Search

Comparative view of the cost function when using the GWO, GD, and PSO algorithms. [Display omitted] • A Policy Iteration Reinforcement Learning-based control approach is proposed. • A Grey Wolf Optimizer algorithm trains the policy neural network. • A comparison with classical Gradient Descent and metaheuristic PSO is offered. • The validation is done on real-time nonlinear servo system position control. • The fast convergence of the suggested approach is proved. This paper presents a new Reinforcement Learning (RL)-based control approach that uses the Policy Iteration (PI) and a metaheuristic Grey Wolf Optimizer (GWO) algorithm to train the Neural Networks (NNs). Due to an efficient tradeoff to exploration and exploitation, the GWO algorithm shows good results in NN training and solving complex optimization problems. The proposed approach is compared to the classical PI RL-based control approach using the Gradient Descent (GD) algorithm, and with the RL-based control approach which uses the metaheuristic Particle Swarm Optimization (PSO) algorithm. The experiments are conducted using a nonlinear servo system laboratory equipment. Each approach evaluated on how well it solves the optimal reference tracking control for an experimental servo system position control system. The policy NNs specific to all three approaches are implemented as state feedback with integrator controllers to remove the steady-state control errors and thus ensure the convergence of the objective function. Because of the random nature of metaheuristic algorithms, the experiments for GWO and PSO algorithms are run multiple times and the results are averaged before the conclusions are presented. The experimental results shows that for the control objective considered in this paper, the GWO algorithm represents a better solution compared to GD and PSO algorithms. [ABSTRACT FROM AUTHOR]

Velocity planning serves as an important issue in motion planning for autonomous vehicles. The presented paper proposes a novel velocity planning method with minimum moving time on the basis of path curvature which is accomplished in three steps. First, the assigned path is divided into some elementary parts based on the path curvature. Second, the velocity planning is transformed into an unconstrained optimization problem by assuming the velocity of vehicle to be a specific cubic polynomial on every elementary part to avoid a sudden acceleration in path switching. Finally, we use a modified projection particle swarm optimization (PPSO) algorithm to obtain the time-optimal velocity profile. The proposed method can generate a smooth time-optimal velocity profile while considering all possible relevant constraints. Three examples are provided on different types of path to demonstrate that the final velocity profile is efficient to avoid the sudden acceleration change. Furthermore, the modified PPSO algorithm in this paper is used to solve the optimization problem with high dimensional variables when its upper bound is known, which can not be achieved by the general PPSO algorithm. [ABSTRACT FROM AUTHOR]

• A combination of Deep Q-Learning algorithm and metaheuristic GSA is offered. • GSA initializes the weights and the biases of the neural networks. • A comparison with classical random, metaheuristic PSO and GWO is carried out. • The validation is done on real-time nonlinear servo system position control. • The drawbacks of randomly initialized neural networks are mitigated. This paper presents a novel Reinforcement Learning (RL)-based control approach that uses a combination of a Deep Q-Learning (DQL) algorithm and a metaheuristic Gravitational Search Algorithm (GSA). The GSA is employed to initialize the weights and the biases of the Neural Network (NN) involved in DQL in order to avoid the instability, which is the main drawback of the traditional randomly initialized NNs. The quality of a particular set of weights and biases is measured at each iteration of the GSA-based initialization using a fitness function aiming to achieve the predefined optimal control or learning objective. The data generated during the RL process is used in training a NN-based controller that will be able to autonomously achieve the optimal reference tracking control objective. The proposed approach is compared with other similar techniques which use different algorithms in the initialization step, namely the traditional random algorithm, the Grey Wolf Optimizer algorithm, and the Particle Swarm Optimization algorithm. The NN-based controllers based on each of these techniques are compared using performance indices specific to optimal control as settling time, rise time, peak time, overshoot, and minimum cost function value. Real-time experiments are conducted in order to validate and test the proposed new approach in the framework of the optimal reference tracking control of a nonlinear position servo system. The experimental results show the superiority of this approach versus the other three competing approaches. [ABSTRACT FROM AUTHOR]

In this paper, optimization of a linear objective function subject to a generalized fuzzy relational inequalities is investigated in which an arbitrary continuous t-norm is considered as fuzzy composition, and fuzzy inequality replaces ordinary inequality in the constraints. Unlike most optimization algorithms, in fuzzy relational inequalities with fuzzy constraints (FRI-FC) we find a near-feasible solution having a better objective value than those resulting from the resolution of the similar problems with crisp (ordinary) inequality constraints. Such solutions are called super-optima in this paper. Subsequently, an algorithm is proposed to find a super-optimum with pre-specified desirable infeasibility. For this purpose, we firstly study some structural properties of the FRI-FC problem and present a new formulization that is independent of t-norms used in the constraints of the problem. This new formulation converts the primary problem into an equivalent problem with simple constraints without considering any penalty parameters. However, it is proved that the transformed equivalent problem enables our algorithm to distinguish unfavorable points and generate a sequence of solutions converging to a super-optimum under a certain sufficient condition. Finally, a modified PSO is presented in which the ability of PSO to solve unconstrained problems with continuous domain, the structure of the transformed problem and some fuzzy structural modifications are combined and form an efficient algorithm to solve the generalized FRI-FC problems. The modified PSO algorithm has been applied to the generalized FRI-FC problem defined with ten well-known continuous t-norms. Additionally, an idea of the FRI-FC problems as outer approximators for FRI problems with ordinary inequalities is also investigated. [ABSTRACT FROM AUTHOR]

Many-objective problems refer to the optimization problems containing more than three conflicting objectives. To obtain a representative set of well-distributed non-dominated solutions close to Pareto front in the objective space remains a challenging problem. Many papers have proposed different Multi-Objective Evolutionary Algorithms to solve the lack of the convergence and diversity in many-objective problems. One of the more promising approaches uses a set of reference points to discriminate the solutions and guide the search process. However, this approach was incorporated mainly in Multi-Objective Evolutionary Algorithms, and there are just some few promising adaptations of Particle Swarm Optimization approaches for effectively tackling many-objective problems regarding convergence and diversity. Thus, this paper proposes a practical and efficient Many-Objective Particle Swarm Optimization algorithm for solving many-objective problems. Our proposal uses a set of reference points dynamically determined according to the search process, allowing the algorithm to converge to the Pareto front, but maintaining the diversity of the Pareto front. Our experimental results demonstrate superior or similar performance when compared to other state-of-art algorithms. [ABSTRACT FROM AUTHOR]

Particle swarm optimization (PSO) is a population-based random-search optimization technique that has become an increasingly important branch of swarm intelligence studies. Population diversity is an effective measurement that shows the distribution of particles in a search space. In this paper, we propose a diversity-based PSO algorithm that is combined with multiple techniques, and population diversity is used as a search-strategy selection criterion for particles. When diversity is high, we suggest implementing a search strategy that has a strong exploration ability. When diversity is low, we suggest implementing a search strategy that has a strong exploitation ability. In addition to our diversity-based method, we introduce a gradient-based local-search technique, multi-crossover operation, and disturbance strategy to help improve the performance of the proposed algorithm. In the experiments, we compare the proposed algorithm with 10 advanced PSO variants based on 40 widely used benchmark functions, including the CEC2017 benchmark. The results indicate that the proposed algorithm yields a better solution accuracy and convergence speed than those of other PSO variants. [ABSTRACT FROM AUTHOR]

• The concept and verification method of affine invariance are presented. • Six algorithms are employed to analyze their affine invariances. • PSO, OFA and DE are affine invariant, GWO, SCA and BOA are not affine invariant. • Multiple experimental comparisons are executed to verify algorithms' affine invariance. An algorithm whose performance depends on the objective function being aligned with a privileged coordinate system is a poor choice in general because it is unlikely that the optimal orientation will be known in advance. In this paper, a property of meta -heuristic algorithms, named affine invariance, is introduced to verify whether the algorithm is depended on the privileged coordinate system or not. The concept of affine invariance is described in detail, and some classical algorithms, efficient in most test and actual problems, are proved to be affine invariant. While some recent algorithms in the literature are proved to be not affine invariant. As a conclusion, particle swarm optimization (PSO), differential evolution (DE) and optimal foraging algorithm (OFA) are affine invariant, while grey wolf optimizer (GWO), sine cosine algorithm (SCA) and butterfly optimization algorithm (BOA) are not affine invariant. Furthermore, comparison tests are designed to support the theoretical analysis results. In these tests, same random numbers and initial population are used to avoid the influence of randomness, thus, the conclusion is reliable. [ABSTRACT FROM AUTHOR]

Real-time services for IoT enabled Disastrous Management Scenarios (DMSs), multicast routing problem has attracted many researchers due to its NP-hard behaviour. The objective is to find a Steiner Tree (S-tree) satisfying the essential constraints. To solve complex problems with minimal overhead and better convergence, Particle Swarm Optimization (PSO) has been widely applied to solve such optimization problems. Data dissemination in DMS poses QoS aware multicast routing as a multi-objective search optimization problem. Conventional Multi-Objective PSO (MOPSO) algorithm has been used to solve multicast routing problem considering multi-objectives i.e., power consumption, delay, and packet loss. However, it consumes more computation power and energy due to iterative searching in complete multidimensional search space and also because of a generic method which converts invalid to valid solutions (S-trees). In this paper, a Forwarding Zone (FZ) enabled MOPSO (FZMOPSO) is proposed which acts in FZ to search the valid solution irrespective of complete search space. An FZ enabled Valid Particle Generator (FZVPG) operator is proposed and integrated with FZMOPSO to convert invalid to valid particles effectively. The performance of FZMOPSO is evaluated considering twelve different network scenarios and the results are compared with several state-of-the-art algorithms. Experimental results evidenced the effectiveness of FZMOPSO. [ABSTRACT FROM AUTHOR]

Abstract As a new evolutionary computation method, the structure of backtracking search optimization algorithm (BSA) is simple and the exploration capability of it is strong. However, the global performance of the BSA is significantly affected by mutation strategies and control parameters. Designing appropriate mutation strategies and control parameters is important to improve the global performance of the BSA. In this paper, an adaptive BSA with knowledge learning (KLBSA) is developed to improve the global performance of the BSA. In the method, an adaptive control parameter based on the global and local information of the swarms in the current iteration is designed to adjust the search step length of individuals, which helps to balance the exploration and exploitation abilities of the algorithm. Moreover, a new mutation strategy based on the guidance of different information is designed to improve the optimization ability of the algorithm. In addition, a multi-population strategy is implemented to thoroughly improve the searching ability of the algorithm for different searching areas. To this end, experiments on three groups of benchmark functions and three real-world problems are implemented to verify the performance of the proposed KLBSA algorithm. The results indicate that the proposed algorithm performs competitively and effectively when compared to some other evolutionary algorithms. [ABSTRACT FROM AUTHOR]

Sport is one of the activities of human being where cooperative, competitive, self-learning and interactive environment helps in the overall improvement of the performance. These learning processes are very effective to regulate the players in a good direction as well as to enhance their capability of exploring the new techniques. Particle Swarm Optimization (PSO) is a popular stochastic optimization algorithm, used for solving real-world engineering problems. However, it usually suffers from local confinement and easily loses its diversity. In this paper, we have integrated the properties of sports with PSO algorithm and proposed an efficient player selection strategy based diversified PSO (EPS-dPSO), which improves the fitness and robustness of the technique without compromising the computational complexity of the algorithm. The properties of player selection is adopted to enhance the diversity within the search phase as well as to incorporate intense searching of the space. We have comprehensively evaluated the performance of proposed EPS-dPSO by applying it on standard benchmark problems. Experimental result shows that it not only tracks the global optimum within the small search interval but also able to obtain good result for large and asymmetrical search space and also insensitive to initialization of the problems. Further, tests are carried out on the benchmark functions from CEC2005, the large dimensional problems of CEC2008 and some real world problems from CEC2011. All the experimental results indicate the effectiveness and efficiency of the proposed EPS-dPSO compared to other traditional algorithms. [ABSTRACT FROM AUTHOR]

Structure learning is a very important problem in the field of Bayesian networks (BNs). It is also an active research area for more than 2 decades; therefore, many approaches have been proposed in order to find an optimal structure based on training samples. In this paper, a Particle Swarm Optimization (PSO)-based algorithm is proposed to solve the BN structure learning problem; named BNC-PSO ( B ayesian N etwork C onstruction algorithm using PSO ). Edge inserting/deleting is employed in the algorithm to make the particles have the ability to achieve the optimal solution, while a cycle removing procedure is used to prevent the generation of invalid solutions. Then, the theorem of Markov chain is used to prove the global convergence of our proposed algorithm. Finally, some experiments are designed to evaluate the performance of the proposed PSO-based algorithm. Experimental results indicate that BNC-PSO is worthy of being studied in the field of BNs construction. Meanwhile, it can significantly increase nearly 15% in the scoring metric values, comparing with other optimization-based algorithms. [ABSTRACT FROM AUTHOR]

High-order graph matching aims at establishing correspondences between two sets of feature points using high-order constraints. It is usually formulated as an NP-hard problem of maximizing an objective function. This paper introduces a discrete particle swarm optimization algorithm for resolving high-order graph matching problems, which incorporates several re-defined operations, a problem-specific initialization method based on heuristic information, and a problem-specific local search procedure. The proposed algorithm is evaluated on both synthetic and real-world datasets. Its outstanding performance is validated in comparison with three state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

In this paper, a dynamic mentoring scheme along with a self-regulation scheme have been incorporated in the standard Particle Swarm Optimization (PSO) algorithm to empower the searching particles with human-like characteristics. The algorithm is referred to as a Dynamic Mentoring and Self-Regulation based Particle Swarm Optimization (DMeSR-PSO) algorithm. Based on their experiences, the particles are divided into three groups, viz., the mentor group, the mentee group and the independent learner group where the number of particles in each group is dynamically changing in every iteration. In human learning psychology, mentoring is regarded as a powerful and effective learning process and independent learners are the ones who do not need mentoring and are capable of performing self-regulation of their own knowledge. Therefore, the particles in each of the above three groups have different learning strategies for their velocity updates where the mentors are equipped with a strong self-belief based search, the mentees are taking guidance from the mentors and the independent learners employ self-perception strategy. The DMeSR-PSO algorithm has been extensively evaluated using the simple unimodal and multimodal benchmark functions from CEC2005, more complex shifted and rotated benchmark functions fromCEC2013 and also based on eight real- world problems from CEC2011. The results have been compared with six state-of-the-art PSO variants and five meta-heuristic algorithms for the CEC2005 problems. Further, a comparative analysis on CEC2013 benchmark functions with different PSO variants has also been presented. Finally, DMeSR-PSO's performance on the real-world problems is compared with the top two algorithms from the CEC2011 competition. The results indicate that the proposed learning strategies help DMeSR-PSO to achieve faster convergence and provide better solutions in most of the problems with a 95% confidence level, yielding an effective optimization algorithm for real-world applications. [ABSTRACT FROM AUTHOR]

The diversity of solutions is of great importance for multi-objective evolutionary algorithms. In this paper, a new multi-objective particle swarm optimization algorithm based on decomposition (MPSO/D) is proposed. Firstly, the objective space of a multi-objective problem is decomposed into a set of sub-regions based on a set of direction vectors. Then MPSO/D makes each sub-region have a solution to maintain the diversity. Secondly, considering the convergence of solutions, MPSO/D uses the crowding distance to calculate the fitness values of the reserved solutions for selection operator, and uses the neighboring particles of a particle to determine the global best historical position ( gbest ) of the particle. The proposed algorithm has been compared with NSGAII, MOEA/D and NNIA on sixteen test sets. The experimental results illustrate that the proposed algorithm outperforms NSGAII, MOEA/D and NNIA in terms of convergence and diversity. [ABSTRACT FROM AUTHOR]

The multi-population method has been widely used to solve unconstrained continuous dynamic optimization problems with the aim of maintaining multiple populations on different peaks to locate and track multiple changing peaks simultaneously. However, to make this approach efficient, several crucial challenging issues need to be addressed, e.g., how to determine the moment to react to changes, how to adapt the number of populations to changing environments, and how to determine the search area of each population. In addition, several other issues, e.g., communication between populations, overlapping search, the way to create populations, detection of changes, and local search operators, should be also addressed. The lack of attention on these challenging issues within multi-population methods hinders the development of multi-population based algorithms in dynamic environments. In this paper, these challenging issues are comprehensively analyzed by a set of experimental studies from the algorithm design point of view. Experimental studies based on a set of popular algorithms show that the performance of algorithms is significantly affected by these challenging issues on the moving peaks benchmark. [ABSTRACT FROM AUTHOR]

Social learning plays an important role in behavior learning among social animals. In contrast to individual (asocial) learning, social learning has the advantage of allowing individuals to learn behaviors from others without incurring the costs of individual trials-and-errors. This paper introduces social learning mechanisms into particle swarm optimization (PSO) to develop a social learning PSO (SL-PSO). Unlike classical PSO variants where the particles are updated based on historical information, including the best solution found by the whole swarm (global best) and the best solution found by each particle (personal best), each particle in the proposed SL-PSO learns from any better particles (termed demonstrators) in the current swarm. In addition, to ease the burden of parameter settings, the proposed SL-PSO adopts a dimension-dependent parameter control method. The proposed SL-PSO is first compared with five representative PSO variants on 40 low-dimensional test functions, including shifted and rotated test functions. The scalability of the proposed SL-PSO is further tested by comparing it with five state-of-the-art algorithms for large-scale optimization on seven high-dimensional (100-D, 500-D, and 1000-D) benchmark functions. Our comparative results show that SL-PSO performs well on low-dimensional problems and is promising for solving large-scale problems as well. [ABSTRACT FROM AUTHOR]