Catálogo de publicaciones - libros

Particle Swarm Optimization (PSO) is a population-based evolutionary search technique, which has comparable performance with Genetic algorithm. The existing PSOs, however, are not global-convergence-guaranteed algorithms, because the evolution equation of PSO, make the particle only search in a finite sampling space. In [10,11], a Quantum-behaved Particle Swarm Optimization algorithm is proposed that outperforms traditional PSOs in search ability as well as having less parameter. This paper focuses on discussing how to select parameter when QPSO is practically applied. After the QPSO algorithm is described, the experiment results of stochastic simulation are given to show how the selection of the parameter value influences the convergence of the particle in QPSO. Finally, two parameter control methods are presented and experiment results on the benchmark functions testify their efficiency.

A new hybrid Particle Swarm Optimization (PSO) algorithm is proposed based on the Nonlinear Simplex Search (NSS) method. At late stage of PSO, when the most promising regions of solutions are fixed, the algorithm isolates particles that are very close to the extrema, and applies the NSS method to them to enhance local exploitation searching. Explicit experimental results on famous benchmark functions indicate that this approach is reliable and efficient, especially on multi-model function optimizations. It yields better solution qualities and success rates compared to other published methods.

Recently, works about Non-Markov environment has attracted increasing attention in autonomous agent control. Based on XCS, XCSM introduced a constant length of bit-register memory into general classifier system structure to record agent’s experience. Then combining the current perception with its past experience, the agent gets suitable action successfully. But when the memory becomes longer, its performance will decrease urgently with the expansion of search space. In this paper, an adaptive XCSM (AXCSM) method has been proposed by adapting variable length of memory. The proposal composes of a hierarchical structure. In the first hierarchy, we learn a suitable memory length for this position using general XCS. Then action of the agent is acquired in the second hierarchy using XCSM. The proposal converges to optimal policy as that of XCSM, within shortened search space.

In this paper a discrete particle swarm optimization (DPSO) algorithm is proposed to solve permutation flowshop scheduling problems with the objective of minimizing the makespan. A discussion on implementation details of DPSO algorithm is presented. The proposed algorithm has been applied to a set of benchmark problems and performance of the algorithm is evaluated by comparing the obtained results with the results published in the literature. Further, it is found that the proposed improvement heuristic algorithm performs better when local search is performed. The results are presented.

The paper proposes a modified particle swarm optimizer for tracking dynamic systems. In the new algorithm, the changed local optimum and global optimum are introduced to guide the movement of each particle and avoid making direction and velocity decisions on the basis of the outdated information. An environment influence factor is put forward based on the two optimums above, which dynamically decide the change of the inertia weight. The combinations of the different local optimum update strategy and local inertia weight update strategy are tested on the parabolic benchmark function. The results on the benchmark function with various severities suggest that modified particle swarm optimizer performs better in convergence speed and aggregation accuracy.

The goal of bipartite subgraph problem is to partition the vertex set of an undirected graph into two parts in order to maximize the cardinality of the set of edges cut by the partition. This paper proposes a modified particle swarm optimization (PSO), called MPPSO (Mutated Personalized PSO), for this NP-hard problem. The proposed MPPSO algorithm contains a key improvement by introducing a personality factor from a psychological standpoint and a mutation operator for global best. Additionally the symmetry issue of solution space of bipartite subgraph problem is coped well with too. A large number of instances have been simulated to verify the proposed algorithm. The results show that the personality factor and mutation operator are efficient and the quality of our algorithm is superior to those of the existing algorithms.

Holonic architectures are more suitable for reconfigurable manufacturing systems compared with hierarchical and heterarchical architectures. A holonic architecture is proposed for reconfigurable manufacturing systems based on the well-known reference architecture PROSA. Considering the special status of the configuration in reconfigurable manufacturing systems, configuration holon is introduced besides the basic holons in PROSA. The basic structure of this holonic architecture, the details of basic holons and cooperation of holons are described in detail. Finally an agent-based holon model is introduced for the realization of the proposed holonic architecture.

Path planning for mobile robots is an important topic in modern robotics studies. This paper proposes a new approach to collision-free path planning problem for mobile robots using the particle swarm optimization combined with chaos iterations. The particle swarm optimization algorithm is run to get the global best particle as the candidate solution, and then local chaotic search iterations are employed to improve the solution precision. The effectiveness of the approach is demonstrated by three simulation examples.

One of the main problems of computational approaches to protein structure prediction is the computational complexity. Many researches use simplified models to represent protein structure. Toy model is one of the simplification models. Finding the ground state is critical to the toy model of protein. This paper applies Particle Swarm Optimization (PSO) Algorithm to search the ground state of toy model for protein folding, and performs experiments both on artificial data and real protein data to evaluate the PSO-based method. The results show that on one hand, the PSO method is feasible and effective to search for ground state of toy model; on the other hand, toy model just can simulate real protein to some extent, and need further improvements.

Inspired by the ongoing research on artificial visual prosthesis, a novel pixelization visual model based on the selection of local attention-drawing features is proposed, and a subjective scoring experiment as a cognitive assessment is designed to evaluate the performance of the model. The results of the experiment reveal that the model can accentuate the areas with prominent features in the original image, so as to give observers a subjective perception of rich visual information. Thus, the model will provide a new approach for future research.