Estimation Of Distribution Algorithm Research Articles

Bound-guided hybrid estimation of distribution algorithm for energy-efficient robotic assembly line balancing

Under the pressure of climate change, energy-efficient manufacturing has attracted much attention. Robotic assembly lines are widely-used in automotive and electronic manufacturing. It is necessary to consider the energy saving and economic criteria simultaneously when robots are utilized to operate assembly tasks replacing human labor. This paper addresses an energy-efficient robotic assembly line balancing (EERALB) problem with the criteria to minimize both the cycle time and total energy consumption. We present a multi-objective mathematical model and propose a bound-guided hybrid estimation of distribution algorithm to solve the problem. When designing the optimization algorithm, we adopt estimation of distribution algorithm (EDA) to tackle the task assignment, and design a non-dominated robot allocation (NGRA) heuristic which is embedded into the EDA to allocate suitable robot to each workstation. Moreover, we propose a bound-guided sampling (BGS) method, which is able to reduce the search space of EDA and focus the search on the promising area. The computational complexity of the proposed algorithm is analyzed and the effectiveness of the proposed NGRA and BGS is tested. In addition, we compare the performances of the proposed mathematical model and the proposed algorithm with those of the existing model and algorithms on a set of widely-used benchmark instances. Comparative results demonstrate the effectiveness of the proposed model and algorithm.

Distributed optimal linear fusion estimators

This paper is concerned with the distributed information fusion estimation problem in multi-sensor environments. A universal distributed optimal linear fusion estimation (DOLFE) algorithm, which has a Kalman-type structure with matrix gains, is presented under the linear unbiased minimum variance criterion. To reduce the computational burden, two suboptimal linear fusion estimation algorithms with diagonal-matrix gains and scalar gains are also presented. Based on the proposed DOLFE algorithm, a distributed optimal linear fusion filter (DOLFF) is presented for multi-sensor linear discrete-time stochastic systems. It has better accuracy than that based on the matrix-weighted fusion of local estimators but worse accuracy than the centralized fusion filter (CFF). The stability and steady-state property of the proposed DOLFF are analyzed. Then, the corresponding multi-step predictor and smoother are also developed based on DOLFF. To obtain the distributed fusion estimators, some estimation error cross-covariance matrices that are used to compute the gains are derived. At last, distributed optimal linear fusion estimators with feedback are also presented. Furthermore, it is strictly proved that the proposed distributed optimal linear fusion filter with feedback (DOLFFWF) has the same accuracy as the CFF. Two simulation examples show the effectiveness of the proposed algorithms.

Event-triggered distributed state estimation over wireless sensor networks

This paper focuses on the event-triggered distributed state estimation over sensor networks, in which each sensor node selectively transmits the local information to its neighbors for the reduced communication bandwidth and the prolonged network lifetime. Based on an individual stochastic triggering condition, an event-triggered minimum mean square error (MMSE) estimator is proposed in a recursive form, and then an event-triggered distributed state estimation algorithm is developed by repeatedly fusing the local information and the event-triggered information. It is shown that, under network connectivity, collective observability and large enough triggering parameters, the distributed estimator in each sensor node is stable with the uniformly bounded estimation error in mean square. Finally, a target tracking example is provided to illustrate the practical effectiveness of the proposed technique.

A GPU-Enabled Compact Genetic Algorithm for Very Large-Scale Optimization Problems

The ever-increasing complexity of industrial and engineering problems poses nowadays a number of optimization problems characterized by thousands, if not millions, of variables. For instance, very large-scale problems can be found in chemical and material engineering, networked systems, logistics and scheduling. Recently, Deb and Myburgh proposed an evolutionary algorithm capable of handling a scheduling optimization problem with a staggering number of variables: one billion. However, one important limitation of this algorithm is its memory consumption, which is in the order of 120 GB. Here, we follow up on this research by applying to the same problem a GPU-enabled “compact” Genetic Algorithm, i.e., an Estimation of Distribution Algorithm that instead of using an actual population of candidate solutions only requires and adapts a probabilistic model of their distribution in the search space. We also introduce a smart initialization technique and custom operators to guide the search towards feasible solutions. Leveraging the compact optimization concept, we show how such an algorithm can optimize efficiently very large-scale problems with millions of variables, with limited memory and processing power. To complete our analysis, we report the results of the algorithm on very large-scale instances of the OneMax problem.

A Distributed Robust Power System State Estimation Approach Using $t$-Distribution Noise Model

In practical power system applications, the distribution of the measurement noise is sometimes unknown and deviates from the assumed Gaussian noise model due to measurement outliers. In such cases, the performances of the state estimators based on Gaussian noise assumption may deteriorate significantly. In this article, we propose a fully distributed and robust power system state estimation approach based on the $t$ -distribution noise model and the maximum likelihood criterion. The $t$ -distribution is used to model Gaussian and non-Gaussian statistics in the field of robust statistics. The matrix-splitting techniques are employed to carry out the extensive matrix inversion of the gain matrix in a distributed way to achieve efficient computation. In the proposed distributed estimation framework, each local control area only requires limited data exchange with its neighboring areas. Simulations on the IEEE 14-bus, 118-bus and 300-bus systems are used to verify the effectiveness and robustness of the proposed distributed state estimation algorithm.

A novel approach to solve AI planning problems in graph transformations

The aim of AI planning is to solve the problems with no exact solution available. These problems usually have a big search space, and planning may not find plans with the least actions and in the shortest time. Recent researches show that using suitable heuristics can help to find desired plans. In planning problems specified formally through graph transformation system (GTS), there are dependencies between applied rules (actions) in the search space. This fact motivates us to solve the planning problem for a small goal (instead of the main goal), extract dependencies from the searched space, and use these dependencies to solve the planning problem for the main goal. In GTS based systems, the nodes of a state (really is a graph) can be grouped due to their type. To create a small (refined) goal, we use a refinement technique to remove the predefined percent of nodes from each group of the main goal. Bayesian Optimization Algorithm (BOA) is then used to solve the planning problem for the refined goal. BOA is an Estimation of Distribution Algorithm (EDA) in which Bayesian networks are used to evolve the solution populations. Actually, a Bayesian network is learned from the current population, and then this network is employed to generate the next population. Since the last Bayesian network learned in BOA has the knowledge about dependencies between applied rules, this network can be used to solve the planning problem for the main goal. Experimental results on four well-known planning domains confirm that the proposed approach finds plans with the least actions and in the lower time compared with the state-of-the-art approaches.

Globally optimal distributed time-varying weight information filter of mobile sensor network

The networked state estimation for a linear dynamic system is studied in this paper. Firstly, a time-varying distributed communication protocol governed by the network topology is introduced for delivering information among the neighboring sensors. Secondly, based on this protocol, a two stage distributed state estimation algorithm is presented to achieve the global optimal results in a least square sense. In the first stage of the proposed algorithm, distributed computation on the sum of global information vector contributions is conducted, and the second stage is to derive the state estimations via utilizing the interaction results offered by the first stage. Thirdly, it is shown that the proposed algorithm only demands finite neighbor interaction times to obtain global optimal estimations. The optimality and stability properties of the proposed algorithm are also proved by using the Riccati equation. Finally, through numerical simulations, it is further illustrated that the proposed algorithm could outperform the previous distributed estimators.

Distributed state estimation for uncertain linear systems: A regularized least-squares approach

This paper addresses the state estimation problem for a discrete-time uncertain system with a network of sensors, where the system is not necessarily observable by each sensor and deterministic uncertainties exist in the system matrices. A new robust estimator is designed for each sensor, using only its own and neighbor’s information, which is fully distributed. Moreover, a novel information fusion strategy is developed to guarantee the estimation performance, based on the collective observability of the sensor network, which greatly relaxes the technical assumption of the proposed estimator. Theoretically, it can be ensured that if the observed system is time-varying, the gains of the estimator will be bounded. Furthermore, if the system is time-invariant, these gains will be convergent. Subsequently, the estimation error covariance will be ultimately bounded if the observed system is quadratically bounded. In the end, the superiority of the proposed robust distributed state estimation algorithm is illustrated by several numerical simulation examples.

Collaborative Learning of Estimation of Distribution Algorithm for RNA secondary structure prediction

Estimation of distribution algorithms (EDAs) are successfully applied in the fields of bioinformatics for tasks such as gene structure analysis, protein structure prediction, and RNA secondary structure prediction. This paper proposes a new method, namely collaborative learning of estimation of distribution algorithms, or Co-EDAs, based on an estimation of distribution algorithm for RNA secondary structure prediction using a single RNA sequence as input. The proposed method consists of two EDAs with minimum free energy objective. The Co-EDAs use both good and poor solutions to improve the algorithm’s to search throughout the search space. Using information from poor solutions can indicate which area is unappealing to explore when searching with high-dimensional data. The Co-EDAs method was tested with 750 known RNA structures from RNA STRAND v2.0. That database includes data with more than 14 RNA types. The proposed method was compared to three prediction programs that are based on dynamic programming algorithms called Mfold, RNAfold, and RNAstructure. These programs are available as services on web servers. The results on average show that the Co-EDAs yields approximately 6% better accuracy than those competitors in all metrics.

Sharp Bounds for Genetic Drift in Estimation of Distribution Algorithms

Estimation of distribution algorithms (EDAs) are a successful branch of evolutionary algorithms (EAs) that evolve a probabilistic model instead of a population. Analogous to genetic drift in EAs, EDAs also encounter the phenomenon that the random sampling in the model update can move the sampling frequencies to boundary values not justified by the fitness. This can result in a considerable performance loss. This article gives the first tight quantification of this effect for three EDAs and one ant colony optimizer, namely, for the univariate marginal distribution algorithm, the compact genetic algorithm, population-based incremental learning, and the max-min ant system with iteration-best update. Our results allow to choose the parameters of these algorithms in such a way that within a desired runtime, no sampling frequency approaches the boundary values without a clear indication from the objective function.

Data Estimation of Injury-Prone Sports by 3D Image Joint Probability Distribution

The objective of the paper is to study and analyze sports injuries from multiple perspectives. The reduction of damage risk factors of young athletes in sports would reduce or avoid the occurrence of sports injuries. The obtained point cloud data is preprocessed by three-dimensional (3D) image technology for the registration of the cloud data to be registered. The parallel Bayesian network structure was established to realize the parallelization of the probability distribution estimation algorithm. The research results have shown that the parallel probability distribution estimation algorithm would provide a great degree of efficiency improvement in terms of solving continuous function optimization and real-time optimization problems, as well as the problems of improving the efficiency of the algorithm to some extent. It could be inferred that the combination of improved rasterized nearest point search method has shortened the search time of the nearest point, thereby improving the efficiency of the algorithm. The correctness and rationality of the algorithm were verified from both theoretical and practical perspectives.

Estudio de factibilidad para la aplicación de algoritmos de estimación de distribución al problema de secuenciación de vehículos

The main objective of an automobile production plant is to deliver on time and form the orders that are received daily. These orders are not homogeneous since they involve large quantities of cars that generally belong to different models and must be painted in different colors. The car sequencing problem that takes these characteristics into account was proposed by the Renault Company in 2005 as part of the ROADEF Challenge. This problem is NP-Hard and various techniques have been proposed to solve it, from exact methods to different heuristic algorithms. This work presents a feasibility study to apply two Distribution Estimation Algorithms (EDAs) to solve this problem. In addition, three important aspects are presented: the adaptation process of the algorithms, a technique for the execution of the algorithms called the "Stepped Approach with Discard" and a methodology that involves tolerance in the substitution of the individuals. The results obtained by the algorithms are also shown. The analysis of the results shows the algorithms adaptation process and the adjustments that can be made to improve their competence with the state of art.

Simulation-Based EDAs for Stochastic Programming Problems

With the rapid growth of simulation software packages, generating practical tools for simulation-based optimization has attracted a lot of interest over the last decades. In this paper, a modified method of Estimation of Distribution Algorithms (EDAs) is constructed by a combination with variable-sample techniques to deal with simulation-based optimization problems. Moreover, a new variable-sample technique is introduced to support the search process whenever the sample sizes are small, especially in the beginning of the search process. The proposed method shows efficient results by simulating several numerical experiments.

The Algorithms of Distributed Learning and Distributed Estimation about Intelligent Wireless Sensor Network

The intelligent wireless sensor network is a distributed network system with high “network awareness”. Each intelligent node (agent) is connected by the topology within the neighborhood which not only can perceive the surrounding environment, but can adjusts its own behavior according to its local perception information to constructs a distributed learning algorithms. Therefore, three basic intelligent network topologies of centralized, non-cooperative, and cooperative are intensively investigated in this paper. The main contributions of the paper include two aspects. First, based on algebraic graph, three basic theoretical frameworks for distributed learning and distributed parameter estimation of cooperative strategy are surveyed: increment strategy, consensus strategy, and diffusion strategy. Second, based on classical adaptive learning algorithm and online updating law, the implementation process of distributed estimation algorithm and the latest research progress of above three distributed strategies are investigated.

Combining information from thresholding techniques through an evolutionary Bayesian network algorithm

Segmentation is an important task in image processing because it could affect the performance of other steps in image analysis. One of the most used methods for segmentation is thresholding which can be formulated as an optimization problem, and evolutionary algorithms (EAs) are alternatives commonly applied to solve it. Estimation of Distribution Algorithms (EDAs) is a branch of EAs that explores the search space by building a probabilistic model, such as Bayesian Networks (BNs). In this article is proposed a BN-based EDA for multilevel image segmentation called BNMTH . The proposed approach iteratively selects the combination of thresholding techniques that permits to find the best configuration of thresholds for a digital image, exploring the inter-dependencies between the decision variables (thresholds) and the different techniques. BNMTH is applied over a set of benchmark images and the results of the segmentation are qualitatively analyzed by using the Peak Signal-to-Noise Ratio (PSNR), the Structure Similarity Index (SSIM) and the Feature Similarity Index (FSIM). Besides, a statistical analysis is provided to compare BNMTH with other state-of-the-art optimization algorithms. The results show that BNMTH is a competitive approach for image segmentation, providing accurate results in almost all the cases. Moreover, the segmented images and the histograms show that the classes are accurately generated even in complex conditions.

Estimation of distribution-based multiobjective design space exploration for energy and throughput-optimized MPSoCs

Modern multicore architectures comprise a large set of components and parameters that require being matched to achieve the best balance between power consumption and throughput performance for a particular application domain. The exploration of design space for finding the best power throughput trade-off is a combinatorial optimization problem with a large number of combinations, and. in general, black its solution is prohibitively difficult to be explored exhaustively. However, fortunately, evolutionary algorithms (EAs) have the potential to efficiently solve this problem with reasonable computational complexity. In this paper, we consider a multiobjective design space exploration (DSE) problem with two conflicting objectives. The first objective corresponds to power consumption minimization while the second objective relates to throughput maximization. We approach this problem by employing the estimation of distribution algorithm (EDA), which belongs to the family of EAs. The proposed EDA-based DSE (EDA-DSE) scheme efficiently selects the design parameters (i.e. cache size, number of cores, and operating frequency) with an efficient power through put ratio. The proposed scheme is verified using cycle accurate simulations over a set of benchmarks and the simulation results show a significant reduction in energy delay product for all benchmark applications when compared to the default baseline configuration and genetic algorithm.

A cooperative coevolution algorithm for multi-objective fuzzy distributed hybrid flow shop

With consideration of uncertainty in the distributed manufacturing systems, this paper addresses a multi-objective fuzzy distributed hybrid flow shop scheduling problem with fuzzy processing times and fuzzy due dates. To optimize the fuzzy total tardiness and robustness simultaneously, a cooperative coevolution algorithm with problem-specific strategies is proposed by reasonably combining the estimation of distribution algorithm (EDA) and the iterated greedy (IG) search. In the EDA-mode search, a problem-specific probability model is established to reduce the solution space and a sample mechanism is proposed to generate new individuals. To enhance exploitation, a specific local search is designed to improve performance of non-dominated solutions. Moreover, destruction and reconstruction methods in the IG-mode search are employed for further exploiting better solutions. To balance exploration and exploitation capabilities, a cooperation scheme for mode switching is designed based on the information entropy and the diversity of elite solutions. The effect of the key parameters on the performances of the proposed algorithm is investigated by Taguchi design of experiment method. Comparative results and statistical analysis demonstrate the effectiveness of the proposed algorithm in solving the problem.

Performance Analysis of Distributed Estimation for Data Fusion Using a Statistical Approach in Smart Grid Noisy Wireless Sensor Networks.

Internet of Things (IoT) can significantly enhance various aspects of today’s electric power grid infrastructures for making reliable, efficient, and safe next-generation Smart Grids (SGs). However, harsh and complex power grid infrastructures and environments reduce the accuracy of the information propagating through IoT platforms. In particularly, information is corrupted due to the measurement errors, quantization errors, and transmission errors. This leads to major system failures and instabilities in power grids. Redundant information measurements and retransmissions are traditionally used to eliminate the errors in noisy communication networks. However, these techniques consume excessive resources such as energy and channel capacity and increase network latency. Therefore, we propose a novel statistical information fusion method not only for structural chain and tree-based sensor networks, but also for unstructured bidirectional graph noisy wireless sensor networks in SG environments. We evaluate the accuracy, energy savings, fusion complexity, and latency of the proposed method by comparing the said parameters with several distributed estimation algorithms using extensive simulations proposing it for several SG applications. Results prove that the overall performance of the proposed method outperforms other fusion techniques for all considered networks. Under Smart Grid communication environments, the proposed method guarantees for best performance in all fusion accuracy, complexity and energy consumption. Analytical upper bounds for the variance of the final aggregated value at the sink node for structured networks are also derived by considering all major errors.

The Estimation of Particle Swarm Distribution Algorithm With Sensitivity Analysis for Solving Nonlinear Bilevel Programming Problems

This paper proposes an estimation of particle swarm distribution algorithm (EPSDA) to solve the nonlinear bilevel programming problem (NBLP) by embedding the estimation distribution algorithm (EDA) into the particle swarm optimization (PSO). One Gaussian function is selected to construct the probability distribution for the superior candidate from the present population before executing the velocity and position update rule at each iteration in PSO. Thus, some new individuals viewed as an offspring population are generated from the probability distribution to replace some inferior particles in the current population for making up a new population. Therefore, this proposed algorithm combines PSO (the local search method) and EDA (the global search method) through updating the population to enhance the efficiency of solving NBLP. In experiments, we select four representative examples for linear, quadratic, nonlinear, high-dimensional nonlinear cases to carry out sensitivity analysis on parameters of the proposed algorithm. The results reveal that linearly decreasing inertia weight and adaptive acceleration coefficients are better than constant parameters. Furthermore, the multivariate Gaussian distribution can achieve better performance compared with the normal distribution. These four examples are also used to compare the performance of EPSDA with ones of separate PSO and EDA by setting constant parameters. The experiments show that EPSDA is better than separate PSO and EDA from both the quality of solution and the computational efficiency. The results on four high-dimensional non-convex nonlinear examples demonstrate feasibility and efficiency of EPSDA to solve the high-dimensional nonlinear bilevel programming from the iteration number and computational time.

A comparative study on evolutionary algorithms for the agent routing problem in multi-point dynamic task

The agent routing problem in multi-point dynamic task (ARP-MPDT) proposed recently is a novel permutation optimisation problem. In ARP-MPDT, a number of task points are located at different places and their states change over time. The agent must go to the task points in turn to execute the tasks, and the execution time of each task depends on the task state. The optimisation objective is to minimise the time for the agent to complete all the tasks. In this paper, five evolutionary algorithms are redesigned and tried to solve this problem, including a permutation genetic algorithm (GA), a variant of the particle swarm optimisation (PSO) and three variants of the estimation of distribution algorithm (EDA). In particular, a dual-model EDA (DM-EDA) employing two probability models was proposed. Finally, comparative tests confirm that the DM-EDA has a stronger adaptability than the other algorithms though GA performs better for the large-scale instances.