Narrow Results

The dynamics of unmanned aerial systems (UAS) are often nonlinear, especially at high speed and high maneuverability flight. The inherent nonlinearity of the system renders conventional linear control techniques inadequate for achieving optimal control outcomes. Consequently, unmanned aerial vehicle (UAV) trajectory planning is fundamentally a nonlinear optimal control challenge, characterized by the difficulty of swiftly obtaining a stable solution and the propensity to converge on suboptimal local solutions during the solving process. In response to this, a method for autonomous UAV obstacle avoidance trajectory planning is introduced, leveraging a convex optimization-based particle swarm algorithm. The UAV uses sensors to record the information about obstacles, adopts a one-dimensional time-parameterized polynomial trajectory to construct the obstacle avoidance control input, combines the safety penalty function between the navigation path and obstacles to generate the obstacle avoidance trajectory planning objective function, and obtains the path points in the obstacle avoidance trajectory; the particle swarm algorithm is used to solve the objective function, the parameters such as obstacle avoidance control input and safety penalty function are used as particles, and the optimal parameter values are obtained through the iterative updating process of position and speed. Aiming at the acquired obstacle avoidance trajectory path points, the convex optimization algorithm is used to construct a non-convex optimal control model that meets the requirements of energy optimization, and is rooted in the concave-convex process, so that the control model’s objective function and inequality constraint are both convex, while the formula constraint is affine. By transforming the control model into a convex optimization problem, it aligns with the energy optimization requirements. The sequential convex optimization framework is employed to solve this problem, enabling the optimization of the UAV’s trajectory for obstacle avoidance. Experimental outcomes demonstrate that this approach effectively captures the coordinate details of obstacles, navigates through various dense obstacle scenarios, and simultaneously guarantees an energy-efficient path to the target destination.

In order to improve the profitability of coal enterprises, ensure the reasonable allocation of coal resources, and meet the dynamic changes of the coal transportation market and environment, a coal transportation cost optimization method based on improved group intelligence algorithm is proposed. Based on the characteristics of coal transportation, analyze its transportation mode, fully combining the cost characteristics of different transportation modes, construct the cost minimization optimization objective function including road transportation, railroad transportation and sea transportation, and determine the corresponding constraints; through the particle swarm optimization (PSO) algorithm to solve the objective function that meets the constraints. In order to better respond to the market changes, the concave function and dynamic adjustment mechanism are introduced to optimize and adjust the inertia weights and acceleration constants, so that the algorithm can automatically adjust the optimization strategy according to the changes of the market and the environment to ensure that the transportation scheme is always kept in the optimal or near-optimal state, and finally obtain the optimal cost strategy for coal transportation. Through the experimental verification, the method can achieve better coal transportation cost optimization, effectively reduce the carbon emission and waste of equipment resources in the transportation process, improve the efficiency of coal transportation, and effectively save the cost of coal transportation.

While no supervised learning algorithm can do well over all functions, we show that it may be possible to adapt a given function to a given supervised learning algorithm so as to allow the learning algorithm to better classify the original function. Although this seems counterintuitive, adapting the problem to the learner may result in an equivalent function that is "easier" for the algorithm to learn. One method of adapting a problem to the learner is to relabel the targets given in the training data. The following presents two problem adaptation methods, SOL-CTR-E and SOL-CTR-P, variants of Self-Oracle Learning with Confidence-based Target Relabeling (SOL-CTR) as a proof of concept for problem adaptation. The SOL-CTR methods produce "easier" target functions for training artificial neural networks (ANNs). Applying SOL-CTR over 41 data sets consistently results in a statistically significant (p < 0.05) improvement in accuracy over 0/1 targets on data sets containing over 10,000 training examples.

Utilizing dynamics system to identify community structure has become an important means of research. In this paper, inspired by the relationship between topology structures of networks and the dynamic Potts model, we present a novel method that describes the conditional inequality forming simple community can be transformed into the objective function F which is analogous to the Hamilton function of Potts model. Likewise, to detect the well performance of partitioning we develop improved-EM algorithm to search the optimal value of the objective function F by successively updating the dynamic process of the membership vector of nodes which is also commonly influenced by the weighting function W and the tightness expression T. Via adjusting relevant parameters properly, our method can effectively detect the community structures. Furthermore, stability as the new measure quality method is applied for refining the partitions the improved-EM algorithm detects and mitigating resolution limit brought by modularity. Simulation experiments on benchmark and real-data network all give excellent results.

Load balancing, which redistributes dynamic workloads across computing nodes within cloud to improve resource utilization, is one of the main challenges in cloud computing system. Most existing rule-based load balancing algorithms failed to effectively fuse load data of multi-class system resources. The strategies they used for balancing loads were far from optimum since these methods were essentially performed in a combined way according to load state. In this work, a fuzzy clustering method with feature weight preferences is presented to overcome the load balancing problem for multi-class system resources and it can achieve an optimal balancing solution by load data fusion. Feature weight preferences are put forward to establish the relationship between prior knowledge of specific cloud scenario and load balancing procedure. Extensive experiments demonstrate that the proposed method can effectively balance loads consisting of multi-class system resources.

Global optimization of a non-convex objective function often appears in large-scale machine learning and artificial intelligence applications. Recently, consensus-based optimization (CBO) methods have been introduced as one of the gradient-free optimization methods. In this paper, we provide a convergence analysis for the first-order CBO method in [J. A. Carrillo, S. Jin, L. Li and Y. Zhu, A consensus-based global optimization method for high dimensional machine learning problems, https://arxiv.org/abs/1909.09249v1]. Prior to this work, the convergence study was carried out for CBO methods on corresponding mean-field limit, a Fokker–Planck equation, which does not imply the convergence of the CBO method per se. Based on the consensus estimate directly on the first-order CBO model, we provide a convergence analysis of the first-order CBO method [J. A. Carrillo, S. Jin, L. Li and Y. Zhu, A consensus-based global optimization method for high dimensional machine learning problems, https://arxiv.org/abs/1909.09249v1] without resorting to the corresponding mean-field model. Our convergence analysis consists of two steps. In the first step, we show that the CBO model exhibits a global consensus time asymptotically for any initial data, and in the second step, we provide a sufficient condition on system parameters — which is dimension independent — and initial data which guarantee that the converged consensus state lies in a small neighborhood of the global minimum almost surely.

In this paper, two new fuzzy clustering algorithms are proposed based on the global optimization metaheuristic, Threshold Accepting. Their effectiveness is demonstrated in the case of five well-known medium sized data sets viz. Iris, Wine, Glass, E.Coli and Olive oil and a large data set Thyroid. In terms of the least objective functions value, these algorithms named TAFC-1 (Threshold Accepting based Fuzzy Clustering) and TAFC-2 outperformed the well-known Fuzzy C-Means (FCM) algorithm in the case of 4 data sets and in the remaining two data sets, FCM marginally outperformed the TAFC. Xie-Beni cluster validity index is used in arriving at the 'optimal' number of clusters for all the algorithms. Here a novel strategy is proposed whereby the FCM is invoked to find alternative decision vectors whenever the neighbourhood search fails in its pursuit. This hybrid scheme has worked well. In conclusion, these new algorithms can be used as viable and efficient alternatives to the FCM algorithm.

In this paper, an optimization method is used to determine the values of partial factors in structural reliability analysis. Once the proper objective function is defined, a group of optimum partial factors, which enable the objective function to take its minimum value, will need to be determined. In the present study, two kinds of objective function are considered. The conditions that have to be satisfied for optimum partial factors of these two kinds of objective function are then derived. In both cases, the result shows that the partial factors of both dead and live loads should satisfy the same proportional expression and should be inversely proportional to the partial factor of resistance force. A simple beam is used as an example to illustrate the computations involved. It is found that the design concept proposed in this paper leads to a design criterion similar to that which applies to the conventional deterministic method. Thus, this concept can be easily used in practice. The illustrative example shows that the values of the dead load and live load have a significant effect on the reliability design criteria.

This paper deals with the simultaneous identification of road roughness and vehicle parameters, considering the effect of vehicle–structure interaction. The proposed technique avoids the use of bridge response data (which has practical implementation difficulties along with the high chances of corruption with environmental noises) and utilizes the vehicle response data (which is relatively easier to record). Further, vehicle calibration is not needed as the roughness is estimated simultaneously. The identification is carried out by the coupling of an unbiased minimum variance estimator with an optimization scheme. This study considers a quarter-car vehicle model and a half-car vehicle model, instrumented to measure the vehicle vibration data. The unbiased minimum variance estimator (MVE) allows a linear temporal evolution of the state variables, incorporating the roughness as an unknown input term such that the need for linearization is avoided, unlike the traditional nonlinear filters. The optimization scheme helps in choosing a set of optimal solutions for the vehicle parameters as designed in the coupled scheme. The best split of the available measurement data to be used in the two schemes (MVE and optimization scheme) is discussed. The effect of different objective functions is also studied. The proposed technique is successful in terms of simultaneously estimating the vehicle parameters, roughness profile and vehicle responses (states) accurately.

In practical civil engineering, structural damage identification is difficult to implement due to the shortage of measured modal information and the influence of noise. Furthermore, typical damage identification methods generally rely on a precise Finite Element (FE) model of the monitored structure. Pointwise mass alterations of the structure can effectively improve the quantity and sensitivity of the measured data, while the data fusion methods can adequately utilize various kinds of data and identification results. This paper proposes a damage identification method that requires only approximate FE models and combines the advantages of pointwise mass additions and data fusion. First, an additional mass is placed at different positions throughout the structure to collect the dynamic response and obtain the corresponding modal information. The resulting relation between natural frequencies and the position of the added mass is sensitive to local damage, and it is thus utilized to form a new objective function based on the modal assurance criterion (MAC) and $l_1$-based sparsity promotion. The proposed objective function is mostly insensitive to global structural parameters, but remains sensitive to local damage. Several approximate FE models are then established and separately used to identify the damage of the structure, and then the Dempster–Shafer method of data fusion is applied to fuse the results from all the approximate models. Finally, fractional data fusion is proposed to combine the results according to the parametric probability distribution of the approximate FE models, which allows the natural weight of each approximate model to be determined for the fusion process. Such an approach circumvents the need for a precise FE model, which is usually not easy to obtain in real application, and thus enhances the practical applicability of the proposed method, while maintaining the damage identification accuracy. The proposed approach is verified numerically and experimentally. Numerical simulations of a simply supported beam and a long-span bridge confirm that it can be used for damage identification, including a single damage and multiple damages, with a high accuracy. Finally, an experiment of a cantilever beam is successfully performed.

This chapter describes the Nelder–Mead (NM) simplex method, the Broyden–Fletcher–Goldfarb–Shanno (BFGS) and conjugate gradient (CG) methods. They are then applied to estimate parameters in a finite element model given the experimental data. The NM method performed better than the other methods.

This chapter describes the particle swarm optimization (PSO) method. Its evolution, theory and applications are explored. Examples on how it has been applied for missing data estimation, operating system scheduling as well as finite element updating are explored. The results obtained showed that the particle swarm optimization gave good results.

A hybrid has two or more components that produce the same or better results, for example, a vehicle powered by both an electric motor and an internal combustion engine as sources of power for the drive train. We apply the hybrid algorithm of the particle swarm optimization (PSO) and Nelder–Mead (NM) simplex method for finite element model (FEM) updating. The results observed showed that on average the hybrid gave results that were more accurate, followed by the PSO and then the NM simplex method.

Clustering is primarily used to uncover the true underlying structure of a given data set. Most algorithms for clustering often depend on initial guesses of the cluster centers and assumptions made as to the number of subgroups presents in the data. In this paper, we propose a method for fuzzy clustering without initial guesses on cluster number in the data set. Our method assumes that clusters will have the normal distribution. Our method can automatically estimate the cluster number and form the clusters according to the number. In it, Genetic Algorithms (GAs) with two chromosomic coding techniques are evaluated. Graph structured coding can derive high fitness value. Linear structured can save the number of generation.