Radial basis function networks-based resource-aware offloading video analytics in mobile edge computing

With the advancement of wireless networks, many mobile applications are becoming increasingly popular. These applications, which include real-time vision processing, intelligent homes, accurate tracking, and traffic monitoring, frequently necessitate an excellent experience that requires expensive computational resources. Video analytics on mobile devices necessitate a significant amount of computational power, resulting in longer processing times. Even though mobile device (MD) performance has steadily improved, running all programs on a single MD still results in excessive energy consumption and delay. The problem can be partially solved by outsourcing processing to the cloud, but uploading videos take a long time. Mobile edge computing can transfer processing to close-by edge servers to lower latency (MEC). Despite this, the performance of video analytics must be improved because edge server processing resources are frequently limited and highly dynamic. This paper aims to maximize utility, a weighted function of frame rate and precision. The lack of server and network expertise and the ever-changing system environment represent formidable obstacles to fixing this issue. However, current resource offloading systems concentrate mainly on average-based performance indicators, so missing the resource's deadline limitation. This research presents a resource-aware offloading video analytics in mobile edge computing and a resource-aware offloading algorithm (ROA) using the radial basis function networks (RBFN) method for enhancing the reward under the resource's deadline restriction. Brute-force search is then used to simplify the computation, with concave processing and convex optimization improving resource allocation and user association. According to extensive experimental data, RBFN-OROA outperforms the comparison algorithms in all parameter settings, demonstrating its resistance to MEC environment state changes.