Research On Computation Offloading And Caching For Edge Networks

Empowered by technologies such as 5G and artificial intelligence,the Internet of Things(Io T)has experienced rapid development,and all parties have also gathered to respond to the edge computing architecture model.However,the limited computational and bandwidth resources of edge nodes and the lack of comprehensive consideration of existing research around AI algorithms for edge network resource allocation,task computation offloading and edge caching have led to a large number of offloading tasks that cannot be efficiently processed,thus causing high task latency problems.For the above deficiencies and establish a more efficient and reasonable network service model,this thesis investigates the computation offloading and caching mechanism for edge networks.The specific contributions include the following three aspects:1)Computation offloading with task delay minimization under energy constraint:To address the challenge of traditional cloud computing architecture being unable to handle a large amount of tasks generated by the quickly increasing number of Io T devices in an efficient and timely manner,a computation offloading with task delay minimization with energy constraints mechanism is proposed.The mechanism takes into account the characteristics of edge computing architecture,which are proximity to device terminals and high reliability,as well as the advantages of task offloading,which are to compensate for local computing power deficits and make full use of system resources.Based on this,a delay minimization optimization problem is formulated that combines task offloading decision-making and computation resource allocation under system energy constraints.To effectively solve this mixed-integer nonlinear problem,the resource allocation for non-binary variables is first solved to determine the optimal computation resource allocation.Secondly,a task delay minimization offloading decision algorithm based on accelerated simulated annealing and genetic algorithm(ASAG)is proposed to efficiently obtain the optimal binary variable-task offloading decision set.This algorithm combines the global search capability of genetic algorithm and the local search capability of simulated annealing algorithm,and dynamically updates the temperature decrease using the number of local optimal solutions to accelerate algorithm convergence.Experimental results show that the proposed solution outperforms the other three schemes in terms of task delay and energy consumption,which demonstrates the effectiveness of the constructed model and the high efficiency of the proposed algorithm.2)Deep reinforcement learning-based collaborative edge computing offloading and caching:In view of the resource limitation of a single edge node,it is difficult to deal with a large number of tasks uploaded by devices in time,and the traditional algorithms are inefficient in solving high-dimensional variable problems and easy to fall into local optimal solutions.Besides,a large number of Io T devices have a certain of repeating requests for the same task.According to the above analysis,a mechanism for collaborative edge computing offloading and caching is proposed.Based on the low-latency effect of caching task results,an edge network computing offloading model is constructed.By jointly optimizing system resources,offloading decision,and caching decision,an optimization problem is designed to minimize the total latency of device tasks.To efficiently solve this optimization problem,a deep reinforcement learning-based collaborative edge computing offloading and caching algorithm(DRL-CECOC)is proposed.The algorithm combines the actor-critic algorithm with a deterministic policy gradient selection of the action network,the evaluation mechanism of the critic network,and the soft update of network parameters to achieve fast and stable convergence and obtain the optimal solution of the system.Finally,simulation results show that the proposed algorithm can quickly achieve stable convergence and obtain the best solution for total task processing latency compared to two benchmark methods.3)Device to Device(D2D)assisted task offloading and caching replacement for edge computing: Aiming at the problem of high task latency caused by the shortage of computing and bandwidth resources in edge networks,and the limitation of edge cache space,this thesis proposes a D2D-assisted edge computing task offloading and caching replacement mechanism that incorporates D2 D communication technology and caching replacement strategies.An optimization problem is formulated to minimize task latency while considering edge server computing and bandwidth resource allocation,offloading decision,and caching decision,in order to further accelerate the solution algorithm to obtain the optimal processing decision and effectively meet the demand of delay-sensitive devices,a priority experience sampling-based task offloading and caching replacement(PES-TOCR)algorithm is proposed to solve this mixed-integer nonlinear optimization problem using the idea of deep deterministic policy gradient algorithm.Unlike the original algorithm which randomly uniformly samples experience pool samples,this algorithm uses a new sample prioritization method,with the absolute value of the time and space error of the samples to increase the probability of selecting samples with larger changes in model network training,can accelerate the network training process to reach stable convergence faster,thereby quickly obtaining the effectiveness of model decision-making.Finally,simulation results show that compared with several benchmark algorithms,the proposed algorithm has significant advantages in network convergence,task latency,and cache hit rate.