سمپوزیوم منطقه ای نوآوری در علم و فناوری

صفحه اصلی / اولین سمپوزیوم منطقه ای نوآوری در علم و فناوری

Increasing resource allocation performance in DRL-based mobile edge calculations

نویسندگان :

Afshin Golchin¹ Vahid Sattari Naini²

1- دانشگاه شهید باهنر کرمان 2- دانشگاه شهید باهنر کرمان

کلمات کلیدی :

increasing resource allocation performance،mobile edge computing،DRL،Strengthen deep learning

چکیده :

Mobile edge computing is a special technology that optimizes the allocation of computing resources. This paper, method to increase resource allocation performance in mobile edge computing using Deep Reinforcement Learning (DRL) is presented. DRL algorithms are used as the main mechanism for allocating resources. Algorithms make decisions about optimal resource allocation using the collected experience that significantly improves system performance compared to traditional methods. Various evaluation criteria are evaluate the accuracy of this research compared to the basic method. Include percentage of resource usage, execution time, number of successful and unsuccessful requests and …. The obtained experiments show that the important parameters in this method include the cost function (Loss), size of batches and prediction of performance increase requests in basic method. Correct setting of these parameters can improve its performance and accuracy. The combination of virtual machine and meta-heuristic algorithm is used as deep reinforcement learning, and with its help, we improve resource allocation and load balance in mobile edge computing. Today, logarithmic increase in the volume of information and the need for new data centers or the development of previous data centers, and increase in speed of energy consumption calculations, information warehouses are among the most important challenges.