• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.2
• /
• pp.407-415
• /
• 2011

In this study, we propose the design of optimized fuzzy controller for the rotary inverted pendulum system by using differential evolution algorithm. The structure of the differential evolution algorithm has a simple structure and its convergence to optimal values is superb in comparison to other optimization algorithms. Also the differential evolution algorithm is easier to use because it have simpler mathematical operators and have much less computational time when compared with other optimization algorithms. The rotary inverted pendulum system is nonlinear and has a unstable motion. The objective is to control the position of the rotating arm and to make the pendulum to maintain the unstable equilibrium point at vertical position. The output performance of the proposed fuzzy controller is considered from the viewpoint of performance criteria such as overshoot, steady-state error, and settling time through simulation and practical experiment. From the result of both simulation and practical experiment, we evaluate and analyze the performance of the proposed optimal fuzzy controller from the comparison between PGAs and differential evolution algorithms. Also we show the superiority of the output performance as well as the characteristic of differential evolution algorithm.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.1 no.1
• /
• pp.125-130
• /
• 2001

In this paper, we propose a new type of evolution strategies combined with reinforcement learning. We use the variances of fitness occurred by mutation to make the reinforcement signals which estimate and control the step length of mutation. With this proposed method, the convergence rate is improved. Also, we use cauchy distributed mutation to increase global convergence faculty. Cauchy distributed mutation is more likely to escape from a local minimum or move away from a plateau. After an outline of the history of evolution strategies, it is explained how evolution strategies can be combined with the reinforcement learning, named reinforcement evolution strategies. The performance of proposed method will be estimated by comparison with conventional evolution strategies on several test problems.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.1C
• /
• pp.20-25
• /
• 2006

In this paper, the performance of Tree-LDPC code is presented based on the min-sum algorithm with scaling and the asymptotic performance in the water fall region is shown by density evolution. We presents that the Tree-LDPC code show a significant performance gain by scaling with the optimal scaling factor which is obtained by density evolution methods. We also show that the performance of min-sum with scaling is as good as the performance of sum-product while the decoding complexity of min-sum algorithm is much lower than that of sum-product algorithm. The Tree-LDPC decoder is implemented on a FPGA chip with a small interleaver size.

• Journal of Computing Science and Engineering
• /
• v.9 no.2
• /
• pp.73-82
• /
• 2015

Understanding how the topic evolves is an important and challenging task. A topic usually consists of multiple related events, and the accurate identification of event evolution relationship plays an important role in topic evolution analysis. Existing research has used the traditional vector space model to represent the event, which cannot be used to accurately compute the semantic similarity between events. This has led to poor performance in identifying event evolution relationship. This paper suggests constructing a semantic aspect-based vector space model to represent the event: First, use hierarchical Dirichlet process to mine the semantic aspects. Then, construct a semantic aspect-based vector space model according to these aspects. Finally, represent each event as a point and measure the semantic relatedness between events in the space. According to our evaluation experiments, the performance of our proposed technique is promising and significantly outperforms the baseline methods.

• Earthquakes and Structures
• /
• v.26 no.3
• /
• pp.175-189
• /
• 2024

Aiming at studying the plastic hinge (PH) evolution regularities and failure mode of rocking wall moment frame (RWMF) structure in earthquakes, the whole-working history analysis of seismic performance state of RWMF structure based on co-operation performance and PH evolution was carried out. Building upon the theoretical analysis of the elastic internal forces and deformations of RWMF structures, nonlinear finite element analysis (FEA) methods were employed to perform both Pushover analysis and seismic response time history analysis under different seismic coefficients (δ). The relationships among PH occurrence ratios (R_ph), inter-story drifts and δ were established. Based on the plotted curve of the seismic performance states, evaluation limits for the R_ph and inter-story drifts were provided for different performance states of RWMF structures. The results indicate that the R_ph of RWMF structures exhibits a nonlinear evolution trend of "fast at first, then slow" with the increasing of δ. The general pattern is characterized by the initial development of beam hinges in the middle stories, followed by the development towards the top and bottom stories until the beam hinges are fully formed. Subsequently, the development of column hinges shifts from the bottom and top stories towards the middle stories of the structure, ultimately leading to the loss of seismic lateral capacity with a failure mode of partial beam yield, demonstrating a global yielding pattern. Moreover, the limits for the R_ph and inter-story drifts effectively evaluate the five different performance states of RWMF structures.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.420-426
• /
• 1998

Evolutionary algorithms are probabilistic optimization algorithms based on the model of natural evolution. Recently the efforts to improve the performance of evolutionary algorithms have been made extensively. In this paper, we introduce the research for improving the convergence rate and search faculty of evolution algorithms by using reinforcement learning. After providing an introduction to evolution algorithms and reinforcement learning, we present adaptive genetic algorithms, reinforcement genetic programming, and reinforcement evolution strategies which are combined with reinforcement learning. Adaptive genetic algorithms generate mutation probabilities of each locus by interacting with the environment according to reinforcement learning. Reinforcement genetic programming executes crossover and mutation operations based on reinforcement and inhibition mechanism of reinforcement learning. Reinforcement evolution strategies use the variances of fitness occurred by mutation to make the reinforcement signals which estimate and control the step length.

• Advances in Computational Design
• /
• v.1 no.4
• /
• pp.315-327
• /
• 2016

Damage detection and localisation in structures is essential since it can be a means for preventive maintenance of those structures under service conditions. The use of structural modal data for detecting the damage is one of the most efficient methods. This paper presents comparative performance of various state-of-the-art meta-heuristics for use in structural damage detection based on changes in modal data. The metaheuristics include differential evolution (DE), artificial bee colony algorithm (ABC), real-code ant colony optimisation (ACOR), charged system search (ChSS), league championship algorithm (LCA), simulated annealing (SA), particle swarm optimisation (PSO), evolution strategies (ES), teaching-learning-based optimisation (TLBO), adaptive differential evolution (JADE), evolution strategy with covariance matrix adaptation (CMAES), success-history based adaptive differential evolution (SHADE) and SHADE with linear population size reduction (L-SHADE). Three truss structures are used to pose several test problems for structural damage detection. The meta-heuristics are then used to solve the test problems treated as optimisation problems. Comparative performance is carried out where the statistically best algorithms are identified.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.04a
• /
• pp.233-236
• /
• 2007

A new evolution method for strong exploration and strong exploitation termed queen-bee and mutant-bee evolution is proposed based on the previous queen-bee evolution [1]. Even though the queen-bee evolution has shown very good performances, two parameters for strong mutation are added to the genetic algorithms. This makes the application of genetic algorithms with queen-bee evolution difficult because the values of the two parameters are empirically decided by a trial-and-error method without a systematic method. The queen-bee and mutant-bee evolution has no this problem because it does not need additional parameters for strong mutation. Experimental results with typical problems showed that the queen-bee and mutant-bee evolution produced nearly similar results to the best ones of queen-bee evolution even though it didn't need to select proper values of additional parameters.

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.1585-1588
• /
• 1997

In this paper, divide total workspace of robot manipulator into several subspaces and construct PID controller ineach subspace. Using EvolutionSTrategy we optimize the gains of PID controller in each subspace. But the gains may have a large difference on the boundary of subspaces, which can cause bad oscillatory performance. So we use Aritificial Neural Network to have continuous gain curves htrough the entire subspaces. Simualtion results show that the proposed method is quite useful.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.32 no.2
• /
• pp.109-121
• /
• 2007

This paper focuses on algorithms based on the evolution, which is applied to various optimization problems. Especially, among these algorithms based on the evolution, we investigate the simple genetic algorithm based on Darwin's evolution, the Lamarckian algorithm based on Lamark's evolution and the Baldwin algorithm based on the Baldwin effect and also Investigate the difference among them in the biological and engineering aspects. Finally, through this comparison, we suggest a new algorithm to find more various solutions changing the genotype or phenotype search space and show the performance of the proposed method. Conclusively, the proposed method showed superior performance to the previous method which was applied to the constrained minimum spanning tree problem and known as the best algorithm.