Multi-objective optimization technique for resource allocation and task scheduling in vehicular cloud architecture: A hybrid adaptive nature inspired approach

With onset of Intelligent Transport Systems, vehicles are equipped with internet enabled powerful computation units that provide smart driving assistance, along with various infotainment applications. These applications require web assistance and high computation power, which cannot be executed by standalone onboard units of the smart vehicles. Third party infrastructures like centralized cloud and cloudlets are introduced, to meet the requirement for such vehicular, web based, resource hungry applications. Offloading jobs to centralized cloud, exhausts network bandwidth and causes network delay, whereas frequent offloading to cloudlet results in resource starvation due to limited cloudlet resources. These problems lead to the introduction of Vehicular Cloud Computing (VCC) where the onboard units of several local smart vehicles collectively form a cloud. The concept of multi layered cloud brings centralized cloud, cloudlet and vehicular cloud together to coexist and provide on demand services to mobile and vehicular users. In this work, a three tier architecture is proposed consisting of vehicular cloud, roadside cloudlet and centralized cloud. We have developed an optimized resource allocation and task scheduling algorithm to efficiently serve huge number of task requests arriving from on road users, while maintaining improved Quality of Service. These task requests are optimally mapped to cloud resources among the three cloud layers. The optimization process is carried out using the proposed Hybrid Adaptive Particle Swarm Optimization (HAPSO) algorithm which is a combination of Genetic Algorithm and Adaptive Particle Swarm Optimization. Further the proposed model is simulated using SUMO 0.30.0, NS 3.26 and MATLAB R2014a. The results show that for this problem domain, HAPSO converges faster than Standard Particle Swarm Optimization and Self Adaptive Particle Swarm Optimization with ~98.92% reduction in mean square error. The results also show ~34% improvement in task response time and reduced energy consumption upto ~32.5%.