Collaborative Task Offloading And Data Caching Based On Cloud-edge

With the development of 5G communication technology,Internet technology,and Internet of Things(IoT)technology,mobile games,image and video processing,streaming media,and other services on mobile devices have become more and more widespread.The demand for the coordinated development of "communication,computation,and storage"in the IoT is becoming increasingly urgent.To optimize the computing resources of mobile devices,computation tasks generated by mobile devices can be offloaded to edge clouds,central clouds,other mobile devices or performed locally through collaborative computing,taking into account the characteristics of different "communication,computation,and storage" resources of cloud computing and edge computing.This technology has attracted more and more attention from professionals in related fields.Currently,optimization algorithms for task offloading and data caching strategies in mobile edge computing often use swarm intelligence algorithms,game theory,reinforcement learning,and other derived methods.However,there are some shortcomings in related research.Firstly,traditional optimal task offloading and data caching strategies do not consider the problem of channel resource allocation,which can lead to inefficient use of limited channel resources.Secondly,current research on task offloading and data caching mainly utilizes end-edge collaboration and end-edge cloud collaboration,but rarely considers incorporating other mobile devices into the target device for task offloading.Finally,the use of DDPG algorithm in solving task offloading and data caching in mobile edge computing is limited by the influence of factors such as the initial state,the generalization ability of neural networks,and the fact that the task offloading problem itself is a non-convex optimization problem,which can easily lead to local optimal solutions and algorithm instability.To address these issues,this paper focuses on the following research topics.Firstly,this paper mainly considers reducing mutual interference between multiple channels in channel resource allocation,and proposes a channel resource allocation algorithm based on an improved firefly algorithm.In the process of improving the firefly algorithm,inspired by the mutation mechanism in genetic algorithms and the inertia coefficient in basic particle swarm algorithms,the inertia coefficient and mutation mechanism are introduced into the firefly algorithm.Compared with the original firefly algorithm,the improved firefly algorithm not only improves the convergence speed of the algorithm and overcomes the disadvantage of being trapped in local optimal solutions but also enhances the global optimization ability in the early stage.Secondly,although the computing performance of mobile devices is weaker than that of edge cloud devices,they can still execute certain tasks with low computing performance requirements.Therefore,this paper incorporates mobile devices that are in sleep mode or have no tasks to perform into the target device for task offloading and proposes a collaborative task offloading and data caching strategy based on end-toend edge cloud collaboration.Through simulation,it is found that compared with offloading tasks to central clouds,edge clouds,local devices,or performing end-edge collaboration or end-edge cloud collaboration,the end-to-end edge cloud collaboration-based task offloading can complete more tasks within the maximum tolerable delay.Thirdly,to solve the dilemma of DDPG algorithm being easily trapped in local optimal solutions when performing task offloading and data caching,this paper proposes an algorithm that integrates DDPG algorithm with the improved firefly algorithm.The main idea of this algorithm is to use the historical best position and global best position of the firefly algorithm to update the policy parameters when integrating the firefly algorithm,so that the algorithm can search for the optimal solution globally.Meanwhile,due to the local optimization ability of the DDPG algorithm,it can help to fine-tune the algorithm after finding the global optimal solution,thereby achieving better performance.Through simulation,it is found that compared with using the DDPG algorithm alone,the algorithm that integrates the DDPG algorithm with the firefly algorithm can not only improve the global search ability and avoid being trapped in local optimal solutions but also increase the convergence speed of the algorithm in the early stage,making the algorithm enter the convergence stage more quickly.