Targeted delivery of gold nanoparticles by neural stem cells to glioblastoma for enhanced radiation therapy: a review

Mogesh Sababathy; Ghayathri Ramanathan; Suat Cheng Tan; Mogesh Sababathy; Ghayathri Ramanathan; Suat Cheng Tan

doi:10.3934/Neuroscience.2022017

AIMS Neuroscience

2022, Volume 9, Issue 3: 303-319. doi: 10.3934/Neuroscience.2022017

Previous Article Next Article

Review

Targeted delivery of gold nanoparticles by neural stem cells to glioblastoma for enhanced radiation therapy: a review

1.
Department of Veterinary Pathology and Microbiology, Faculty of Veterinary Medicine, Universiti Putra Malaysia, Serdang, Malaysia
2.
Institute of Biological Sciences, Faculty of Science, University of Malaya, Kuala Lumpur, Malaysia
3.
School of Health Sciences, Health Campus, Universiti Sains Malaysia, 16150 Kubang Kerian, Kelantan, Malaysia

Received: 29 March 2022 Revised: 09 June 2022 Accepted: 01 July 2022 Published: 08 July 2022

Glioblastoma (GB) is the most malignant subtype of brain cancer derived from astrocytes in the brain. Radiotherapy is one of the standard treatments for GB patients, but its effectiveness is often limited by the radioresistance of aggressive GB cells. Higher dose of radiation needs to be applied to GB patients to eliminate these stubborn cells, but this also means more side effects on the adjacent healthy cells because the radiation beam could indistinguishably harm all cells exposed to it. In order to address this problem, various strategies have been studied to enhance the radiosensitivity among the radioresistant cell populations for targeted eradication of GB without harming other surrounding healthy cells. One of the promising strategies for radiosensitization is to use gold nanoparticles (AuNPs) which can enhance photoelectric effects within the radioresistant cells for higher killing efficiency even at low doses of radiation. Nonetheless, there is no evidence showing the capability of these nanoparticles to travel to brain tumor cells, therefore, the application of this nanotechnology is very much dependent on the development of a suitable carrier to deliver the AuNPs to the GB tumor sites specifically. In this review article, we discussed the potentials of neural stem cells (NSCs) as biological carriers to carry AuNPs to targeted GB tumor sites and provided new insights into the potential of NSC-based targeted delivery system for GB treatment. The information reported here may pave a new direction for clinical transformation of next-generation nanoparticle-assisted radiotherapy to optimize the efficacy of radiotherapy for GB treatment.

Keywords:

Citation: Mogesh Sababathy, Ghayathri Ramanathan, Suat Cheng Tan. Targeted delivery of gold nanoparticles by neural stem cells to glioblastoma for enhanced radiation therapy: a review[J]. AIMS Neuroscience, 2022, 9(3): 303-319. doi: 10.3934/Neuroscience.2022017

Related Papers:

[1]	Wenyang Gan, Lixia Su, Zhenzhong Chu . A PSO-enhanced Gauss pseudospectral method to solve trajectory planning for autonomous underwater vehicles. Mathematical Biosciences and Engineering, 2023, 20(7): 11713-11731. doi: 10.3934/mbe.2023521
[2]	Junjie Liu, Hui Wang, Xue Li, Kai Chen, Chaoyu Li . Robotic arm trajectory optimization based on multiverse algorithm. Mathematical Biosciences and Engineering, 2023, 20(2): 2776-2792. doi: 10.3934/mbe.2023130
[3]	Zhiqiang Wang, Jinzhu Peng, Shuai Ding . A Bio-inspired trajectory planning method for robotic manipulators based on improved bacteria foraging optimization algorithm and tau theory. Mathematical Biosciences and Engineering, 2022, 19(1): 643-662. doi: 10.3934/mbe.2022029
[4]	Juan Du, Jie Hou, Heyang Wang, Zhi Chen . Application of an improved whale optimization algorithm in time-optimal trajectory planning for manipulators. Mathematical Biosciences and Engineering, 2023, 20(9): 16304-16329. doi: 10.3934/mbe.2023728
[5]	Xingjia Li, Jinan Gu, Zedong Huang, Wenbo Wang, Jing Li . Optimal design of model predictive controller based on transient search optimization applied to robotic manipulators. Mathematical Biosciences and Engineering, 2022, 19(9): 9371-9387. doi: 10.3934/mbe.2022436
[6]	Tian Xue, Liu Li, Liu Shuang, Du Zhiping, Pang Ming . Path planning of mobile robot based on improved ant colony algorithm for logistics. Mathematical Biosciences and Engineering, 2021, 18(4): 3034-3045. doi: 10.3934/mbe.2021152
[7]	Jinhua Zhou, Shan Jia, Jinbao Chen, Meng Chen . Motion and trajectory planning modeling for mobile landing mechanism systems based on improved genetic algorithm. Mathematical Biosciences and Engineering, 2021, 18(1): 231-252. doi: 10.3934/mbe.2021012
[8]	Xiaoyong Xiong, Haitao Min, Yuanbin Yu, Pengyu Wang . Application improvement of A* algorithm in intelligent vehicle trajectory planning. Mathematical Biosciences and Engineering, 2021, 18(1): 1-21. doi: 10.3934/mbe.2021001
[9]	Hongmin Chen, Zhuo Wang, Di Wu, Heming Jia, Changsheng Wen, Honghua Rao, Laith Abualigah . An improved multi-strategy beluga whale optimization for global optimization problems. Mathematical Biosciences and Engineering, 2023, 20(7): 13267-13317. doi: 10.3934/mbe.2023592
[10]	Rui Zhang, Shuicai Wu, Weiwei Wu, Hongjian Gao, Zhuhuang Zhou . Computer-assisted needle trajectory planning and mathematical modeling for liver tumor thermal ablation: A review. Mathematical Biosciences and Engineering, 2019, 16(5): 4846-4872. doi: 10.3934/mbe.2019244

Abstract

1. Introduction

With the continuous development of modern industry and the improvement of automation level, more and more industrial robots are applied in modern manufacturing. In order to pursue a more efficient and stable robot arm, trajectory planning is one of the important parameters to evaluate the motion of the robot arm. At present, a lot of research has been carried out for the smooth trajectory, optimal time, shortest path and lowest energy consumption of robotic arms. There are different constraints for different task requirements. In the casting sorting industry, for the handling and sorting of small castings with a large number of sand falls, there are certain requirements for the handling speed and smoothness of the mechanical arm, and time is an important factor to be considered to ensure that the execution efficiency of the casting sorting robot is maximized within the allowed range.

Generally, the motion path points used for trajectory planning are given in Cartesian space or joint space, thus ensuring that all target positions are reachable during the robotic arm is functioning. Under the constraint condition, it is the main goal to find a trajectory line with the shortest time between multi-segment waypoints. In recent years, a plethora of methods have been studied to optimize trajectories. Smooth trajectories can be constructed using interpolation functions, such as spline functions and polynomial functions ^{[1,2,3,4,5,6]}. In existing studies, it has been proposed to use 3rd-degree polynomial and Bessel curve to compound smooth trajectories ^[7], and effectively reduced the speed and acceleration. However, compared with B-spline curves, Bessel curve lacks local modification, and the trajectory optimized by non-uniform B-spline curves can adjust the curvature of local curves ^[8]. To ensure acceleration continuity, a method of generating trajectories involves segmenting and discretizing the joint angle sequence using fifth-order polynomial interpolation ^[9]. Zhao et al. conducted kinematic analysis and simulation comparisons of 5th order and 3rd order polynomials to validate the planning effectiveness of the 5th order polynomial through five iterations of verification. This approach was then applied to six types of industrial robotic arms ^[10]. Smooth trajectories for the robotic arm can also be achieved through the combination of traditional interpolation functions. By utilizing segment-wise cubic interpolation functions, such as Hermite-type functions, that also can be generated smooth trajectories ^[11,12]. Alternatively, a novel interpolation planning technique called "5-4-5" or "5-4-5-5" can be employed. By arranging different segments of polynomial interpolation function sequences, the smoothness and stability of the trajectories can be significantly improved ^[13].

These methods improve the motion performance of the robot through interpolation or curvature calculations. However, to ensure both smooth trajectories and minimal execution time, adjustments need to be made to the motion parameters and the methods employed for generating smooth trajectories. Generally, the trajectory planning problem is transformed into nonlinear function optimization problem by intelligent optimization algorithm. In the realm of optimization, there are two major directions: Deterministic and non-deterministic. Deterministic algorithms, often gradient-based, have some optimization effectiveness for linear or nonlinear function problems. Traditional heuristic algorithms also fall into this category. However, due to their fixed inputs and outputs, as well as repetitive computations, they are prone to becoming trapped in local optima ^[14]. On the other hand, non-deterministic methods rely on a combination of randomization and heuristic solutions, such as genetic algorithms and particle swarm optimization (PSO). These methods leverage their inherent randomness, diversity, and independence to break free from the vicious cycle of local optima and effectively solve and optimize problems. In recent years, many scholars have adopted these methods to optimize trajectory planning problems. Using a genetic algorithm, Zhang et al. conducted a global optimization of fifth-order interpolation joint nodes and computed the optimal time intervals between the interpolation points of each joint ^[15]. Shi et al. proposed an improved simulated annealing algorithm to overcome challenges associated with high computational complexity and local minima. This algorithm was designed for dynamic path planning, aiming to provide time-optimal solutions for the robotic arm in both static and dynamic environments ^[16]. Zhang et al. adopted Sparrow Search Algorithm (SSA) based on Tent Chaos mapping to achieve real-time optimization of robot arm trajectory ^[17]. Wang et al. introduced an enhanced Whale Optimization Algorithm to optimize the objective function defined by time and pulses. This approach ensures the smooth continuity and operational efficiency of the robot ^[18]. Ivan Carvajal et al. optimized the path length of a 6-axis robot based on nonlinear regression of cubic polynomial combinations to represent trajectories. They validated the feasibility of this approach through simulation ^[19]. Özge and Bekir utilized the Particle Swarm Optimization (PSO) algorithm to optimize fifth-order polynomial interpolation functions. Their approach specifically aimed to determine the shortest path for point-to-point tasks and achieve time optimization ^[20]. There are many other optimization algorithms and joint trajectory planning methods for different types of robots, not limited to a single algorithm or approach. For different robots and different optimization goals, there has been considerable relevant research ^[21,22,23]. In terms of bio-inspired optimization algorithms, with the continuous emergence of non-deterministic intelligent optimization algorithms, such as genetic algorithms, particle swarm algorithms, or fish swarm algorithms, the field has embarked on analyzing the behaviors of various organisms and deriving relevant mathematical models. This has resulted in a diverse range of bio-inspired optimization algorithms, including the whale algorithm, zebra algorithm, and social spider algorithm. These algorithms have shown promise in addressing conventional engineering application problems, but they also come with their own strengths and weaknesses. Some intelligent optimization algorithms may suffer from premature convergence, lack of robust global search capabilities, or a tendency to become trapped in local optima. However, a novel nature-inspired heuristic algorithm proposed by Gang Hu and colleagues, based on the behavior of Genghis Khan sharks, offers a unique approach ^[24]. By emulating the hunting, movement, foraging, and self-protection behaviors of GKS, this algorithm effectively executes optimization tasks for intelligent agents and accomplishes desired optimization objectives. While there are numerous case studies analyzing specific mathematical problems related to this method, there is limited research on its application in practical robotic arm trajectory planning. Hence, this article holds significant research value in exploring the potential of this approach in such real-world applications.

We present a method for optimizing the trajectory of a casting sorting robotic arm using a nature-inspired heuristic algorithm based on the behavior of Genghis Khan sharks. Calculating the basic parameters of the robotic arm for casting sorting through robot kinematics. With the constraints of task path angles, velocity, and acceleration in mind, a segmented polynomial interpolation method called "3-5-3" is utilized for joint trajectory planning. Objective functions are then constructed, and the optimization problem is solved using an equivalent mathematical model that mimics the four behaviors of GKS. The algorithm seeks to find the optimal time solution for the trajectory problem. The obtained results are compared with those from particle swarm optimization and improved particle swarm optimization algorithms. To validate the effectiveness of the algorithm, experimental simulations are conducted.

2. Robot kinematics

2.1. Kinematic calculations

The robotic arm consists of a series of rotating or translating joints combined with multiple linkages. By manipulating the angles between the joints, the desired pose of the end effector can be determined. Kinematic calculations involve transforming between joint angle space and Cartesian space to establish the relationship between angles and positions. The coordinate system of the mechanical arm is established with the rotating joint of the mechanical arm as the Z axis. Figure 1 shows the relationship between the coordinate systems on the six-axis mechanical arm. Through the Angle and displacement parameters of each connecting rod, the position and attitude of the end can be described by the transformation matrix.

Figure 1. Six-axis robotic arm coordinate system.

DownLoad: Full-Size Img PowerPoint

Through forward kinematics, joint angles and distances can be converted into an end effector pose transformation matrix. Inverse kinematics calculations can transfer the joint angles, based on this matrix. By utilizing the provided parameters, the kinematic equations for the robot can be established

1) Inter-bar transformation matrix

${}_{\text{i}}^{{\text{i - }}1}T = \left[ {\begin{array}{*{20}{c}} {\cos {\theta _i}}&{ - \sin {\theta _i}}&0&{{a_{i - 1}}} \\ {\sin {\theta _i}\cos {\alpha _{i - 1}}}&{\cos {\theta _i}\cos {\alpha _{i - 1}}}&{ - \sin {\alpha _{i - 1}}}&{ - \sin {\alpha _{i - 1}}{d_i}} \\ {\sin {\theta _i}\sin {\alpha _{i - 1}}}&{\cos {\theta _i}\sin {\alpha _{i - 1}}}&{\cos {\alpha _{i - 1}}}&{\cos {\alpha _{i - 1}}{d_i}} \\ 0&0&0&1 \end{array}} \right]$

(1)

2) The kinematic equation is established by multiplying multiple transformation matrices to obtain the transformation matrix of the angle and displacement of the end effector.

${}_6^0T = {}_1^0T{}_2^1T{}_3^2T{}_4^3T{}_5^4T{}_6^5T = \left[ {\begin{array}{*{20}{c}} {{r_{11}}}&{{r_{12}}}&{{r_{13}}}&{{P_x}} \\ {{r_{21}}}&{{r_{22}}}&{{r_{23}}}&{{P_y}} \\ {{r_{31}}}&{{r_{32}}}&{{r_{33}}}&{{P_z}} \\ 0&0&0&1 \end{array}} \right]$

(2)

where r_{i, j} (i = 1, 2, 3, j = 1, 2, 3) is the rotation Angle transformation matrix and P_x, P_y, P_z is the displacement transformation matrix. When the end attitude is given, the joint Angle is inversely solved using the inverse transformation of the connecting rod matrix. The inverse solution obtained is typically not unique. By comparing the motion parameters of each joint and sequentially eliminating possibilities, the correct angles can be determined. Axes 1, 2, and 3 are responsible for determining the precise positioning of the robot hand within the operational area, while axes 4, 5, and 6 play a crucial role in establishing the desired orientation of the end effector. Algebraic methods are employed to solve for the first three axes. By left-multiplying the equation _i-1ⁱT with the inverse transformation matrix, the joint variables are separated, enabling their determination. By gradually solving for θ₁ to θ₂ using element-wise comparisons, there can be up to 8 possible solutions. However, due to structural limitations and constraints on joint angles, some solutions can be ruled out. In situations with multiple solutions, a suitable set of solutions can be selected based on specific criteria, such as minimizing joint motion or ensuring motion accuracy ^{[25,26,27,28]}. This selection process aims to fulfill the operational requirements of the robotic arm.

2.2. Robotic workspace

The dimensions of a small-scale casting sorting robotic arm are determined based on the desired working space required for factory tasks. This workspace includes the region where the robotic arm can effectively grip and place small castings. The working space of the robotic arm is defined by considering the limitations on joint angles, as well as the constraints on velocity and acceleration. These limitations and constraints are outlined in Table 2.

Table 1. Denavit-Hartenberg table of 6-DOF robotic arm.

Link number	Link twist angle α_i-1 (rad)	Link length a_i-1 (mm)	Joint distance d_i (mm)	Joint angle θ_i-1 (rad)
1	-pi/2	32	80	0
2	0	108	0	- pi/2
3	-pi/2	20	0	0
4	pi/2	0	175	0
5	-pi/2	0	0	0
6	0	0	20	0

| Show Table

DownLoad: CSV

Table 2. Robotic arm parameter limits.

Joint	1	2	3	4	5	6
Working range (rad)	-1.92~2.79	-1.75~1.75	-2.09~1.05	-3.14~3.14	-3.49~2.09	-6.28~6.28
Maximum speed (rad/s)	1.48	1.05	1.13	3.49	3.49	7.85
Maximum Acceleration (rad²/s)	0.87	0.78	0.78	1.31	1.31	1.40

| Show Table

DownLoad: CSV

Traditional graphical methods and numerical methods have limitations in accurately describing the working space of a mechanical system. Numerical methods often involve excessive computational complexity, and their results may not be sufficiently reliable for regions near the boundaries. Therefore, a statistical simulation method called the Monte Carlo method is utilized to construct the comprehensive working space of the robotic arm as accurately as possible ^[29]. This approach relies on the joint angle limitations for solving the problem.

The fundamental principle of this solution is to randomly sample angles within the range of joint angles for the robotic arm and use the accumulation of these random points to construct the overall working space. This method is applicable to the calculation of the working space for robotic arms with various joint configurations. In this study, a total of 30,000 random points is generated. Figure 2 depicts the working space of the small-scale casting sorting robotic arm.

Figure 2. Casting sorting robotic arm workspace.

DownLoad: Full-Size Img PowerPoint

3. Trajectory planning

3.1. Cubic polynomial interpolation function

The point-to-point motion mode (PTP) of a small casting sorting robot can utilize various common polynomial interpolation functions, such as cubic polynomials, fifth-degree polynomials, seventh-degree polynomials, etc. In situations where the initial and end points have zero velocity, the third-order polynomial enables interpolation between two points to create intermediate points while also defining the joint angle and time. Therefore, we adopt third-order polynomials. By sequentially processing these intermediate points, the following expression is obtained:

$q(t) = {a_0} + {a_1}(t - {t_0}) + {a_2}{(t - {t_0})^2} + {a_3}{(t - {t_0})^3}$

(3)

In this equation, a_i (where i = 0, 1, 2, 3) represents the coefficients of the polynomial. Taking the derivative of the polynomial allows us to derive the expressions for the joint velocity and angular acceleration of the robotic arm:

$\left\{ {\begin{array}{*{20}{c}} {\dot q(t) = {a_1} + 2{a_2}(t - {t_0}) + 3{a_3}{{(t - {t_0})}^2}} \\ {\ddot q(t) = 2{a_2} + 6{a_3}(t - {t_0})} \end{array}} \right.$

(4)

To meet the constraints of joint angle limitations and angular velocity restrictions at the initial and final time instants, it is possible to set constraints and solve for the four unknown coefficients in the cubic polynomial.

$\left\{ {\begin{array}{*{20}{c}} {q({t_0}) = {q_0}} \\ {q({t_f}) = {q_f}} \\ {\dot q({t_0}) = {{\dot q}_0}} \\ {\dot q({t_f}) = {{\dot q}_f}} \end{array}} \right.$

(5)

where t₀ and t_f represent the initial and final times for each segment, q₀ and q_f represent the initial and final angles, ${\dot q_0}$ and ${\dot q_f}$ represent the initial and final velocities, respectively. By obtaining the analytical expression for a_i, it can be substituted into the trajectory, velocity, and acceleration expressions for further calculation.

$\left\{ {\begin{array}{*{20}{c}} {{a_0} = {q_0}} \\ {{a_1} = {{\dot q}_0}} \\ {{a_2} = \frac{{3({q_f} - {q_0})}}{{{t_f} - {t_0}}} - \frac{{(2{{\dot q}_0} + {{\dot q}_f})}}{{{t_f} - {t_0}}}} \\ {{a_3} = - \frac{{2({q_f} - {q_0})}}{{{{({t_f} - {t_0})}^3}}} + \frac{{({q_0} + {q_f})}}{{{t_f} - {t_0}}}} \end{array}} \right.$

(6)

3.2. Quintic polynomial interpolation function

In certain specific working scenarios, there are requirements for the robot's motion trajectory to have constraints on acceleration. In these cases, the acceleration function of the cubic polynomial can change only at a fixed rate, following a first-order form. However, the cubic polynomial may not satisfy the requirement of continuous acceleration. To address this, a higher-order quintic polynomial interpolation method can be employed for trajectory planning. The expression for quintic polynomial interpolation is defined, and by taking derivatives, expressions for trajectory position, velocity, and acceleration can be obtained.

$\left\{ {\begin{array}{*{20}{c}} {q(t) = {a_0} + {a_1}(t - {t_0}) + {a_2}{{(t - {t_0})}^2} + {a_3}{{(t - {t_0})}^3} + {a_4}{{(t - {t_0})}^4} + {a_5}{{(t - {t_0})}^5}} \\ {\dot q(t) = {a_1} + 2{a_2}(t - {t_0}) + 3{a_3}{{(t - {t_0})}^2} + 4{a_4}{{(t - {t_0})}^3} + 5{a_5}{{(t - {t_0})}^4}} \\ {\ddot q(t) = 2{a_2} + 6{a_3}(t - {t_0}) + 12{a_4}{{(t - {t_0})}^2} + 20{a_5}{{(t - {t_0})}^3}} \end{array}} \right.$

(7)

where q(t) represents the joint angle trajectory, and a_i (i = 0, 1, …, 5) are the coefficients of the quintic polynomial. Similar to the cubic polynomial, the quintic polynomial takes into account the initial and final joint angle positions and the constraints on joint angle velocities. However, it also includes an additional equation to ensure continuous acceleration, denoted by $\ddot q({t_0}) = {\ddot q_0}$ , $\ddot q({t_f}) = {\ddot q_f}$ . By considering the constraint conditions, the expression for the polynomial coefficient parameters can be derived.

$\left\{ {\begin{array}{*{20}{c}} {{a_0} = {q_0}} \\ {{a_1} = {{\dot q}_1}} \\ {{a_2} = \frac{{{{\ddot q}_0}}}{2}} \\ {{a_3} = \frac{{20({q_f} - {q_0}) - (8{{\dot q}_f} + 12{{\dot q}_0})({t_f} - {t_0}) - 3({{\ddot q}_0} - {{\ddot q}_f}){{({t_f} - {t_0})}^2}}}{{2{{({t_f} - {t_0})}^3}}}} \\ {\begin{array}{*{20}{c}} {{a_4} = \frac{{ - 30({q_f} - {q_0}) + (14{{\dot q}_f} + 16{{\dot q}_0})({t_f} - {t_0}) + (3{{\ddot q}_0} - 2{{\ddot q}_f}){{({t_f} - {t_0})}^2}}}{{2{{({t_f} - {t_0})}^4}}}} \\ {{a_5} = \frac{{12({q_f} - {q_0}) - 6({{\dot q}_f} + {{\dot q}_0})({t_f} - {t_0}) + ({{\ddot q}_0} - {{\ddot q}_f}){{({t_f} - {t_0})}^2}}}{{2{{({t_f} - {t_0})}^5}}}} \end{array}} \end{array}} \right.$

(8)

3.3. Piecewise polynomial interpolation

In the case of multi-node trajectory interpolation, the constraints of continuous position, velocity, and acceleration can only be satisfied when at least five polynomials are used for cubic polynomial interpolation. When interpolating multi-segment trajectory points using quintic polynomials, each segment requires more constraints due to its higher order ^{[30,31,32,33]}. This results in a large amount of calculations and high time cost, making it unsuitable for tasks with fewer path points for small castings sorting robot arm. Therefore, a "3-5-3" mixed piecewise polynomial interpolation function is proposed by combining cubic and quintic polynomials. This approach not only addresses the issue of angle mutation in cubic polynomial interpolation programming but also solves the problem of excessive angular velocity in quintic polynomial interpolation programming while reducing calculation complexity and saving time. The work task of the sorting robot arm for small castings is simplified by converting it into four path points in Cartesian space. These path points are then converted into joint angles through inverse kinematics calculations (Section 2.1). The entire path is divided into three sections, each consisting of two adjacent path points. A general formula describing how joint angles vary with time under certain constraints is derived.

$\left\{ {\begin{array}{*{20}{c}} {{q_{i1}}(t) = {a_{i10}} + {a_{i11}}t + {a_{i12}}{t^2} + {a_{i13}}{t^3}\begin{array}{*{20}{c}} {}&{}&{}&{} \end{array}, t \in ({t_0}, {t_1})} \\ {{q_{i2}}(t) = {a_{i20}} + {a_{i21}}t + {a_{i22}}{t^2} + {a_{i23}}{t^3} + {a_{i24}}{t^4} + {a_{i25}}{t^5}, t \in ({t_1}, {t_2})} \\ {{q_{i3}}(t) = {a_{i30}} + {a_{i31}}t + {a_{i32}}{t^2} + {a_{i33}}{t^3}\begin{array}{*{20}{c}} {\begin{array}{*{20}{c}} {}&{} \end{array}}&{}&{} \end{array}, t \in ({t_2}, {t_3})} \end{array}} \right.$

(9)

In this context, q_i1(t), q_i2(t) and q_i3(t) represent the trajectories of the "3-5-3" segmented polynomial interpolation. The coefficients of each segment function are denoted by a_i, where i = 0, 1, …, n and n represent the number of joints.

$M = \left[ {\begin{array}{*{20}{c}} {{t_1}^3}&{{t_1}^2}&{t1}&1&0&0&0&0&{ - 1}&0&0&0&0&0 \\ {3{t_1}^2}&{2{t_1}}&1&0&0&0&0&{ - 1}&0&0&0&0&0&0 \\ {6{t_1}}&2&0&0&0&0&{ - 2}&0&0&0&0&0&0&0 \\ 0&0&0&0&{{t_2}^5}&{{t_2}^4}&{{t_2}^3}&{{t_2}^2}&{{t_2}}&1&0&0&0&{ - 1} \\ 0&0&0&0&{5{t_2}^4}&{4{t_2}^3}&{3{t_2}^2}&{2{t_2}}&1&0&0&0&{ - 1}&0 \\ 0&0&0&0&{20{t_2}^3}&{12{t_2}^2}&{6{t_2}}&2&0&0&0&{ - 2}&0&0 \\ 0&{}&{}& \cdots &{}&{}& \cdots &{}&{}&0&{{t_3}^3}&{{t_3}^2}&{{t_3}}&1 \\ 0&{}&{}& \cdots &{}&{}& \cdots &{}&{}&0&{3{t_3}^2}&{2{t_3}}&1&0 \\ 0&{}&{}& \cdots &{}&{}& \cdots &{}&{}&0&{6{t_3}}&2&0&0 \\ 0&0&0&1&0&{}&{}& \cdots &{}&{}& \cdots &{}&{}&0 \\ 0&0&1&0&0&{}&{}& \cdots &{}&{}& \cdots &{}&{}&0 \\ 0&1&0&0&0&{}&{}& \cdots &{}&{}& \cdots &{}&{}&0 \\ 0&0&0&0&0&0&0&0&0&0&0&0&0&1 \\ 0&0&0&0&0&0&0&0&0&1&0&0&0&0 \end{array}} \right]$

(10)

The joint angles of the four intermediate points, as well as the velocity and acceleration of both the initial and end points, are all known to be zero. Additionally, there is continuity in velocity and acceleration between adjacent path points. To find the 14 unknown coefficients, a combination of cubic polynomials and quintic polynomials is used to derive polynomial coefficients. M represents a polynomial matrix, where t₁, t₂, t₃ denote the time used by three respective polynomials. q₀, q₁, q₂, q₃ represent trajectory joint angles for each of the four waypoints, respectively. A denotes a matrix composed of unknown parameter a_i. By multiplying matrix M with matrix Q, we obtain a matrix equation that can solve for the coefficient a_i.

$Q = {\left[ {\begin{array}{*{20}{c}} 0&0&0&0&0&0&{{q_3}}&0&0&{{q_0}}&0&0&{{q_2}}&{{q_1}} \end{array}} \right]^T}$

(11)

$A = {M^{ - 1}}*Q$

(12)

4. Genghis Khan shark optimizer (GKSO)

4.1. GKSO background

The Genghis Khan shark optimization algorithm is a population-based optimization algorithm that takes inspiration from the individual and collective behaviors of Genghis Khan sharks in their oceanic habitat. It is designed to emulate their hunting, movement, foraging, and self-protective escape behaviors, which collectively form the basis of a comprehensive mathematical model for optimization. These four biological behaviors provide mathematical expressions for the four stages of the algorithm. The diagram below illustrates the four behaviors of Genghis Khan sharks ^[24].

Figure 3. Diagram of the four biological behaviors of the Genghis Khan shark.

DownLoad: Full-Size Img PowerPoint

4.2. Mathematical model

4.2.1. Hunting stage

In order to ensure their safety and effectively search for food in unfamiliar marine environments, Genghis Khan sharks tend to linger near the seabed, conducting broad spatial searches. During the initial phase, the school of sharks patrols in different directions until they locate the optimal position for their target prey. This space is bounded by upper and lower limits, within which random positions are marked as optimal. The mathematical formula for initializing and updating these positions is as follows:

$X_i^{\text{j}}(t + 1) = X_i^j + \frac{{l{b_j} + rand*(u{b_j} - l{b_j})}}{{iter}}, i = 1, 2, \cdots , N, j = 1, 2, \cdots , D, iter = 1, 2, \cdots , T.$

(13)

Where X_i^j(t + 1) represents the position of the i-th individual in the j-th dimension at time t + 1, ub_j and lb_j are the upper and lower bounds in the j-th dimension, rand is a random number in the range of [0, 1], N represents the total population size, D represents the problem dimension, iter denotes the current iteration number, and T is the maximum number of iterations.

4.2.2. Moving stage

After identifying the initial optimal position, Genghis Khan sharks rely on their exceptional olfactory senses to progressively close in on the prey's location. The mathematical formula for their movement towards the target position is as follows:

$\widehat X_i^j(t + 1) = s*(X_{best}^j(t) - X_i^j(t))$

(14)

In the equation, s represents the olfactory influence factor during the movement towards the target, X_best^j represents the optimal target position at time t, and s is determined by the concentration of odor emitted by the target prey. The mathematical model associated with this is as follows:

$s = m{I^r}$

(15)

Among these parameters, m is a non-negative constant that holds special significance as its value directly affects the overall convergence speed of the GKSO algorithm. After thorough comparison, it has been determined that setting m to 1.5 yields optimal algorithm performance. Another parameter, I, represents the attribute intensity, which is contingent upon the individual's capabilities. Additionally, r is a random number within the range of [0, 1] and reflects the individual's degree of absorption towards odor information. The value of r directly influences the behavior of GKSO. When r = 0, it signifies that the algorithm has not detected the target prey and should continue exploring outward. Conversely, when r = 1, it indicates complete absorption of the odor, prompting the algorithm to move directly towards the target prey. At this point, the algorithm is likely to converge to the optimal solution. However, it is noteworthy that when r = 1, the value of s may also be relatively high. While this brings the algorithm closer to the optimal solution, it simultaneously reduces the overall exploration capacity. Furthermore, there is a possibility that a newly discovered optimal solution may surpass the current global optimum, limiting the algorithm's ability to further evolve. Hence, the choice of r holds significant importance.

To better approach the optimal solution, the algorithm retains the position of the previous best solution during the optimization process and updates the positions of other best solutions based on it. The position update formula is as follows:

$X_i^j(t + 1) = \frac{{\widehat X_i^j(t + 1) + X_{i - 1}^j(t)}}{2}$

(16)

4.2.3. Foraging stage

During the foraging process, Genghis Khan sharks adopt a parabolic-shaped attack strategy for cooperative hunting, as shown in the diagram. The position update formula for this strategy is as follows:

$X_i^j(t + 1) = X_{best}^j(t) + rand*(X_{best}^j(t) - X_i^j(t)) + \lambda *{p^2}*(X_{best}^j(t) - X_i^j(t))$

(17)

Among them, the value of λ is a random number that can take either 1 or -1. The parameter p controls the step size of Genghis Khan sharks during the foraging phase. A larger value of p tends to favor global exploration, while gradually reducing the value of p shifts the focus towards local exploitation. The calculation formula for p is as follows:

$p = 2*\left\{ {1 - {{\left( {\frac{t}{T}} \right)}^{\frac{1}{4}}} + \left| {\omega (t + 1)} \right|*[{{(\frac{t}{T})}^{\frac{1}{4}}} - {{(\frac{t}{T})}^3}]} \right\}$

(18)

The weight coefficient |ω(t + 1)| at time t + 1 is calculated using the following formula:

$\left| {\omega (t + 1)} \right| = 1 - 2{\omega ^4}(t)$

(19)

The initial value of ω(0) is typically set to 0.1.

4.2.4. Self-protection stage

During foraging, Genghis Khan sharks may encounter factors that pose a threat to their safety. In such situations, they exhibit a behavior of lightening their body color and swiftly escaping. The mathematical model for this behavior is as follows:

$\left\{ {\begin{array}{*{20}{c}} {X_i^j(t + 1) = X_i^j(t) + {k_1}({z_1}X_{best}^j(t) - {z_2}X_k^j(t)) + {k_2}\rho ({a_3}(X2_i^j(t) - X1_i^j(t))) + \frac{{{a_2}}}{2}(X_{u1}^j(t) - X_{u2}^j(t)){\text{ }}if{\text{ }}{z_1} < 0.5} \\ {X_i^j(t + 1) = X_{best}^j(t) + {k_1}({z_1}X_{best}^j(t) - {z_2}X_k^j(t)) + {k_2}\rho ({z_3}(X2_i^j(t) - X1_i^j(t))) + \frac{{{z_2}}}{2}(X_{u1}^j(t) - X_{u2}^j(t)){\text{ otherwise}}} \end{array}} \right.$

(20)

Among them, k₁ is a random number between -1 and 1, and k₂ follows a normal distribution with a mean of 0 and a standard deviation of 1. z₁, z₂, and z₃ are three random numbers. The calculation formula is as follows:

$\left\{ {\begin{array}{*{20}{c}} {{z_1} = {l_1}*2*rand + (1 - {l_1})} \\ {{z_2} = {l_1}*rand + (1 - {l_1})} \\ {{z_3} = {l_1}*rand + (1 - {l_1})} \end{array}} \right.$

(21)

The value of l₁ is a binary random number that can take either 0 or 1. ρ is an adaptive coefficient, and its calculation formula is as follows, which is related to the calculation of α and β:

$\alpha = \left| {\beta *\sin (\frac{{3\pi }}{2} + \sin (\frac{{3\pi \beta }}{2}))} \right|$

(22)

$\beta = {\beta _{\min }} + ({\beta _{\max }} - {\beta _{\min }})*{(1 - {(\frac{{iter}}{T})^3})^2}$

(23)

$\rho = \alpha *(2*rand - 1)$

(24)

β_min is 0.2 and β_max is 1. The formulas for calculating X1_i^j and X2_i^j in the equation are as follows:

$X1_i^j(t) = l{b_j} + rand*(u{b_j} - l{b_j})$

(25)

$X2_i^j(t) = l{b_j} + rand*(u{b_j} - l{b_j})$

(26)

The formula for calculating X_k^j(t) is as follows:

$X_k^j(t) = {l_2}*(X_p^j(t) - X_r^j(t)) + X_r^j(t)$

(27)

Among them, X_p^j(t) represents the randomly selected solution at time t (p ∈ 1, 2, ..., N). Similarly, l₂ and l₁ are binary random numbers that can take either 0 or 1. X_r^j(t) is a set of randomly generated initial solutions at time t. X_u1^j and X_u2^j are two randomly selected solutions. By decomposing the position update expression, we can see that the self-defense phase actually involves searching the space near the optimal solution using multiple solutions. This can be considered as a form of neighborhood search. This self-defense mechanism helps to prevent the GKSO algorithm from getting trapped in local optima and improves exploration efficiency.

4.2.5. Summary of the algorithm

The complexity of this algorithm primarily depends on the initialization of random solutions, position updates, and fitness evaluation. In GKSO, there are stages where new initialization updates occur, which enhance global search capability by introducing random solutions for position updates. The position update formula at each step is subject to upper and lower bounds, and the generation of certain solutions relies on constraints. The specific constraints are tailored to the characteristics of the function problem. In this study, we optimize and impose constraints on the position, velocity, and acceleration. The specific constraint conditions are designed based on the fitness function, but they will not be discussed in detail here.

To better illustrate the structure of the GKSO algorithm, a flowchart has been created, as shown in Figure 4.

Figure 4. GKSO algorithm flow chart.

DownLoad: Full-Size Img PowerPoint

5. Experiment

5.1. Objective function

Based on Chapter 3, an interpolation function for the robotic arm's trajectory can be derived. However, in practical trajectory planning, explicit time intervals are not provided, making it challenging to determine the unknown coefficients for each interpolation segment. To address this issue, the GKSO algorithm is employed in this study to determine the optimal interpolation times that satisfy velocity and acceleration constraints.

During the trajectory optimization process, an effective approach is to directly optimize the time variable, t. This reduces the dimensionality of the objective function, significantly reducing the computational complexity of the GKSO algorithm. The fitness function and boundary constraints established for the optimization process are as follows:

$\begin{array}{*{20}{c}} {{T_{all}} = {t_1} + {t_2} + {t_3} = \sum\limits_{j = 0}^3 {{t_j}} } \\ {\left\{ {\begin{array}{*{20}{c}} {\left| {{q_{ij}}(t)} \right| \leqslant Q\max , j = 1, 2, 3} \\ {\begin{array}{*{20}{c}} {\left| {{{\dot q}_{ij}}(t)} \right| \leqslant V\max , j = 1, 2, 3} \\ {\left| {{{\ddot q}_{ij}}(t)} \right| \leqslant A\max , j = 1, 2, 3} \end{array}} \end{array}} \right.} \end{array}$

(28)

T_all represents the total operation time of the robotic arm, while t₁, t₂, and t₃ denote the running times of the three interpolation functions. The objective function is designed to comply with boundary constraints. By utilizing the "3-5-3" interpolation function described in Chapter 3, we derive the problem matrix and simultaneously optimize the position, velocity, and acceleration. During the experimental process, specific constraint conditions are set, and adjustments can be made to accommodate portions that exceed the boundaries in the calculation of the relevant fitness function.

In this study, our main goal is to minimize the total time, with shorter durations considered more optimal. For portions that exceed the boundaries, the GKSO algorithm replaces them with boundary values during the hunting phase. Subsequent position updates depend on the design of the fitness function. In this study, we handle portions that exceed the boundaries by replacing their fitness values with infinity, ensuring that they are not included in the point position list during subsequent updates. This approach allows for the overall optimization calculation to proceed without interference.

5.2. Work task

In order meet the requirements of a small-scale casting sorting robotic arm, specific motion trajectories need to be defined. The current robotic arm used in casting production factories primarily focuses on the task of transporting and sorting castings that have been piled up on a conveyor belt. It operates within the workspaces of machine tools, conveyor belts, and processing areas. The workspace is determined based on known joint parameters, and relative task objectives have been established, as depicted in Figure 5. The end effector is required to move sequentially through four designated waypoints, labeled A, B, C, and D, and place the items in the adjacent processing area. It then follows the same path back for sorting purposes. Point A is located on the conveyor belt. Points B and C are located closer to the robotic arm, about 400 mm apart from each other with some height difference between the two points. Point D is located at the target position. As shown in Figure 5, the total shortest linear distance is about 2736 mm. The expected target is to make the trajectory as close to the shortest path as possible while keeping the trajectory smooth without shock or vibration during operation.

Figure 5. Work task path.

DownLoad: Full-Size Img PowerPoint

The coordinates of these four waypoints are calculated using the robot kinematics principles outlined in Chapter 2, converting them from Cartesian coordinates to joint angle space. The resulting joint angles are as follows.

5.3. Experimental simulation

Based on the given DH parameters, a model of the casting sorting robotic arm was constructed. Four provided waypoints were used as reference points, and Table 1 specifies the angle, angular velocity, and angular acceleration constraints for each joint. To facilitate the experiment, the average values were used. Considering time optimization, a "3-5-3" polynomial interpolation with significant fluctuations in the acceleration at the transition points was selected to avoid detrimental vibrations to the robotic arm. The maximum allowable joint velocity was set at 1.5 rad/s, and the maximum allowable joint acceleration was set at 2 rad/s². The population size was set to 50, and the maximum number of iterations was set to 50.

Assuming initial times t₁, t₂, and t₃ are 3 seconds, 8 seconds and 3 seconds respectively, a comparative analysis was conducted between the results obtained from the "3-5-3" configuration without PSO planning, the traditional PSO method, the improved PSO algorithm (referred to as I-PSO), and the GKSO algorithm. The PSO algorithm parameters were set as follows: the inertial influence factor was set to 0.9, and the individual and social learning factors were both set to 2. In the I-PSO algorithm, a linearly decreasing inertia weight and dynamic learning factor improvement were implemented ^{[34,35,36,37,38,39,40,41,42,43]}. Since the interpolation coefficients of the polynomial segments are time-dependent, each optimization may yield different combinations. Therefore, after conducting multiple experiments, an optimal time solution graph was obtained. The total time obtained from multiple iterations was compared in Table 4 for a comprehensive evaluation.

Table 3. Joint angular path points.

	1	2	3	4	5	6
A	-1.57	-0.79	0.17	0.35	0.52	0.70
B	-0.79	-1.05	-0.34	0	0.17	-0.35
C	0.61	-0.52	-0.61	-0.35	0.35°	0.17
D	1.40	0.17	-1.05	0.26	0.52	0.52

| Show Table

DownLoad: CSV

Table 4. Comparison of initial solution computation.

	Experiment number	Joint 1	Joint 2	Joint 3	Joint 4	Joint 5	Joint 6
PSO algorithm Initial time solution (s)	1	8.4098	7.1704	8.9544	8.1209	7.8547	8.8324
	2	7.9198	6.6726	8.2748	7.9653	7.7378	8.6228
	3	8.7916	7.6681	7.5922	9.0822	6.5264	9.1036
	4	8.1696	6.1719	7.6175	8.4861	7.6593	9.8066
	5	7.5208	6.9103	8.8100	7.1629	6.1390	9.7259
	6	6.9100	8.1284	8.0368	6.8938	7.0409	9.2878
	7	7.7727	7.2811	7.5904	9.0667	6.1309	10.1242
	8	7.9547	7.1014	7.3895	7.2746	6.9102	9.0220
IPSO algorithm Initial time solution (s)	1	6.5781	6.6019	8.2768	9.0012	5.8138	10.0710
	2	7.5423	6.4004	7.4877	7.1267	7.1999	8.8870
	3	6.7233	6.5709	7.9847	8.6081	5.4322	7.5025
	4	8.3360	6.2058	7.2968	8.2743	5.5754	9.3354
	5	6.2108	6.8971	7.8647	7.0906	6.0873	8.9467
	6	6.3753	7.3511	7.8584	7.5814	4.9933	8.7924
	7	8.1775	7.2881	6.4751	6.8552	5.6674	9.4112
	8	8.2808	6.6139	6.1443	7.2670	7.6661	7.9648
GKSO algorithm Initial time solution (s)	1	5.5387	6.1930	7.5979	6.3257	4.8661	7.9096
	2	5.1007	6.9222	7.2373	7.0316	4.9755	7.4689
	3	4.9898	6.4846	7.4874	6.1898	5.4697	8.0156
	4	6.1106	6.0770	7.3171	6.6270	4.8989	7.3879
	5	4.7932	7.0215	7.1610	7.5530	4.9587	7.7137
	6	4.9024	6.5184	7.1402	6.0367	5.5395	7.8925
	7	5.9273	6.6714	7.2846	7.1147	5.3979	8.0432
	8	5.2854	6.6528	7.1431	6.8349	4.7269	7.8857

| Show Table

DownLoad: CSV

In Figures 6–8, a) represents trajectory planning using only the "3-5-3" segmented interpolation, b) represents the traditional PSO algorithm, c) represents the I-PSO algorithm, and d) represents the GKSO algorithm. From the trajectory, velocity, and acceleration, it is evident that the GKSO algorithm significantly shortens the time required to reach the target waypoints. The original total running time for a single joint is 18 seconds, but the optimized total running time can be reduced to a minimum of 4.109 seconds, which represents an optimization rate of approximately 77.2% compared to the unoptimized case. When compared to the PSO algorithm, the GKSO algorithm shows superior optimization results. Analyzing multiple joints, apart from joint 3 where the I-PSO algorithm outperforms, the difference between the GKSO algorithm and the I-PSO algorithm is relatively small for the other joints. However, specific numerical comparisons show that the GKSO algorithm exhibits slightly better optimization results. The trajectory and velocity obtained from the "3-5-3" interpolation planning demonstrate excellent continuity overall. Although the optimized positions show slight variations compared to the original positions, the velocities and accelerations remain within acceptable ranges, closely approaching V_max and a_max, with minimal fluctuations. The trajectory appears smooth overall, satisfying the constraints of each joint, and confirming the feasibility and effectiveness of the algorithm in enhancing the efficiency of the robotic arm.

Figure 6. Comparison of trajectory and time change.

DownLoad: Full-Size Img PowerPoint

Figure 7. Velocity versus time change.

DownLoad: Full-Size Img PowerPoint

Figure 8. Acceleration and time change.

DownLoad: Full-Size Img PowerPoint

Here is a smoothed version:

Based on comparisons of the trajectory data, in the PSO algorithm joint 3 reaches the target position first, followed by joints 5, 1, 2, 4, and 6, with a final time of 7.45 seconds. The reaching order is the same for the I-PSO algorithm. The running time for joint 5 is optimized to approximately 4.49 seconds in the PSO algorithm and 4.48 seconds in the I-PSO algorithm. In the GKSO algorithm, the running time is around 4.44 seconds, showing little difference. The optimization of joints 3 and 6 results in more noticeable changes to running time. Joint 6 has the longest running time of 7.28 seconds. With the GKSO algorithm, joint 5 is the first to complete its motion, followed by joints 1, 2, 5, 3, and 6, with an overall running time for joint 6 of 7.10 seconds - shorter than the optimization times of the other two algorithms.

Each joint reduces a certain amount of running time. For example, joints 1, 2, 4 and 5 reach their positions ahead of the equivalent interpolation points for the other joints during runtime. After arriving at the position, they no longer change and maintain the joint angle, waiting for the next interpolation point to arrive. Finally, the remaining time is spent waiting for the last joint to complete its motion and finish rotating the joint angle.

In the velocity comparison graph, between 1.75 s and 3.00 s, joint 1 exhibits more noticeable changes. Under algorithm optimization, the difference between other joints except for joints 1 and 3 was not significant. However, the optimization time of the GKSO algorithm was slightly better. The GKSO algorithm displays smaller fluctuations and a smaller maximum amplitude in comparison to the other two algorithms. After optimization using the GKSO algorithm, the maximum speed was reached at 1.97 s and 2.97 s, approximately 1.47 rad/s. For the PSO algorithm, the maximum speed of 1.4 rad/s was reached at 2.24 s. As for the I-PSO algorithm, the maximum speeds of approximately 1.77 rad/s were reached around 1.85 s and 2.74 s. In terms of velocity fluctuation, the GKSO algorithm and the optimized PSO algorithm exhibit a bimodal pattern, while the IPSO algorithm follows a unimodal pattern. The PSO and IPSO algorithms demonstrate larger variations, whereas the GKSO algorithm shows smaller variations. The optimization objective of the employed algorithm is the total running time of each joint. Thus, while shortening the time and adhering to velocity and acceleration constraints, the inherent nature of the "3-5-3" polynomial interpolation function may lead to noticeable fluctuations at the connection points. Nevertheless, by setting lower acceleration constraint values, the vibrations caused by fluctuations in the robotic arm are reduced, and even the maximum vibration does not impede the arm's operation. To mitigate unnecessary vibrations, non-uniform B-spline optimization is applied at the connection points. As illustrated in the Figure 8, the unoptimized acceleration graph demonstrates desirable continuity, with minimal numerical fluctuations within an acceptable range. Building upon optimized time, the PSO and IPSO algorithms exhibit substantial variations in each joint, particularly joint 3, during velocity analysis. While the GKSO algorithm performs comparatively worse in terms of optimized time, it showcases superior stability and smaller fluctuation amplitudes. Considering each joint, the GKSO algorithm yields more stable acceleration variations, characterized by smooth curves devoid of inflection points. When completing the same route task, the GKSO algorithm attains shorter time durations, more stable velocities, and reduced vibrations, thereby exemplifying superior algorithmic performance. Similarly, after the trajectory is completed within the optimized runtime, the angular position is not changed. The subsequent velocity and acceleration are both zero.

The convergence graphs in Figure 9, a) and b) represent the convergence of time and iteration for different joints. The graphs show the convergence of GKSO, traditional PSO, and I-PSO algorithms in optimizing time. From the graphs, it is evident that the GKSO algorithm converges the fastest, followed by the I-PSO algorithm. The GKSO algorithm demonstrates excellent overall convergence, almost reaching the optimal solution, indicating a significant improvement in convergence compared to the other algorithms. Furthermore, the improved algorithm outperforms the traditional PSO algorithm by achieving smaller optimal time solutions and avoiding the trap of local optima. Moreover, the GKSO algorithm shows better optimization than the I-PSO algorithm. After 50 iterations, the convergence is already close to the optimal solution, and the GKSO algorithm converges faster. Compared to the traditional PSO and I-PSO algorithms, the GKSO algorithm achieves smaller optimal time solutions, making it a superior algorithm.

Figure 9. Variations in partial joint runtime with iteration number.

DownLoad: Full-Size Img PowerPoint

To validate the GKSO algorithm's capability in seeking initial optimal solutions, 8 groups of experiments were conducted as shown in Table 4. Under equal population sizes and iteration counts, the solutions were obtained and the times of the initial solutions from the first calculation were recorded for comparison. As shown, the GKSO algorithm generated initial calculation times that were mostly better than PSO and I-POS, with a small minority slightly lower possibly due to random numbers coincidentally being selected close to optimal. The results not only demonstrate GKSO's advantage in the early stage but also validate the algorithm's robustness.

It is challenging to discern the time differences between the traditional PSO algorithm and the improved algorithm from the trajectory, velocity, and acceleration graphs. To compare the differences between the two algorithms, the initial times for the cubic polynomial interpolation are adjusted to 5 s, while the initial times for the quintic polynomial interpolation are set to 8 s. Table 5 provides a separate breakdown of the running times for each joint to facilitate comparison. Additionally, multiple experiments are conducted to compare the results for different iteration numbers and population sizes. Based on multiple comparisons, the GKSO algorithm and the I-PSO algorithm show minimal differences. The optimal solutions for joints 1, 2, 4, and 6 differ by no more than 0.6 s. However, overall, the GKSO algorithm achieves slightly better optimization objectives. For joint 3, the GKSO algorithm successfully avoid local optima through its defensive measures, resulting in an optimal time solution with minimal fluctuations and demonstrating good performance in terms of acceleration. Furthermore, as the iteration numbers and population sizes increase, the optimal solutions show a decreasing trend, further enhancing the effectiveness of the optimization algorithm.

Table 5. Iterative optimization time comparison.

	Algorithm	1	2	3	4	5	6
Population size 50, iterations 50	Basical PSO (time/s)	4.72	5.91	5.10	6.39	4.70	7.35
	Improved PSO (time/s)	4.61	5.73	4.75	6.21	4.42	7.38
	GKSO (time/s)	4.53	5.68	5.81	6.01	4.41	7.29
Population size 50, iterations 100	Basical PSO (time/s)	4.69	5.79	5.20	6.31	4.83	7.31
	Improved PSO (time/s)	4.39	5.71	4.74	5.98	4.36	7.23
	GKSO (time/s)	4.33	5.70	5.42	5.72	4.22	7.16
Population size 70, iterations 50	Basical PSO (time/s)	4.65	5.81	5.06	6.29	4.71	7.29
	Improved PSO (time/s)	4.31	5.66	4.75	6.03	4.43	7.25
	GKSO (time/s)	4.27	5.32	5.41	5.98	4.39	7.11

| Show Table

DownLoad: CSV

5.4. Experiment

To validate the practical applicability of the optimization algorithm in real-world scenarios for the robotic arm, the algorithm's pre- and post-optimized trajectory points are implemented in the casting sorting robotic arm. The trajectory plot depicted in the figure illustrates the simulated trajectories in MATLAB before and after optimization using the GKSO and PSO algorithms. The blue line represents the trajectory obtained from the GKSO algorithm, the black line represents the trajectory obtained from the I-PSO algorithm, and the red line represents the trajectory obtained from the traditional PSO optimization.

It is evident that, in the case of the I-PSO algorithm and PSO optimization, the middle section of the trajectory is longer. The trajectory lines are divided into multiple points and fed into the robotic arm for actual operation. The trajectory lengths optimized by different algorithms through 30, 50,100,150, and 200 iterations of optimized runtime were respectively calculated, as shown in Figure 11. All algorithms exhibited rapid convergence behavior with increasing iterations. Among them, the initial solution of the GKSO algorithm was better in the first 150 iterations, with convergence closer to the optimal solution. After 200 iterations, there was little difference in the overall optimization effect of the algorithms on this robotic arm trajectory planning problem. At this point, they all approached the shortest trajectory length, which to some extent validated the certain advantages of the GKSO algorithm in trajectory planning. To account for the challenge of real-time adjustment of acceleration in the control segment, linear averaging is applied between each acceleration and deceleration segment. Three experiments were conducted without specifying interpolation time, prioritizing safety by setting the extremum at 30% of the robotic arm's maximum speed and acceleration. The trajectories obtained after 30, 60, and 100 iterations are imported into the robotic arm for sorting, as shown in Figure 12. The results are documented in Table 6. The overall operation unfolds smoothly without significant fluctuations, and there are no instances of jamming at the transition points of the trajectory, which further solidifies the reliability of the desired trajectory.

Figure 10. Trajectory simulation after algorithm optimization.

DownLoad: Full-Size Img PowerPoint

Figure 11. Trajectory length comparison data graph.

DownLoad: Full-Size Img PowerPoint

Figure 12. Experiment of robotic arm trajectory operation.

DownLoad: Full-Size Img PowerPoint

Table 6. Trajectory running time comparison.

Iterations	30	60	100
GKSO	39.87 s	36.57 s	33.05 s
Improved PSO	43.08 s	38.85 s	33.59 s
PSO	54.72 s	48.89 s	38.46 s

| Show Table

DownLoad: CSV

In order to consider the runtime of the algorithm itself, a timing function is implemented to measure the computation time for obtaining the optimal solution. With 50 iterations and a population size of 50, multiple optimizations are performed, and the average runtime is calculated as shown in the Table 7. The overall difference in runtime is within 2 seconds. Furthermore, after incorporating dynamic learning factors and adaptive inertia weights, the difference in runtime for the I-PSO algorithm is within 1 second. Although there is a slight disparity, the time saved from multiple repetitions of optimization can offset this relatively small difference in industrial production settings.

Table 7. The running time of the algorithm for each joint.

	GKSO algorithm runtime (s)	PSO algorithm runtime (s)	I-PSO algorithm runtime (s)
Joint 1	7.939962	5.004269	6.093648
Joint 2	7.777234	5.16282	6.37203
Joint 3	7.768724	5.100063	6.552508
Joint 4	7.774511	6.36556	6.872196
Joint 5	7.958547	5.187522	6.685268
Joint 6	8.028946	6.087946	6.443644

| Show Table

DownLoad: CSV

It is important to note that we did not focus on optimizing and simplifying the GKSO algorithm itself, resulting in the use of code that may not be the most concise. Additionally, in cases where multiple nested formulas are present, parallel computing cannot be utilized to reduce runtime.

6. Conclusions

We present a time-optimized trajectory planning method for the casting sorting robotic arm using the GKSO (Genghis Khan Shark Optimization) algorithm. We utilize the Monte Carlo method to compute the robotic arm's workspace and employ a "3-5-3" segmented polynomial interpolation technique for end-effector trajectory planning. Furthermore, we compare and analyze the simulated position, velocity, and acceleration profiles of the trajectories optimized using the GKSO algorithm, the traditional PSO (Particle Swarm Optimization) algorithm, and the I-PSO (Improved Particle Swarm Optimization) algorithm. The results confirm that the GKSO algorithm outperforms the other two algorithms, yielding the best time optimization results.

Moreover, the GKSO algorithm exhibits remarkable continuity in the optimized trajectories, adhering to the robotic arm's kinematic constraints. Compared to the other algorithms, the GKSO algorithm demonstrates lower fluctuations, higher stability, and enhanced reliability. Finally, when the calculated trajectory points are implemented in the actual operation of the robotic arm, minimal fluctuations are observed. In conclusion, qw successfully achieve the objective of optimizing the time-optimal trajectories for the casting sorting robotic arm.

In terms of future research directions, there are several potential areas to explore. First, as the algorithm is in the early stages of research, there may be a tendency for the generation of initial solutions to be concentrated, thereby limiting the algorithm's global search capability in the initial phase. To address this, one possible direction is to consider the integration of chaotic mapping techniques. This could help improve the generation of initial solutions and enhance the algorithm's ability to explore the global solution space.

Second, further studies can be conducted to investigate and optimize the parameters associated with the four stages of the algorithm. By refining and fine-tuning these parameters, it is possible to improve the overall performance and efficiency of the algorithm.

Additionally, there is research value in exploring strategies to optimize and simplify the algorithm itself. By streamlining the algorithm, it can become more efficient and easier to implement in practical applications.

Furthermore, in real-world industrial production scenarios, there may be situations that involve multiple points and requirements beyond the scope of this study. Therefore, it is worth exploring how the algorithm can be further optimized to handle such multi-point scenarios and provide accurate and effective solutions.

Use of AI tools declaration

The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

Acknowledgments

This paper was supported by Anhui Province Graduate Education Quality Project (2022cxcysj107) and Natural Science Foundation of Anhui Province: 2208085ME128.

Conflict of interest

The authors declare that there are no conflicts of interest.

Acknowledgments

The authors thank Universiti Sains Malaysia (USM) for the support in granting permission to conduct the research. We thank Universiti Sains Malaysia (USM) libraries (Perpustakaan Hamdan Tahir and Perpustakaan Hamzah Sendut), Universiti Putra Malaysia (UPM) library (Perpustakaan Sultan Abdul Samad), and Universiti Malaya (UM) Main Library for providing the resources to write the systematic review. We also appreciate all the scientists whose works contributed to this review article.

Funding

This article was funded by the Ministry of Higher Education Malaysia for Fundamental Research Grant Scheme with Project Codes: FRGS/2/2013/SKK01/USM/03/1.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

M.S and G.R: literature review search and write original draft; S.C.T: literature review search, review and edit original draft. All authors have read and agreed to the published version of the manuscript.

References

[1]	D'Alessio A, Proietti G, Sica G, et al. (2019) Pathological and molecular features of glioblastoma and its peritumoral tissue. Cancers 11: 469. https://doi.org/10.3390/cancers11040469
[2]	Arbab AS, Ali MM (2020) Glioblastoma: Targeting Angiogenesis and Tyrosine Kinase Pathways. Novel Approaches Cancer Study 4: 398. https://doi.org/10.31031/nacs.2020.04.000596
[3]	Davis ME (2016) Glioblastoma: Overview of disease and treatment. Clin J Oncol Nurs 20: 1-8. https://doi.org/10.1188/16.CJON.S1.2-8
[4]	Miller KD, Ostrom QT, Kruchko C, et al. (2021) Brain and other central nervous system tumor statistics, 2021. CA-Cancer J Clin 71: 381-406. https://doi.org/10.3322/caac.21693
[5]	Batash R, Asna N, Schaffer P, et al. (2017) Glioblastoma Multiforme, Diagnosis and Treatment; Recent Literature Review. Curr Med Chem 24: 3002-3009. https://doi.org/10.2174/0929867324666170516123206
[6]	Silantyev AS, Falzone L, Libra M, et al. (2019) Current and Future Trends on Diagnosis and Prognosis of Glioblastoma: From Molecular Biology to Proteomics. Cells 8: 863. https://doi.org/10.3390/cells8080863
[7]	Ma C, Nguyen HPT, Jones JJ, et al. (2022) Extracellular Vesicles Secreted by Glioma Stem Cells Are Involved in Radiation Resistance and Glioma Progression. Int J Mol Sci 23: 2770. https://doi.org/10.3390/ijms23052770
[8]	Vogelbaum MA (2012) Does extent of resection of a glioblastoma matter?. Clin Neurosur 59: 79-81. https://doi.org/10.1227/NEU.0b013e31826b2e75
[9]	Perrini P, Gambacciani C, Weiss A, et al. (2017) Survival outcomes following repeat surgery for recurrent glioblastoma: a single-center retrospective analysis. J Neuro-Oncol 131: 585-91. https://doi.org/10.1007/s11060-016-2330-7
[10]	Wann A, Tully PA, Barnes EH, et al. (2018) Outcomes after second surgery for recurrent glioblastoma: a retrospective case-control study. J Neuro-Oncol 137: 409-415. https://doi.org/10.1007/S11060-017-2731-2
[11]	Montemurro N, Fanelli GN, Scatena C, et al. (2021) Surgical outcome and molecular pattern characterization of recurrent glioblastoma multiforme: A single-center retrospective series. Clin Neurol Neurosur 207: 106735. https://doi.org/10.1016/j.clineuro.2021.106735
[12]	Sacko O, Benouaich-Amiel A, Brandicourt P, et al. (2021) The impact of surgery on the survival of patients with recurrent glioblastoma. Asian J Neurosur 16. https://doi.org/10.4103/ajns.ajns_180_20
[13]	Brennan PM, Borchert R, Coulter C, et al. (2021) Second surgery for progressive glioblastoma: a multi-centre questionnaire and cohort-based review of clinical decision-making and patient outcomes in current practice. J Neuro-Oncol 153. https://doi.org/10.1007/s11060-021-03748-0
[14]	Robin AM, Lee I, Kalkanis SN (2017) Reoperation for Recurrent Glioblastoma Multiforme. Neurosurg Clin N Am 28: 407-28. https://doi.org/10.1016/J.NEC.2017.02.007
[15]	Stupp R, Mason WP, van den Bent MJ, et al. (2005) Radiotherapy plus Concomitant and Adjuvant Temozolomide for Glioblastoma. New Engl J Med 352: 987-996. https://doi.org/10.1056/nejmoa043330
[16]	Fan CH, Liu WL, Cao H, et al. (2013) O6-methylguanine DNA methyltransferase as a promising target for the treatment of temozolomide-resistant gliomas. Cell Death Dis 4: e876. https://doi.org/10.1038/cddis.2013.388
[17]	Strobel H, Baisch T, Fitzel R, et al. (2019) Temozolomide and Other Alkylating Agents in Glioblastoma Therapy. Biomedicines 7: 69. https://doi.org/10.3390/biomedicines7030069
[18]	Auffinger B, Spencer D, Pytel P, et al. (2015) The role of glioma stem cells in chemotherapy resistance and glioblastoma multiforme recurrence. Expert Rev Neurother 15: 741-752. https://doi.org/10.1586/14737175.2015.1051968
[19]	Alves AL V, Gomes INF, Carloni AC, et al. (2021) Role of glioblastoma stem cells in cancer therapeutic resistance: a perspective on antineoplastic agents from natural sources and chemical derivatives. Stem Cell Res Ther 12: 206. https://doi.org/10.1186/s13287-021-02231-x
[20]	Singh N, Miner A, Hennis L, et al. (2021) Mechanisms of temozolomide resistance in glioblastoma - a comprehensive review. Cancer Drug Resist 4: 17-43. https://doi.org/10.20517/cdr.2020.79
[21]	Armocida D, Frati A, Salvati M, et al. (2020) Is Ki-67 index overexpression in IDH wild type glioblastoma a predictor of shorter Progression Free survival? A clinical and Molecular analytic investigation. Clin Neurol Neurosur 198: 106126. https://doi.org/10.1016/j.clineuro.2020.106126
[22]	Alkhaibary A, Alassiri AH, AlSufiani F, et al. (2019) Ki-67 labeling index in glioblastoma; does it really matter?. Hematology/Oncology Stem Cell Ther 12: 82-88. https://doi.org/10.1016/j.hemonc.2018.11.001
[23]	Smith HR (2015) Depression in cancer patients: Pathogenesis, implications and treatment (review). Oncol Lett 9: 1509-14. https://doi.org/10.3892/ol.2015.2944
[24]	Huh DH, Kim S (2020) History of Radiation Therapy Technology. Korean Soc Med Phys 31: 124-134. https://doi.org/10.14316/PMP.2020.31.3.124
[25]	Baskar R, Lee KA, Yeo R, et al. (2012) Cancer and radiation therapy: Current advances and future directions. Int J Med Sci 9: 193-199. https://doi.org/10.7150/ijms.3635
[26]	National Cancer InstituteExternal Beam Radiation Therapy (2020). https://doi.org/10.32388/h102yn
[27]	Chalmers AJ, Ruff EM, Martindale C, et al. (2009) Cytotoxic Effects of Temozolomide and Radiation are Additive- and Schedule-Dependent. Int J Radiat Oncol 75: 1511-1519. https://doi.org/10.1016/j.ijrobp.2009.07.1703
[28]	Ali MY, Oliva CR, Noman ASM, et al. (2020) Radioresistance in glioblastoma and the development of radiosensitizers. Cancers 12: 1-29. https://doi.org/10.3390/cancers12092511
[29]	Higby GJ (1982) Gold in medicine - A review of its use in the west before 1900. Gold Bull 15: 130-40. https://doi.org/10.1007/BF03214618
[30]	Fricker SP (1996) Medical uses of gold compounds: Past, present and future. Gold Bull 29: 53-60. https://doi.org/10.1007/bf03215464
[31]	Giljohann DA, Seferos DS, Daniel WL, et al. (2010) Gold nanoparticles for biology and medicine. Angew Chem Int Edit 49: 3280-94. https://doi.org/10.1002/anie.200904359
[32]	Hu X, Zhang Y, Ding T, et al. (2020) Multifunctional Gold Nanoparticles: A Novel Nanomaterial for Various Medical Applications and Biological Activities. Front Bioeng Biotech 8: 990. https://doi.org/10.3389/FBIOE.2020.00990/BIBTEX
[33]	Hu X, Zhang Y, Ding T, et al. (2020) Multifunctional Gold Nanoparticles: A Novel Nanomaterial for Various Medical Applications and Biological Activities. Front Bioeng Biotech 8: 990. https://doi.org/10.3389/fbioe.2020.00990
[34]	Frens G (1973) Controlled Nucleation for the Regulation of the Particle Size in Monodisperse Gold Suspensions. Nature Phys Sci 241: 20-2. https://doi.org/10.1038/physci241020a0
[35]	Brust M, Walker M, Bethell D, et al. (1994) Synthesis of thiol-derivatised gold nanoparticles in a two-phase Liquid–Liquid system. J Chem Soc Chem Commun, : 801-802. https://doi.org/10.1039/C39940000801
[36]	Yeh YC, Creran B, Rotello VM (2012) Gold Nanoparticles: Preparation, Properties, and Applications in Bionanotechnology. Nanoscale 4: 1871-1880. https://doi.org/10.1039/C1NR11188D
[37]	Kim M, Lee JH, Nam JM (2019) Plasmonic Photothermal Nanoparticles for Biomedical Applications. Adv Sci 6: 1900471. https://doi.org/10.1002/advs.201900471
[38]	Kim SE, Lee BR, Lee H, et al. (2017) Near-Infrared Plasmonic Assemblies of Gold Nanoparticles with Multimodal Function for Targeted Cancer Theragnosis. Sci Rep 7: 1-10. https://doi.org/10.1038/s41598-017-17714-2
[39]	Longmire M, Choyke PL, Kobayashi H (2008) Clearance properties of nano-sized particles and molecules as imaging agents: Considerations and caveats. Nanomedicine 3: 703-717. https://doi.org/10.2217/17435889.3.5.703
[40]	Gerosa C, Crisponi G, Nurchi VM, et al. (2020) Gold nanoparticles: A new golden Era in oncology?. Pharmaceuticals 13: 192. https://doi.org/10.3390/ph13080192
[41]	Gupta R, Xie H (2018) Nanoparticles in daily life: Applications, toxicity and regulations. J Environ Pathol Tox 37: 209-230. https://doi.org/10.1615/JEnvironPatholToxicolOncol.2018026009
[42]	Kong FY, Zhang JW, Li RF, et al. (2017) Unique roles of gold nanoparticles in drug delivery, targeting and imaging applications. Molecules 22: 1445. https://doi.org/10.3390/molecules22091445
[43]	Singh P, Pandit S, Mokkapati VRSS, et al. (2018) Gold nanoparticles in diagnostics and therapeutics for human cancer. Int J Mol Sci 19: 1979. https://doi.org/10.3390/ijms19071979
[44]	Chen YS, Hung YC, Hong MY, et al. (2012) Control of in vivo transport and toxicity of nanoparticles by tea melanin. J Nanomater 2012. https://doi.org/10.1155/2012/746960
[45]	Lasagna-Reeves C, Gonzalez-Romero D, Barria MA, et al. (2010) Bioaccumulation and toxicity of gold nanoparticles after repeated administration in mice. Biochem Bioph Res Co 393: 649-655. https://doi.org/10.1016/j.bbrc.2010.02.046
[46]	Sani A, Cao C, Cui D (2021) Toxicity of gold nanoparticles (AuNPs): A review. Biochem Biophys Rep 26: 100991. https://doi.org/10.1016/j.bbrep.2021.100991
[47]	Sun PP, Lai CS, Hung CJ, et al. (2021) Subchronic oral toxicity evaluation of gold nanoparticles in male and female mice. Heliyon 7: E06577. https://doi.org/10.1016/j.heliyon.2021.e06577
[48]	Berce C, Lucan C, Petrushev B, et al. (2016) In vivo assessment of bone marrow toxicity by gold nanoparticle-based bioconjugates in CrL:CD1(ICR) mice. Int J Nanomed 11: 4261-4273. https://doi.org/10.2147/IJN.S108711
[49]	Schaeublin NM, Braydich-Stolle LK, Schrand AM, et al. (2011) Surface charge of gold nanoparticles mediates mechanism of toxicity. Nanoscale 3: 410-420. https://doi.org/10.1039/c0nr00478b
[50]	Schneider T, Westermann M, Glei M (2017) In vitro uptake and toxicity studies of metal nanoparticles and metal oxide nanoparticles in human HT29 cells. Arch Toxicol 91: 3517-3527. https://doi.org/10.1007/s00204-017-1976-z
[51]	Bahamonde J, Brenseke B, Chan MY, et al. (2018) Gold Nanoparticle Toxicity in Mice and Rats: Species Differences. Toxicol Pathol 46. https://doi.org/10.1177/0192623318770608
[52]	Glazer ES, Zhu C, Hamir AN, et al. (2011) Biodistribution and acute toxicity of naked gold nanoparticles in a rabbit hepatic tumor model. Nanotoxicology 5: 459-468. https://doi.org/10.3109/17435390.2010.516026
[53]	Kim KT, Zaikova T, Hutchison JE, et al. (2013) Gold nanoparticles disrupt zebrafish eye development and pigmentation. Toxicol Sci 133: 275-288. https://doi.org/10.1093/toxsci/kft081
[54]	Pompa PP, Vecchio G, Galeone A, et al. (2011) In Vivo toxicity assessment of gold nanoparticles in Drosophila melanogaster. Nano Res 4. https://doi.org/10.1007/s12274-011-0095-z
[55]	Castillo MH, Button TM, Doerr R, et al. (1988) Effects of radiotherapy on mandibular reconstruction plates. Am J Surg 156: 261-263. https://doi.org/10.1016/S0002-9610(88)80287-3
[56]	Choi J, Kim G, Cho SB, et al. (2020) Radiosensitizing high-Z metal nanoparticles for enhanced radiotherapy of glioblastoma multiforme. J Nanobiotechnol 18: 122. https://doi.org/10.1186/S12951-020-00684-5
[57]	Cunningham C, de Kock M, Engelbrecht M, et al. (2021) Radiosensitization Effect of Gold Nanoparticles in Proton Therapy. Front Public Health 9. https://doi.org/10.3389/fpubh.2021.699822
[58]	Chen Y, Yang J, Fu S, et al. (2020) Gold nanoparticles as radiosensitizers in cancer radiotherapy. Int J Nanomed 2020: 9407-9430. https://doi.org/10.2147/IJN.S272902
[59]	Penninckx S, Heuskin AC, Michiels C, et al. (2020) Gold nanoparticles as a potent radiosensitizer: A transdisciplinary approach from physics to patient. Cancers 12: 2021. https://doi.org/10.3390/cancers12082021
[60]	Mousavie Anijdan SH, Mahdavi SR, Shirazi A, et al. (2013) Megavoltage X-ray Dose Enhancement with Gold Nanoparticles in Tumor Bearing Mice. Int J Mol Cell Med 2: 118-123.
[61]	Choi BJ, Jung KO, Graves EE, et al. (2018) A gold nanoparticle system for the enhancement of radiotherapy and simultaneous monitoring of reactive-oxygen-species formation. Nanotechnology 29. https://doi.org/10.1088/1361-6528/aae272
[62]	Ahmed AU, Tyler MA, Thaci B, et al. (2011) A comparative study of neural and mesenchymal stem cell-based carriers for oncolytic adenovirus in a model of malignant glioma. Mol Pharmaceut 8: 1559-1572. https://doi.org/10.1021/mp200161f
[63]	Schnarr K, Mooney R, Weng Y, et al. (2013) Gold Nanoparticle-Loaded Neural Stem Cells for Photothermal Ablation of Cancer. Adv Healthc Mater 2: 976-982. https://doi.org/10.1002/adhm.201300003
[64]	Danyu M, Huile G, Wei G, et al. (2011) Anti glioma effect of doxorubicin loaded liposomes modified with Angiopep-2. Afr J Pharm Pharmaco 5: 409-414. https://doi.org/10.5897/AJPP11.068
[65]	Zhang Y, Huang Y, Li S (2014) Polymeric micelles: Nanocarriers for cancer-targeted drug delivery. AAPS Pharm Sci Tech 15: 862-71. https://doi.org/10.1208/s12249-014-0113-z
[66]	Clavreul A, Pourbaghi-Masouleh M, Roger E, et al. (2017) Human mesenchymal stromal cells as cellular drug-delivery vectors for glioblastoma therapy: A good deal?. J Exp Clin Canc Res 36. https://doi.org/10.1186/s13046-017-0605-2
[67]	Upadhyay RK (2014) Drug delivery systems, CNS protection, and the blood brain barrier. BioMed Res Int 2014. https://doi.org/10.1155/2014/869269
[68]	Kaneko N, Kako E, Sawamoto K (2011) Prospects and limitations of using endogenous neural stem cells for brain regeneration. Genes 2: 107-30. https://doi.org/10.3390/genes2010107
[69]	Reynolds BA, Weiss S (1992) Generation of neurons and astrocytes from isolated cells of the adult mammalian central nervous system. Science 255: 1707-1710. https://doi.org/10.1126/SCIENCE.1553558
[70]	Nagato M, Heike T, Kato T, et al. (2005) Prospective characterization of neural stem cells by flow cytometry analysis using a combination of surface markers. J Neurosci Res 80: 456-66. https://doi.org/10.1002/jnr.20442
[71]	Yuan SH, Martin J, Elia J, et al. (2011) Cell-surface marker signatures for the Isolation of neural stem cells, glia and neurons derived from human pluripotent stem cells. PLoS ONE 6. https://doi.org/10.1371/journal.pone.0017540
[72]	Deshpande K, Saatian B, Martirosian V, et al. (2019) Isolation of Neural Stem Cells from Whole Brain Tissues of Adult Mice. Curr Protoc Stem Cell Biol 49. https://doi.org/10.1002/cpsc.80
[73]	Casarosa S, Bozzi Y, Conti L (2014) Neural stem cells: ready for therapeutic applications?. Mol Cell Ther 2: 31. https://doi.org/10.1186/2052-8426-2-31
[74]	Ottoboni L, von Wunster B, Martino G (2020) Therapeutic Plasticity of Neural Stem Cells. Front Neurol 11: 148. https://doi.org/10.3389/fneur.2020.00148
[75]	Haus DL, López-Velázquez L, Gold EM, et al. (2016) Transplantation of human neural stem cells restores cognition in an immunodeficient rodent model of traumatic brain injury. Exp Neurol 281: 1-16. https://doi.org/10.1016/j.expneurol.2016.04.008
[76]	Zhang S, Xie R, Zhao T, et al. (2014) Neural stem cells preferentially migrate to glioma stem cells and reduce their stemness phenotypes. Int J Oncol 45: 1989-96. https://doi.org/10.3892/ijo.2014.2629
[77]	Jeong YJ, Jeung HA, Seung UK, et al. (2008) Migration of human neural stem cells toward an intracranial glioma. Exp Mol Med 40: 84-91. https://doi.org/10.3858/emm.2008.40.1.84
[78]	Reagan MR, Kaplan DL (2011) Concise review: Mesenchymal stem cell tumor-homing: Detection methods in disease model systems. Stem Cells 29: 920-7. https://doi.org/10.1002/stem.645
[79]	Aboody KS, Brown A, Rainov NG, et al. (2000) Neural stem cells display extensive tropism for pathology in adult brain: Evidence from intracranial gliomas. P Natl Acad Sci USA 97: 12846-51. https://doi.org/10.1073/pnas.97.23.12846
[80]	Mooney R, Roma L, Zhao D, et al. (2014) Neural stem cell-mediated intratumoral delivery of gold nanorods improves photothermal therapy. ACS Nano 8: 12450-60. https://doi.org/10.1021/nn505147w
[81]	Baghban R, Roshangar L, Jahanban-Esfahlan R, et al. (2020) Tumor microenvironment complexity and therapeutic implications at a glance. Cell Commun Signal 18. https://doi.org/10.1186/s12964-020-0530-4
[82]	Albini A, Bruno A, Gallo C, et al. (2015) Cancer stem cells and the tumor microenvironment: Interplay in tumor heterogeneity. Connect Tissue Res 56. https://doi.org/10.3109/03008207.2015.1066780
[83]	Hill BS, Pelagalli A, Passaro N, et al. (2017) Tumor-Educated mesenchymal stem cells promote Pro-Metastatic phenotype. Oncotarget 8: 73296-73311. https://doi.org/10.18632/oncotarget.20265
[84]	Carey-Ewend AG, Hagler SB, Bomba HN, et al. (2021) Developing Bioinspired Three-Dimensional Models of Brain Cancer to Evaluate Tumor-Homing Neural Stem Cell Therapy. Tissue Eng Pt A 27: 857-866. https://doi.org/10.1089/ten.tea.2020.0113
[85]	Xu JT, Qian Y, Wang W, et al. (2020) Effect of stromal cell-derived factor-1/CXCR4 axis in neural stem cell transplantation for Parkinson's disease. Neural Regen Res 15: 112-9. https://doi.org/10.4103/1673-5374.264470
[86]	Heese O, Disko A, Zirkel D, et al. (2005) Neural stem cell migration toward gliomas in vitro. Neuro-Oncology 7: 476-84. https://doi.org/10.1215/S1152851704000754
[87]	Schmidt NO, Przylecki W, Yang W, et al. (2005) Brain tumor tropism of transplanted human neural stem cells is induced by vascular endothelial growth factor. Neoplasia 7: 623-30. https://doi.org/10.1593/neo.04781
[88]	Miska J, Lesniak MS (2015) Neural Stem Cell Carriers for the Treatment of Glioblastoma Multiforme. EBioMedicine 2: 774-5. https://doi.org/10.1016/j.ebiom.2015.08.022

This article has been cited by:

1.	Mina Emami Khansari, Saeed Sharifian, A deep reinforcement learning approach towards distributed Function as a Service (FaaS) based edge application orchestration in cloud-edge continuum, 2025, 233, 10848045, 104042, 10.1016/j.jnca.2024.104042
2.	Fatima Daqaq, Siham Ouhimmou, 2024, Genghis Khan Shark Optimizer Based Approach for Multi-Objective Engineering Problems, 979-8-3503-8735-3, 1, 10.1109/ICOA62581.2024.10754310
3.	Haytham Hesham Elmousalami, Nehal Elshaboury, Ahmed Hussien Ibrahim, Ahmed Hussien Elyamany, Bayesian Optimized Ensemble Learning System for Predicting Conceptual Cost and Construction Duration of Irrigation Improvement Systems, 2024, 12267988, 100014, 10.1016/j.kscej.2024.100014
4.	Jinxian Yang, Xiaojian Yang, Ying Zhang, Online Error Compensation of Accelerometer While Drilling Based on MISO, 2025, 25, 1530-437X, 9130, 10.1109/JSEN.2025.3529766

Reader Comments

Your name:*

Email:*
© 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

AIMS Neuroscience

3.1 4.2

Metrics

Article views(2935) PDF downloads(331) Cited by(3)

Preview PDF

Download XML

Export Citation

Article outline

Show full outline

Figures and Tables

Figures(1)

AIMS Neuroscience

Targeted delivery of gold nanoparticles by neural stem cells to glioblastoma for enhanced radiation therapy: a review

Related Papers:

Abstract

1. Introduction

2. Robot kinematics

2.1. Kinematic calculations

2.2. Robotic workspace

3. Trajectory planning

3.1. Cubic polynomial interpolation function

3.2. Quintic polynomial interpolation function

3.3. Piecewise polynomial interpolation

4. Genghis Khan shark optimizer (GKSO)

4.1. GKSO background

4.2. Mathematical model

4.2.1. Hunting stage

4.2.2. Moving stage

4.2.3. Foraging stage

4.2.4. Self-protection stage

4.2.5. Summary of the algorithm

5. Experiment

5.1. Objective function

5.2. Work task

5.3. Experimental simulation

5.4. Experiment

6. Conclusions

Use of AI tools declaration

Acknowledgments

Conflict of interest

References

This article has been cited by:

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Catalog

AIMS Neuroscience

Targeted delivery of gold nanoparticles by neural stem cells to glioblastoma for enhanced radiation therapy: a review

Related Papers:

Abstract

1. Introduction

2. Robot kinematics

2.1. Kinematic calculations

2.2. Robotic workspace

3. Trajectory planning

3.1. Cubic polynomial interpolation function

3.2. Quintic polynomial interpolation function

3.3. Piecewise polynomial interpolation

4. Genghis Khan shark optimizer (GKSO)

4.1. GKSO background

4.2. Mathematical model

4.2.1. Hunting stage

4.2.2. Moving stage

4.2.3. Foraging stage

4.2.4. Self-protection stage

4.2.5. Summary of the algorithm

5. Experiment

5.1. Objective function

5.2. Work task

5.3. Experimental simulation

5.4. Experiment

6. Conclusions

Use of AI tools declaration

Acknowledgments

Conflict of interest

References

This article has been cited by:

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Related pages

Tools

Export File

Citation

Format

Content

Catalog