Abstract: In order to make use of limited computing resources to provide efficient computing services, a cloud-edge-end collaborative task offloading framework based on Docker is proposed In MEC (mobile edge computing) to solve the problems of multi-access MEC collaborative offloading and computing resource allocation. In order to improve the execution rate of tasks and the resource utilization of each node, preprocessing of tasks: Kahn algorithm added to the requirements of full rank matrix sets the execution sequence of output tasks in combination with the parallel calculation of tasks. Based on these, the task computing mathematical models of end, edge and cloud are established respectively, and the objective functions of the joint optimization system delay and energy consumption are designed by assigning weights. In order to solve the optimal offloading decision, the parameter of "full merit rate" and particle bee are introduced to propose APS (artificial particle swarm) algorithm as offloading decision algorithm. The experiments show that multi-task processing proves the effectiveness of APS algorithm. Compared with the five modes of local computing, edge computing, cloud computing, end-edge union computing and random processing, the low latency and low energy consumption of the proposed scheme proves its advantages in providing efficient services under multi-access conditions.