A Cloud-edge Collaborative Framework for Computing Tasks Based on Load Forecasting and Resource Adaptive Allocation

Abstract

In recent years, cloud-edge collaboration has attracted extensive attention and research. Robot target tracking is a common application scenario for collaborative computing, which usually involves visual location, target analysis, route planning, attitude control and other task types with different computational complexities and real-time requirements. Given different computing capabilities, service loads, communication overheads, area crossing of different nodes at different time, the problem of how to allocate various services of the robot among the local, edge servers and cloud servers effectively according to the optimization objectives involves complex cloud edge cooperation strategy in the process of robot target tracking. To solve this problem, this paper proposes a cloud-edge collaborative framework for computing tasks based on load forecasting and resource adaptive allocation, which can efficiently allocate tasks of different types and complexities according to the current relative position of the robot with the servers, the computing resources and current loads of edge servers and cloud server. A large number of experiments show that this method can effectively shortens the average response time of various services and improves the service quality compared with traditional methods.