Multi-objective resource allocation for Edge Cloud based robotic workflow in smart factory

摘要

Multi-robotic services are widely used to enhance the efficiency of Industry 4.0 applications including emergency management in smart factory. The workflow of these robotic services consists of data hungry, delay sensitive and compute intensive tasks. Generally, robots are not enriched in computational power and storage capabilities. It is thus beneficial to leverage the available Cloud resources to complement robots for executing robotic workflows. When multiple robots and Cloud instances work in a collaborative manner, optimal resource allocation for the tasks of a robotic workflow becomes a challenging problem. The diverse energy consumption rate of both robot and Cloud instances, and the cost of executing robotic workflow in such a distributed manner further intensify the resource allocation problem. Since the tasks are inter-dependent, inconvenience in data exchange between local robots and remote Cloud also degrade the service quality. Therefore, in this paper, we address simultaneous optimization of makespan, energy consumption and cost while allocating resources for the tasks of a robotic workflow. As a use case, we consider resource allocation for the robotic workflow of emergency management service in smart factory. We design an Edge Cloud based multi-robot system to overcome the limitations of remote Cloud based system in exchanging delay sensitive data. The resource allocation for robotic workflow is modelled as a constrained multi-objective optimization problem and it is solved through a multi-objective evolutionary approach, namely, NSGA-II algorithm. We have redesigned the NSGA-II algorithm by defining a new chromosome structure, pre-sorted initial population and mutation operator. It is further augmented by selecting the minimum distant solution from the non-dominated front to the origin while crossing over the chromosomes. The experimental results based on synthetic workload demonstrate that our augmented NSGA-II algorithm outperforms the state-of-the-art works by at least 18% in optimizing makespan, energy and cost attributes on various scenarios. (C) 2019 Elsevier B.V. All rights reserved.