쿠버네티스 클러스터 기반 다수 로봇 원격 제어 및 관리

Abstract

This paper proposes a method to remotely control and manage a Kubernetes-based multi-robot swarm efficiently while ensuring high availability of the robots’ internal controller modules. We evaluate the effectiveness of the proposed approach by experimenting with a self-balancing robot and robot arm. Each robot is configured as a worker node in a Kubernetes cluster. A master node is configured to remotely monitor their statuses and manage the cluster by running, stopping, and updating the controller module. We also develop a labeling module to automate the process of imprinting labels on all the robots in the cluster according to their hardware, software, and functions. A control module required for each robot is automatically deployed remotely, and the appropriate module for robot is selected when the control module is deployed. Furthermore, we experimentally evaluate a control feature designed to ensure high resource availability and fault tolerance. The Kubernetes system is configured to detect problems with the controllers of the self-balancing robots and to create new pods to solve these problems automatically. Our results demonstrate that multi-robots can be controlled simultaneously using an existing cloud-computing container-orchestration system.