This paper presents an evaluation of different methods used to deliver virtual machines capable of being accessed remotely by thin-clients. In this paper, the satellite payloads’ remote testing system can able to support the function of real-time processing and analyzing for large amount of data on the remote side, so that the limited expert power can monitoring and testing the satellites been laid on the launching place from the testing hall of Beijing. In the last part of this paper, the author gives a concise introduction to the improvement and optimization direction of this platform. In the second part of this paper, the author introduced the application and characteristics of this system in detail. In the first part of this paper, the author introduces the overall architecture design of the system Then describe the design idea, implementation method and running process of the remote-test-service- platform and remote-terminal in detail At the same time, the design and implementation of image partition compression technology and data frame format for data communication within the platform are described in detail Finally, the operation process and operation logic of the platform are introduced. This paper presents one method and application for remote real-time processing & analyzing of payload data, describes the implementation principle of this method and the details of each functional module in the system. Also, since the company found some opportunities to improve the accuracy of their own data, the new strategy was compared against two other algorithms using other test problems as complex as the original one, comparing both their computer resource usage (memory and processor) as well as the hypervolume for the Pareto front approximations and the approximations in themselves. With low resource usage and with faster results arose. While this algorithm was executed and provided viable results, it used too much computer resources and took too long to completely run due to the large number of variables in the problem.Īsafeedbackoftheseoutcomes,theneedtocreateamulti-objectiveoptimizationalgorithm As such, a multi-objective optimization algorithm could address the scenario considering the maximization of the reliability of the overall plan (determined from the reliability rates of the slaughterhouses) as another objective. However, this strategy did not consider the reliability of each slaughterhouse - each slaughterhouse covered only partial quantities of the proposed production plan determined by the algorithm and, therefore, the profits informed by the algorithm differed from the real values. This strategy originated from the situation of a large meat company which used a single objective optimization algorithm to generate the production plan for its slaughterhouses, being the objective the maximization of the theoretical profits. Other than this, it also must be able to attempt to quickly reach the real Pareto front and be easy to evaluate, adapt and port. containing tens of thousands of variables or more). With these strategies, the proposed strategy aims to allow the evaluation of problems containing a large number of variables (i.e. For example, in the proposed strategy, some parameters enable the reduction on the number of solutions being transferred from one generation to the other or the number of solutions being evaluated in the local, stochastic search algorithm. By computer resources, it is understood as the memory and processor usage. The proposed hybrid strategy is intended to have lower computer resource usage through parallelization and new parameters to be used in order to better use these resources as needed. Considering these reasons, this research proposes a new hybrid (combining a genetic algorithm approach with local search) optimization strategy for multi-objective continuous problems having such characteristics while keeping an acceptable performance to be deployed in industrial applications such as in the meat industry. It happens due to their low performance combined with a high resource usage as well as sometimes being unable to reach and generate an adequate solution set. However, these strategies do not adapt very well to some real problems that have a high amount of variables. The latter are able to solve multi-objective problems with varying amounts of variables and constraints. In the optimization field it is common to find algorithms originally capable of addressing problems with a single objective (such as direct search methods including Nelder-Mead simplex method, Hooke-Jeeves and pattern search) and other strategies known to address multi-objective problems (multi-objective evolutionary algorithms, for instance).
0 Comments
Leave a Reply. |