Um framework orientado a contexto e algoritmos de decisão para offloading computacional em computação de borda veicular

• Autor(a) principal: Souza, Alisson Barbosa de
• Data de Publicação: 2021
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional da Universidade Federal do Ceará (UFC)
• Texto Completo: http://www.repositorio.ufc.br/handle/riufc/59823
• topic: Computation offloading; Vehicular edge computing; Contextual information; Artifical bee colony; Vehicular networks

Autonomous vehicles and complex vehicular applications have become increasingly popular and require massive computational resources. Although vehicles are becoming more connected and intelligent, they still do not have enough computation power to satisfy these demands satisfactorily. One option to deal with this challenge is to allow computing resources from neighboring vehicles and edge servers coupled to base stations to be used through vehicular edge computing systems. Then, vehicles can send tasks, or smaller parts of applications, to these remote servers through the computation offloading technique. In this technique, such servers execute the tasks and return the processing result to the initial vehicle. Although this technique aims to reduce application execution time, performing it in vehicular scenarios is challenging due to the fast movement of network nodes and the frequent disconnections. In such cases, contextual information that characterizes the situation of network devices and vehicles helps to deal with these challenges by assisting offloading decision processes in delivering better results. Thus, we propose a context-oriented framework and task assignment algorithms to reduce the execution time of vehicular applications reliably through computation offloading in vehicular edge computing systems. The framework manages all stages of the offloading process and provides a failure recovery mechanism. The main module of this framework allows the proposed algorithms to assign application tasks to different servers, using contextual parameters and WAVE and 5G networks. Experimental results show that our solutions can significantly reduce the execution time of vehicular applications. Based on the artificial bee colony metaheuristic, our best algorithm achieves that this average reduction reaches up to 75.6% compared to local execution and up to 57.9% compared to literature algorithms, with up to 0.0% of failures. These results show that the proposed solutions are a promising alternative to enable the execution of complex vehicular applications.