Research On Crowdsourcing Task Allocation Mechanism In Edge Cloud Environment

In recent years,with the continuous growth of the number of mobile intelligent terminals and the emergence of various data collection methods under the edge cloud,the crowd collaboration sensing has been further developed.The goal of crowd collaboration is to achieve a high-quality goal through the mutual-cooperation among workers,tasks and platforms,task allocation has become an important work in crowd collaboration.Due to the resource limitation of the hardware server and the continuous increase of the transmission amount in the sensing process,the traditional cloud environment can no longer meet the current computing demand.The emergence of edge computing and reinforcement learning effectively alleviates such problems.At present,most of the existing research on task allocation focuses only on dynamic and static environment.Dynamic scenarios cannot meet the requirements of summarizing global information,and static scenarios cannot match the dynamic requirements of real situations.This paper focuses on the hybrid design of two scenarios.Mobile edge computing is a promising computing paradigm that migrates computation-intensive tasks from resource-constrained mobile smart devices to nearby edge service servers for lower latency and energy consumption.However,it is challenging to coordinate computing task offloading among multiple users,considering channel conditions and different latency requirements of various computing tasks.In order to solve the above problems,we use multiple edge nodes as task offload servers to maximize the use of service resources to complete the calculation and reduce the delay.The main contents of this paper are as follows:(1)To solve the task allocation problem in hybrid scenarios,a latency-based online and offline task allocation(LTB-TAOO)framework is proposed.The framework dynamically receives employees and tasks from the perspective of the entire assignment process.A static allocation method is used throughout the task allocation algorithm.To solve the problem of latency computation in LTB-TAOO framework,a latency task algorithm based on Qlearning(QL-LTC)is proposed.In order to improve the efficiency and accuracy of LTBTAOO framework,a new LTC task allocation algorithm based on state transition Q-learning(NSQL-LTC)is proposed by optimizing the state transition matrix and reward mechanism.(2)A multi-user mobile edge computing(MEC)system is considered and an autonomous system task offloading system is proposed to solve the resource-constrained task allocation problem in edge environment.Inspired by the recent development of reinforcement learning,two unloading strategies based on RL are proposed to automatically optimize the delay performance.Specifically,a Q-learning algorithm is first performed to provide partial unloading decisions.Then,a more flexible task offloading strategy is adopted to further optimize the system performance,and a continuous task offloading decision is provided based on the deep deterministic policy gradient.