate that a neural network can accurately model
learned dynamics highly nonlinear behaviors for large time horizons, from data
collected in only 25 minutes of interactions on two distinct robots:
the Boston Dynamics Spot and an RC car. The learned model is then

15
used to perform complex and dynamic maneuvers through gradient-
based trajectory optimization.
Ahmadianfar et al. Gradient-based [41] In this method, the operator’s named gradient search rule (GSR) is
optimizer: A new used to improve the exploration tendency and accelerate the
metaheuristic convergence rate to reach better locations in the search space and
optimization local escaping operator (LEO) to allow it to escape from local
algorithm optima. Because of its improved capabilities for exploration,
exploitation, convergence, and excellent avoidance of local optima,
GBO produced very promising results.

16
3.1.2 Convexification-Based Optimization

This technique involves approximating the objective function with a convex function and solving
the resulting convex optimization problem using efficient algorithms. Convexification-based
methods are effective for spacecraft trajectory optimization problems with non-convex constraints
and objectives[42, 43]. As mentioned by Jansson and Harris [42] in the title “Convex optimization-
based techniques for trajectory design and control of nonlinear systems with polytopic range”, a
nonlinear control problem with convex state and control constraints are discussed. The paper
presents a set of sufficient conditions under which the problem can be solved as a single convex
program. The paper also describes a practical algorithm for implementing the theoretical results.
The main finding in this paper is that a convex polytope can be used to bind the range of the
nonlinear function, and with each vertex of the polytope, an "auxiliary" linear system can be
associated. The sufficient conditions presented in this paper guarantee that a feasible control for
the nonlinear system can be interpolated from the "auxiliary" linear controls obtained via convex
optimization. It also outlines a "resetting approach" that applies even when sufficient conditions
are violated. This study applies the theory and algorithms to two problems in aerospace
engineering: a spherically constrained relative orbital motion problem and an attitude control
problem using the dynamics of the Euler axis. In this article, it is concluded that the proposed
approach can be used to solve a wide range of nonlinear control problems with convex state and
control constraints. Whereas “a convex approach to rocket ascent trajectory optimization”
proposed by Benedikter et al.[44] revealed that the convex programming approach is the best way
to optimize the ascent trajectory of a multistage launch vehicle. The problem is complex due to
highly nonlinear dynamics and multiple mission constraints. The proposed methodology combines
lossless and successive convexification techniques to transform the original nonconvex problem
into a sequence of convex problems that converge to the same solution. The main advantage of the
convexification-based method is that it can guarantee global optimality under certain conditions
[45, 46]. However, the convexification process can be computationally expensive and may
introduce approximation errors.
Moreover, extra works performed by some other scholars using convexification-based
optimization techniques are presented below in table-3 which shows their main findings, the results
they have found at the end of their work, and the area of application of these techniques.

17
Table 3. A summary of convexification-based optimization techniques used by different authors

Authors Title Source Main Findings


Pinson and Lu Trajectory design [43] This paper discusses the challenges of designing a propellant optimal
employing convex powered descent trajectory for landing spacecraft on irregularly shaped
optimization for landing asteroids and proposes a solution by using convex optimization, a gravity
on irregularly shaped model with higher fidelity than Newtonian, and an iterative solution
asteroids process for a fixed final time problem. The success of the algorithm is
demonstrated by designing powered descent trajectories for the
elongated binary asteroid Castalia.
Li et al. A convex approach to [45] The approach used in this research is aimed to handle the online
trajectory optimization for trajectory optimization problem of boost back vertical take-off/vertical
boost back of vertical take- landing reusable launch. To accomplish high-precision landing of launch
off or vertical landing vehicles, trajectory optimization during the boost-back flight phase is
reusable launch vehicles performed, especially in emergencies. The trajectory optimization
problem is posed as an optimal control problem with minimal fuel
consumption, and then it is turned into a series of convex optimization
subproblems that can be solved accurately and quickly using the primal-
dual interior-point approach. The technique provided in this paper
performs well in mission change and thrust drop simulation studies. The
suggested convex technique has a strong potential for onboard
application in reusable launch vehicles and other space vehicles due to
its high computational efficiency and exceptional durability.

18
Hu et al. Successive convexification [46] The article proposes an algorithm that converges quickly to the optimal
for online ascent trajectory solution even with a coarse initial guess. The method is demonstrated
optimization through iterative solutions of the ascent trajectory optimization problem
for a small guided rocket, and its accuracy is verified by comparison with
optimal control solvers. The proposed algorithm can be implemented
online and is validated by simulations.

19
3.1.3 Dynamic Programming-Based Optimization

Dynamic optimization possesses solving the Bellman equation recursively to find the optimal
control policy. These methods are particularly effective for trajectory optimization problems with
complex dynamics and non-convex objectives. Zhang et al. [47] proposed a trajectory planning
based on non-convex global optimization for serial manipulators. The proposed method is applied
to the continuous trajectory planning problem for haptic feedback manipulators in virtual reality
systems. The results show that the proposed method can generate smooth velocity curves and each
joint can reach the interacting point in the interacting time. Overall, the paper provides a promising
approach to reducing the calculation time for trajectory planning in frequently changing
environments. Recent studies have shown that dynamic programming-based method can be used
for a wide range of trajectory optimization problems, including satellite trajectory design and
planetary exploration missions[48]. In another study conducted by Lin et al. [49], dynamic
programming is used for space station orbit design. The approach divides the long-duration orbit
design process into several planning periods, each consisting of five basic flight segments. Each
planning period is modeled as a multi-step decision process with consideration of the interaction
effects of different flight segments, and the dynamic programming approach is used to optimize
the total propellant consumption. The results show that the proposed approach can effectively
optimize the design of altitude strategies and can save considerable propellant consumption for the
space station compared to previous studies. However, the paper does not provide specific
numerical results or data on the amount of propellant saved. As mentioned in the article [50],
dynamic programming is crucial to the applications of flight trajectory optimization. This paper
examines two application examples of a winged space vehicle’s reentry flight and the fuel-efficient
flight of a jet passenger aircraft to demonstrate the advantages of DP and discusses how to
overcome its disadvantages. It is indicated that this technique provides a global optimization,
predictable computational time, feasibility in handling inequality constraints, and simplicity of
computer programming. This dynamic programming technique has been used in numerous
applications in the space industry. However, these methods require more time to compute and
cause quantization errors which may require careful discretization of the state and control spaces.

20
3.1.4 DERIVATIVE-FREE (HEURISTIC-BASED) OPTIMIZATION

In this method searching for the optimal solution using heuristic search algorithms is applied that
do not require derivatives of the objective function. Particle swarm optimization and genetic
algorithms are heuristic methods that are not dependent on the gradient of the system and do not
easily relent to local minima, and can be applied to a wide range of problems[51]. The optimization
method proposed in this article aims to optimize the complete trajectory from launch through re-
entry. Unlike the typical multiphase trajectory optimization technique, which directly employs the
original dynamical model, a surrogate model is substituted for the original time-consuming model
to increase iteration efficiency and reduce computational cost. Derivative-free methods are
particularly effective for optimization problems with non-differentiable or discontinuous objective
functions and constraints[52, 53]. A study titled Surrogate-based entire trajectory optimization for
full space mission from launch to reentry by He et al. [51] used a derivative-free technique. In this
study, the authors tested 22 leading software implementations of state-of-the-art derivative-free
optimization algorithms on a large collection of publicly available problems and found that
attaining the best solutions even for small problems is a challenge for most current derivative-free
solvers. However, computational results show that there is no single solver whose performance
dominates that of all others. In addition, all solvers provided the best solution possible for at least
some of the test problems. Problem dimensionality and non-smoothness were found to rapidly
increase the complexity of the search and decrease performance for all solvers. Finally, from a
starting point close to a solution, TOMLAB/OQNLP, NEWUOA, and TOMLAB/MULTIMIN
showed the fastest convergence toward the solution. Missing bounds on the variables are found to
affect significantly the performance of all solvers, particularly the stochastic ones. Another study
conducted by Rios and Sahinidis [52] provides a review of derivative-free optimization algorithms
and compares their performance on various test problems. The algorithms are divided into two
major groups: those that do not partition the search space and those that do. The algorithms that
do not partition the search space use information from a subset of points and the next point to be
evaluated may be located anywhere in the search space. Examples of methods using information
from a single point include simulated annealing and pattern search. The algorithms that partition
the search space divide the search space into subsets and use information from each subset to
determine the next point to be evaluated. Examples of methods that partition the search space

21
include mesh adaptive direct search (MADS) and compass search. The article also emphasizes the
importance of considering the structure of the search space when selecting an optimization
algorithm. Some algorithms may perform better on certain types of problems, such as those with
a smooth or non-smooth objective function, or those with a large or small number of variables. All
in all, derivative-free methods can be used in a trajectory optimization problem including
spacecraft re-entry and orbital transfer missions. However, these methods may not guarantee
global optimality and may require careful tuning of the search parameters.

22
4. Discussion and Recommendation

In summary, each optimization technique discussed in this study exhibits both advantages and
drawbacks, making the selection of a specific technique contingent upon the nature of the problem
being addressed. Recent studies have demonstrated the efficacy of these techniques across a wide
range of spacecraft trajectory optimization problems. To address the gaps identified in previous
research, a recommended solution involves adopting a hybrid approach that combines different
optimization techniques. This approach capitalizes on the strengths of each technique to overcome
their respective limitations, yielding more efficient and effective solutions for spacecraft trajectory
optimization problems. For instance, a hybrid approach can leverage the speed and efficiency of
gradient-based methods alongside the global optimality offered by convexification-based methods.
By utilizing a gradient-based method to identify a promising initial solution, followed by refining
the solution using a convexification-based method, the hybrid approach achieves both
computational speed and the assurance of global optimality.
Likewise, a hybrid approach can merge the global optimality inherent in dynamic programming-
based methods with the computational efficiency provided by derivative-free methods. This can
be accomplished by employing dynamic programming-based methods to generate an initial
solution, and subsequently employing derivative-free methods to fine-tune the solution, handling
scenarios involving non-differentiable or discontinuous objective functions and constraints. By
combining these techniques in a hybrid framework, researchers and engineers can augment the
capabilities of spacecraft trajectory optimization, offering more robust and adaptable solutions that
encompass the strengths of multiple optimization approaches. This approach holds promise for
enhancing the efficiency and reliability of trajectory planning in space missions.
One area for future investigation in the field of rocket trajectory optimization for maximum
payload delivery is the development of machine learning-based approaches. Machine learning
algorithms, such as deep learning and reinforcement learning, have shown great potential in
various domains. Applying these techniques to rocket trajectory optimization can lead to more
intelligent and adaptive optimization strategies. Future work should explore the use of machine
learning models to learn from historical data and real-time flight data, enabling the development
of predictive models that can efficiently optimize rocket trajectories for maximum payload

23
delivery. Additionally, combining machine learning with traditional optimization algorithms can
potentially yield hybrid approaches that leverage the strengths of both methods.

24
5. Conclusion

The optimization of a rocket's trajectory for maximum payload delivery is an important problem
in aerospace engineering. Different methods have been used in previous studies to deal with the
trajectory problem. However, these studies focused on optimizing the trajectory for minimal fuel
consumption or minimal launch cost, rather than for maximum payload delivery. This research
aimed to address the gaps in previous research by using a combination of optimization techniques
to provide a more flexible and powerful solution to spacecraft trajectory optimization problems by
combining the strengths of multiple techniques. This review has the potential to contribute
significantly to the field of rocket trajectory optimization and improve the efficiency and reliability
of rocket launches.

25
References

[1] F. Wang, S. Yang, F. Xiong, Q. Lin, and J. Song, "Robust trajectory optimization using
polynomial chaos and convex optimization," Aerospace Science and Technology, vol. 92,
pp. 314-325, 2019/09/01/ 2019.
[2] J. F. Seabra and J. F. Seabra, "Spacecraft Trajectory Optimization," 2015.
[3] S. A. Darani and O. Abdelkhalik, "Space Trajectory Optimization Using Hidden Genes
Genetic Algorithms," vol. 55, no. 3, pp. 764-774, 2018.
[4] F. Marchetti, E. Minisci, and A. Riccardi, "Single-stage to orbit ascent trajectory
optimization with reliable evolutionary initial guess," Optimization and Engineering, vol.
24, no. 1, pp. 291-316, 2023/03/01 2023.
[5] A. Shirazi, J. Ceberio, and J. Lozano, "Spacecraft trajectory optimization: A review of
models, objectives, approaches, and solutions," Progress in Aerospace Sciences, vol. 102,
08/01 2018.
[6] J. T. Betts, "Survey of Numerical Methods for Trajectory Optimization," Journal of
Guidance, Control, and Dynamics, vol. 21, no. 2, pp. 193-207, 1998/03/01 1998.
[7] O. von Stryk and R. Bulirsch, "Direct and indirect methods for trajectory optimization,"
Annals of Operations Research, vol. 37, no. 1, pp. 357-373, 1992.
[8] S. W. Hur, S. H. Lee, Y. H. Nam, and C.-J. Kim, "Direct dynamic-simulation approach to
trajectory optimization," Chinese Journal of Aeronautics, vol. 34, no. 10, pp. 6-19, 2021.
[9] J. Fageot, S. Aziznejad, M. Unser, and V. Uhlmann, "Support and approximation properties
of Hermite splines," Journal of Computational and Applied Mathematics, vol. 368, p.
112503, 2020/04/01/ 2020.
[10] O. Von Stryk and R. Bulirsch, "Direct and Indirect Methods for Trajectory Optimization,"
Annals of Operations Research, vol. 37, pp. 357-373, 12/01 1992.
[11] B. Conway, "A Survey of Methods Available for the Numerical Optimization of
Continuous Dynamic Systems," J. Optimization Theory and Applications, vol. 152, pp.
271-306, 02/01 2012.
[12] G. Yang, "Earth-moon Trajectory Optimization Using Solar Electric Propulsion," Chinese
Journal of Aeronautics, vol. 20, pp. 452-463, 10/01 2007.
[13] L. Ravikumar, R. Padhi, and N. K. Philip, "Trajectory optimization for Rendezvous and
Docking using Nonlinear Model Predictive Control⁎⁎This research work have been carried
as a part of Ph D program at IISc Aerospace Engineering," IFAC-PapersOnLine, vol. 53,
no. 1, pp. 518-523, 2020/01/01/ 2020.
[14] Y. Kayama, K. Howell, M. Bando, and S. Hokamoto, Low-Thrust Trajectory Design With
Convex Optimization For Libration Point Orbits. 2021.
[15] Elsabeth Howell. (2022, 6th May 2023). The History of Rockets. Available:
https://www.space.com/29295-rocket-history.html
[16] K. Teague. (July 27, 2007). History of NASA. Available: https://history.nasa.gov/ap11-
35ann/kippsphotos/apollo.html
[17] W. contributors. (April 26, 2023). Sputnik 1
Available: https://en.wikipedia.org/w/index.php?title=Sputnik_1&oldid=1151931716
[18] SpaceX. (2021). Falcon heavy. Available: https://www.spacex.com/vehicles/falcon-9/
[19] D. Koelle, "Reusable Launch Vehicles (RLVs) and Cost Engineering Principles," 2001, pp.
157-166.

26
[20] G. Gstattenbauer, M. Franke, and J. Livingston, Cost Comparison of Expendable, Hybrid,
and Reusable Launch Vehicles. 2006.
[21] M. Lorell, R. Leonard, and A. Doll, Extreme Cost Growth: Themes from Six U.S. Air Force
Major Defense Acquisition Programs. 2015.
[22] K. Dresia et al., "Multidisciplinary Design Optimization of Reusable Launch Vehicles for
Different Propellants and Objectives," Journal of Spacecraft and Rockets, vol. 58, no. 4,
pp. 1017-1029, 2021.
[23] A. Thompson. (April 05, 2019). Spaceflight. Available: https://www.space.com/spacex-
falcon-heavy
[24] J. Zhu, E. Trélat, and M. Cerf, "Geometric Optimal Control and Applications to
Aerospace," Pacific Journal of Mathematics for Industry, vol. 9, 01/22 2017.
[25] K. O. Hartman and S. Morrow, "Solid Propellants," in Encyclopedia of Physical Science
and Technology (Third Edition), R. A. Meyers, Ed. New York: Academic Press, 2003, pp.
277-293.
[26] M. J. Casiano, J. R. Hulka, and V. Yang, "Liquid-Propellant Rocket Engine Throttling: A
Comprehensive Review," Journal of Propulsion and Power, vol. 26, no. 5, pp. 897-923,
2010/09/01 2010.
[27] S. P. Sethi, "What Is Optimal Control Theory?," in Optimal Control Theory: Applications
to Management Science and Economics, S. P. Sethi, Ed. Cham: Springer International
Publishing, 2021, pp. 1-23.
[28] M. H. A. Biswas, M. Huda, M. Ara, and M. Rahman, "Optimal Control Theory and Its
Applications in Aerospace Engineering," International Journal of Academic Research, Vol.
3, No. 2, Part II, pp. 349-357, 2011., 01/01 2011.
[29] G. Sushnigdha, "Spacecraft Reentry Trajectory Optimization using Search Space
Reduction Technique," IFAC-PapersOnLine, vol. 55, no. 1, pp. 46-51, 2022/01/01/ 2022.
[30] M. J. S. R. Kelly, "An Introduction to Trajectory Optimization: How to Do Your Own
Direct Collocation," vol. 59, pp. 849-904, 2017.
[31] S. S and S. Swaminathan, "Real Time Trajectory Optimization of L1 Optimal Control
Problem using Gauss-Pseudo-spectral Method," IFAC-PapersOnLine, vol. 53, no. 1, pp.
258-265, 2020/01/01/ 2020.
[32] R. Chai, A. Savvaris, A. Tsourdos, S. Chai, and Y. Xia, "A review of optimization
techniques in spacecraft flight trajectory design," Progress in Aerospace Sciences, vol. 109,
06/01 2019.
[33] J. Laurent-Varin, M. Haddou, and C. Talbot, "An Interior-Point Approach to Trajectory
Optimization," Journal of Guidance Control and Dynamics - J GUID CONTROL DYNAM,
vol. 30, pp. 1228-1238, 09/01 2007.
[34] M. Campana, F. Lamiraux, and J.-P. Laumond, "A gradient-based path optimization
method for motion planning," Advanced Robotics, pp. 1126-1144, 04/11 2016.
[35] Q. Xia, S. Liu, M. Guo, H. Wang, Q. Zhou, and X. Zhang, "Multi-UAV trajectory planning
using gradient-based sequence minimal optimization," Robotics and Autonomous Systems,
vol. 137, p. 103728, 2021/03/01/ 2021.
[36] B. Subburaj, U. M. Jayachandran, V. Arumugham, and M. J. A. Suthanthira Amalraj, "A
Self-Adaptive Trajectory Optimization Algorithm Using Fuzzy Logic for Mobile Edge
Computing System Assisted by Unmanned Aerial Vehicle," vol. 7, no. 4, p. 266, 2023.
[37] A. Miele and S. Mancuso, "OPTIMAL TRAJECTORIES FOR EARTH–MOON–EARTH
FLIGHT," Acta Astronautica, vol. 49, no. 2, pp. 59-71, 2001/07/01/ 2001.

27
[38] M. Sadegh Mohammadi and A. Naghash, "Robust optimization of impulsive orbit transfers
under actuation uncertainties," Aerospace Science and Technology, vol. 85, pp. 246-258,
2019/02/01/ 2019.
[39] R. Chai, A. Savvaris, A. Tsourdos, S. Chai, and Y. Xia, "Improved Gradient-Based
Algorithm for Solving Aeroassisted Vehicle Trajectory Optimization Problems," Journal
of Guidance, Control, and Dynamics, vol. 40, pp. 2091-2099, 08/01 2017.
[40] B. Sukhija et al., "Gradient-Based Trajectory Optimization With Learned Dynamics," p.
arXiv: 2204.04558, 2022.
[41] I. Ahmadianfar, O. Bozorg-Haddad, and X. Chu, "Gradient-based optimizer: A new
metaheuristic optimization algorithm," Information Sciences, vol. 540, pp. 131-159,
2020/11/01/ 2020.
[42] O. Jansson and M. W. Harris, "Convex Optimization-Based Techniques for Trajectory
Design and Control of Nonlinear Systems with Polytopic Range," vol. 10, no. 1, p. 71,
2023.
[43] R. M. Pinson and P. Lu, "Trajectory Design Employing Convex Optimization for Landing
on Irregularly Shaped Asteroids," vol. 41, no. 6, pp. 1243-1256, 2018.
[44] B. Benedikter, A. Zavoli, and G. Colasurdo, A Convex Approach to Rocket Ascent
Trajectory Optimization. 2019.
[45] Y. Li, C. Wei, Y. He, and R. Hu, "A convex approach to trajectory optimization for boost
back of vertical take-off/vertical landing reusable launch vehicles," Journal of the Franklin
Institute, vol. 358, no. 7, pp. 3403-3423, 2021/05/01/ 2021.
[46] C. Hu, X. Bai, S. Zhang, and H. Yang, "Successive Convexification for Online Ascent
Trajectory Optimization," IEEE Access, vol. PP, pp. 1-1, 10/15 2021.
[47] S. Zhang, S. Dai, A. M. Zanchettin, and R. Villa, "Trajectory planning based on non-convex
global optimization for serial manipulators," Applied Mathematical Modelling, vol. 84, pp.
89-105, 2020/08/01/ 2020.
[48] F. Salvioli, F. Capasso, E. Ferrentino, and P. Chiacchio, "Globally-optimal whole-body
motion planning under nonholonomic constraints using dynamic programming," Acta
Astronautica, vol. 193, pp. 619-626, 2022/04/01/ 2022.
[49] K.-P. Lin, Y. Luo, and G.-J. Tang, "Space station orbit design using dynamic
programming," Acta Astronautica, vol. 89, pp. 195–204, 08/01 2013.
[50] A. Harada and Y. Miyazawa, "Dynamic Programming Applications to Flight Trajectory
Optimization," IFAC Proceedings Volumes, vol. 46, no. 19, pp. 441-446, 2013/01/01/ 2013.
[51] X. He, X. Zuo, Q. Li, M. Xu, and J. Li, "Surrogate-based entire trajectory optimization for
full space mission from launch to reentry," Acta Astronautica, vol. 190, pp. 83-97,
2022/01/01/ 2022.
[52] L. Rios and N. Sahinidis, "Derivative-free optimization: A review of algorithms and
comparison of software implementations," Journal of Global Optimization, vol. 56, 11/10
2009.
[53] F. Forouzanfar and A. C. Reynolds, "Well-placement optimization using a derivative-free
method," Journal of Petroleum Science and Engineering, vol. 109, pp. 96-116, 2013/09/01/
2013.

28

You might also like