CCRP: A Novel Clone-Based Cloud Robotic Platform for Multi-Robots

Recently, the cloud computing paradigm has evolved from various research fields. A new path of research, cloud robotics, has emerged which allows robots to inherit the enormous computing and storage capability of cloud. Advances in cloud computing technologies, networking, parallel computing and other evolving technologies, and the integration with multi-robot systems, make it possible to design systems with new capabilities. The main advantages of cloud robotics are in overcoming the limitations of on-board robot computing and storage capabilities and in improving energy efficiency. Nevertheless, there is a lack of cloud robotics frameworks that can provide a secured environment for multi-robot application. The implementation of a robust cloud robotic platform capable of handling multi-robot applications has been shown to be challenging.

This research proposes a novel Clone-based Cloud Robotic Platform architecture (CCRP) which assigns a Virtual Machine (VM) clone of each individual robot's operating system in the cloud, enabling fast and efficient collaboration between them via the cloud's inner-network. The system utilises Robot Operating System (ROS) as a middleware and programmable environment for robot development. This model is using the OpenVPN as a communication protocol between the robot and the VM, which provides considerable enhancement for the security and additional network for the system to allow multi-master ROS deployment. The Quality of Service (QoS) for the system has been tested and evaluated in terms of performance, compatibility and scalability via comparison study, which examines the CCRP performance against a local system and a proxy-based cloud system.

Two case studies have been deployed for different robot scenarios. Case study 1 was focused on a navigation task which includes the process of mapping and teleoperation implemented in Google public cloud. The real time response has been examined by using the CCRP to teleoperate the NAO and Turtlebot robots. A response time and video streaming delays were measured to assess the overall QoS performance. Case study 2 is composed of a face recognition task performed using the CCRP in a private cloud on an Openstack platform. The objective of this task was to evaluate the system ability of running the tasks in the cloud effectively and to assess the collaborative learning capability. During the CCRP development and deployment stages an optimization study was conducted to determine optimal parameters for data offloading to the cloud and energy efficiency of a low-cost robot.

The result of the CCRP performance evaluation proved that it is capable of running on a public and private cloud platform for self-configuring and programmable robotic systems, as well as executing various applications on different robot types. The CCRP is facilitating the improvements to QoS performance, compatibility and scalability and is providing a secure cloud computing environment for on-board robots.