Yanmin Liu

Hybrid learning particle swarm optimizer with genetic disturbance

Abstract Particle swarm optimizer (PSO) is a population-based stochastic optimization technique which has already been successfully applied to the engineering and other scientific fields. This paper presents a modification of PSO (hybrid learning PSO with genetic disturbance, HLPSO-GD for short) intended to combat the problem of premature convergence observed in many PSO variants. In HLPSO-GD, the swarm uses a hybrid learning strategy whereby all other particles’ previous best information is adopted to update a particle׳s position. Additionally, to better make use of the excellent particle׳s information, the global external archive is introduced to store the best performing particle in the whole swarm. Furthermore, the genetic disturbance (simulated binary crossover and polynomial mutation) is used to cross the corresponding particle in the external archive, and generate new individuals which will improve the swarm ability to escape from the local optima. Experiments were conducted on a set of traditional multimodal test functions and CEC 2013 benchmark functions. The results demonstrate the good performance of HLPSO-GD in solving multimodal problems when compared with the other PSO variants.

A Multi-objective Particle Swarm Optimization Based on Decomposition

Decomposition is a classic method in traditional multi-objective optimization problems (MOPs). However, so far it has not yet been widely used in multi-objective particle swarm optimization (MOPSO). This paper proposes a MOPSO based on decomposition strategy (MOPSO-D), in which MOPs is decomposed into a number of scalar optimization sub-problems by a set of even spread weight vectors, and each sub-problem is optimized by a particle (here, it is viewed as a sub-swarm) personal history best position (pbest) and global best position in the its all neighbors (gbest) in a single run. By computing the Euclidean distances between any two weight vectors corresponding to a particle, the neighborhood identification strategy of each particle is assigned. The method of decomposition inherited the traditional method merits and makes MOPSO-D have lower computational complexity at each generation than NSMOPSO and OMOPSO. Simulation experiments on multi-objective 0-1 knapsack problems and continuous multi-objective optimization problems show MOPSO-D outperforms or performs similarly to NSMOPSO and OMOPSO.

Location Selection of Multiple Logistics Distribution Center Based on Particle Swarm Optimization

Distribution center is an important pivot position at the logistics system. This paper presents the site selection’s model of logistics distribution center based on decision matrix C, which adopts particle swarm optimization (PSO) to find the best combination of site selection by structuring the iteration of decision matrix. At the same time, simulation experiments are conducting for the site selection of logistics distribution center with 4 candidate centers and 10 distribution points, and the results show that PSO can get the best solution of distribution center in 90 % success rate of the best solution and the average research time of the approximate 3.5 s. From simulation experiment, PSO is efficient, accurate and suitable for the model optimization of distribution center, and therefore, it can be regarded as an effective method for the site selection’s model of logistics distribution.

An Improved PSO for Multimodal Complex Problem

As the multimodal complex problem has many local optima, basic PSO is difficult to effectively solve this kind of problem. To conquer this defect, firstly, we adopt Monte Carlo method to simulate the fly trajectory of particle, and conclude the reason for falling into local optima. Then, by defining distance, average distance and maximal distance between particles, an adaptive control factor (Adaptive rejection factor, ARF) for pp and pg was proposed to increase the ability for escaping from local optima. In order to test the proposed strategy, three test benchmarks were selected to conduct the analysis of convergence property and statistical property. The simulation results show that particle swarm optimizer based on adaptive rejection factor (ARFPSO) can effectively avoid premature phenomenon. Therefore, ARFPSO is available for complex multimodal problems.

Guided chemotaxis-based bacterial colony algorithm for three-echelon supply chain optimisation

On the base of the existing research study, a multi-period, multi-product, multi-supplier, single-manufacture, and multi-distributor supply chain model is considered in the paper. In the three-echelon model, a variety of decision-making activities involved in the procurement, production and distribution process are integrated at the operational level, giving rise to the non-deterministic polynomial-time hard computational complexity for model optimisation. For tackling the difficult model, this paper proposes a new optimisation method called guided chemotaxis-based bacterial colony algorithm, characterised by centre learning communication mechanism. More specifically, centre learning communication mechanism, where all the bacteria are enforced to learn towards the centre position of the swarm, is designed for the global exploration ability of algorithm. Chemotaxis, which guides the bacterium to fine-tune the solution in an increasingly favourable fitness landscape, is used to enhance the local exploitation ability of algorithm. Numerical experiments on a variety of simulated scenarios show the effectiveness and efficiency of the proposed algorithm in terms of both quality solution and computational time, by comparing it with some existing state-of-the-art solution approaches.

Effects of Simulated Annealing Strategy on Swarm Intelligence Algorithm

Swarm intelligence algorithm (SI) is a kind of stochastic search algorithm based on swarm. Similar to other evolutionary algorithm, when solving the complicated multimodal problem using SI, it is easy to have premature convergence. So, to promote the optimization of swarm intelligence algorithm, the typical algorithm (Particle swarm optimizer) of swarm intelligence algorithm is selected to explore some strategies how to improve the performance. In this paper, we explore the follow research: firstly, the mutation operation is introduced to produce new learn example for each individual in itself evolution process; secondly, in the view of the idea of simulated annealing, the range strategy of fitness of each individual is proposed; finally, to make best use of each individual information, the comprehensive learning strategy is adopted to improve each individual evolution mechanism.

Particle Swarm Optimizer with Full Information

In order to improve the particle swarm optimizer (PSO) for solving complex multimodal problems, an improved PSO with full information and mutation operator (PSOFIM) is proposed base basic PSO and mutation thought. In PSOFIM, a novel mutation is adopted to improve the history optimal position of particle (pbest) by disturbance in operation of each dimension. Additionally, a full information strategy for each particle is introduced to make best use of each dimension of each particle to ensure the information utility for swarm topology where each particle learns from its neighborhood information for his optimal position to improve itself study ability, whose strategies improve the swarm fly to the probability of the optimal solution. The simulation experiment results of benchmark function tests show PSOFIM has better performance than the basic PSO algorithm.

Application of Disturbance of DNA Fragments in Swarm Intelligence Algorithm

Optimization problem is one of the most important problems encountered in the real world. In order to effectively deal with optimization problem, some intelligence algorithms have been put forward, for example, PSO, GA, etc. To effectively solve this kind of problem, in this paper, crossing strategy of DNA fragments is proposed to explore the effect on intelligence algorithms based on American genetic biologist Morgan theory. We mainly focus on DNA fragment decreasing strategy and DNA fragment increasing strategy based on disturbance in PSO. In order to test the role of the DNA mechanism, three test benchmarks were selected to conduct the analysis of convergence property and statistical property. The simulation results show that the PSO with DNA mechanism have an advantage on algorithm performance efficiency compared with the original proposed PSO. Therefore, DNA mechanism is an effective method for improving swarm Intelligence algorithm performance.

Multi-objective PSO Based on Grid Strategy

In multi-objective optimization problem (MOP), keeping solution diversity is key case for solution quality. To improve the MOP quality, the diversity maintenance threshold value (λα) is proposed to keep solutions diversity based on adaptive grid strategy. These strategies can adaptive maintain the non-inferior diversity to improve swarm individual fly to the global optimal. Four test problems are selected to test the proposed strategy compared with other classical methods, and three performance metrics are chosen to explore the algorithm effectiveness.

Simulation Study of Swarm Intelligence Based on Life Evolution Behavior

Swarm intelligence (SI) is a new evolutionary computation technology, and its performance efficacy is usually affected by each individual behavior in the swarm. According to the genetic and sociological theory, the life evolution behavior process is influenced by the external and internal factors, so the mechanisms of external and internal environment change must be analyzed and explored. Therefore, in this paper, we used the thought of the famous American genetic biologist Morgan, “life = DNA