GB/T 45301-2025The classification and reference architecture of cloud platform for service robot (English PDF)

1 Scope

This document gives the service content, the classification method and the reference architecture (infrastructure layer, platform layer, application layer, security management, operation and maintenance management, and evaluation and assessment) of cloud platforms for service robots.

This document applies to the building and use of service robot cloud platforms by service robot cloud platform developers and by service robot application developers, manufacturers, integrators and users.

2 Normative references

The contents of the following documents constitute indispensable provisions of this document through the normative references made to them in the text. For dated references, only the edition corresponding to that date applies to this document; for undated references, the latest edition (including all amendments) applies to this document.

GB/T 12643-2013 Robots and robotic devices - Vocabulary

GB/T 32400-2015 Information technology - Cloud computing - Overview and vocabulary

GB/T 38834.1-2020 Robotics - Performance criteria and related test methods for service robots - Part 1: Locomotion for wheeled robots

3 Terms and definitions

The terms and definitions given in GB/T 12643-2013, GB/T 32400-2015 and GB/T 38834.1-2020 and the following apply to this document.

3.1 service robot: A robot that performs useful tasks for humans or equipment, excluding industrial automation applications. [SOURCE: GB/T 12643-2013, 2.10]

3.2 cloud platform: A collection of software and hardware that provides one or more capabilities through the defined interfaces of cloud computing.

3.3 architecture: The organisational structure of a system or a component. [SOURCE: GB/T 11457-2006, 2.73]

3.4 cloud computing: A paradigm in which a scalable and elastic pool of shared physical and virtual resources is provided and managed in an on-demand self-service manner. Note: the resources include servers, operating systems, networks, software, applications and storage devices. [SOURCE: GB/T 32400-2015, 3.2.5]

4 Abbreviations

The following abbreviations apply to this document.

API: Application Programming Interface.

IaaS: Infrastructure as a Service.

PaaS: Platform as a Service.

SaaS: Software as a Service.

5 Services provided by a service robot cloud platform

A service robot cloud platform mainly provides the following services.

a) Computing and storage services: they provide the computing and storage resources a service robot needs so that it can carry out various tasks on the cloud platform, such as speech recognition, natural language processing and data analysis.

b) Security services: they provide the security measures that protect the data of the service robot at every stage of the whole data life cycle, such as data encryption and identity authentication.

c) Interface services: they provide open and extensible interfaces, including secondary development of software interfaces and data transmission protocols.

d) Collaboration services: they rapidly assemble the various system services according to what the user needs and provide a collaborative solution.

e) Sharing services: they allow outside parties, once they have obtained permission, to call the services of the platform.

f) Other services: services other than those listed above are supported.

6.1 Classification by the users the cloud platform is aimed at

6.1.1 Cloud platform aimed at service robot developers. It is aimed at the people who develop service robots and provides the various tools and general service components that service robot development calls for, including service robot software development tools, software development kits and APIs, a virtual simulation environment for service robots, and cloud training services for service robot models.

6.1.2 Cloud platform aimed at business users of service robots. It provides users with platform services for the deployment, management, operation and maintenance of service robots, and supports continuous integration, security, deployment of computing resources and management of cluster service pools, so that users can quickly build a service robot system, manage the computing resources and the data of the service robots, and make the service robots intelligent, for example by achieving visualisation and automated data management through data analysis.

6.2 Classification by the way the cloud is connected

6.2.1 Cloud-device centralised cloud platform. Each service robot sets up a communication link directly with the cloud platform, transmits its operating data and calls the services on the cloud platform, as shown in Figure 1.

6.2.2 Cloud-edge-device distributed cloud platform. Each service robot sets up a communication link with an edge server; the edge server is responsible for the preliminary filtering, analysis and storage of local data and then uploads the data to the cloud, and the cloud platform is responsible for the analysis and storage of the data, as shown in Figure 2.

Users, cloud service platform developers and robot developers choose the deployment method of the cloud service according to factors such as the scale of the cloud service, its real-time behaviour, its security and the way the services are called. In a cloud-edge-device distributed cloud platform, the computing and storage services are generally placed both in the cloud and at the edge, the security services are placed at the edge, and the interface services, collaboration services and sharing services are placed in the cloud.

7.1 Overview

The reference architecture of a service robot cloud platform includes the infrastructure layer, the platform layer, the application layer, security management, operation and maintenance management, and evaluation and assessment, as shown in Figure 3. The architecture defines a complete device abstraction capability, a general service extension capability and a flexible capability for aggregating solutions. The different classes of cloud service platform described in Clause 6 can be built on this architecture, forming different user interfaces by the way the function list is organised, and the function list can be adjusted as use requires.

The infrastructure layer is mainly the layer that provides the basic physical resources for the service robot, and it includes computing infrastructure, storage infrastructure and network infrastructure. The provider of the cloud computing infrastructure integrates large numbers of servers, network devices and storage devices into available basic resources by means of distributed technology, virtualisation technology and the like, and abstracts them outwards as standardised cloud infrastructure. The user of cloud computing only has to apply for the kind of resource it needs and can then obtain the required resource allocation from the infrastructure service in the cloud. A service robot, or an application, can obtain these resources on demand while running, and the cloud platform can deliver the lowest-level devices and resources to the user over the internet.

The platform layer includes modules for service robot and device management, data communication, resource management, data storage, perception, cognition, decision, control and collaborative computing. It performs the modelling and access management of service robots and devices and the unified management and scheduling of computing resources, provides the service robots connected to the platform with intelligent algorithm services for perception, cognition and autonomous decision, and controls the service robot on the basis of the decision results. Relying on the collaborative computing module, the platform layer schedules cloud, edge and device heterogeneous computing units together, and deploys and runs each computing task in a distributed way according to the current resource load of the system and the demand for computing resources, producing a reliable output.

The application layer abstracts and encapsulates the resources of the two layers below into application services or complete solutions, and provides pre-integrated products and combinations of capabilities for particular industries, meeting the needs of different scenarios for moving to the cloud.

Security management of the cloud platform includes infrastructure security, IaaS security, PaaS security, SaaS security, cloud security management, cloud security supervision, and the process security between the service robot and the cloud platform, so that the secure operation of the platform is managed and monitored.

Operation and maintenance management of the cloud platform provides monitoring of the operation and maintenance process of the platform and safeguards the quality of service.

Evaluation and assessment of the cloud platform provides evaluation services covering the function, performance, security and reliability of the platform.

The emphasis of the reference architecture differs between types of cloud platform. A cloud platform aimed at service robot developers should provide at least access rights to the platform layer and the related interface services; a cloud platform aimed at business users of service robots should include at least access rights to the application layer and the related collaboration services and sharing services.

7.2 Infrastructure layer

7.2.1 Computing infrastructure. The computing infrastructure includes servers, processors and the like. All available resources are virtualised in a unified way as virtual resources in a resource pool, and the cloud service provider can supply computing capability services according to what the user needs.

7.2.2 Storage infrastructure. The storage infrastructure includes hard disks, solid state disks, storage area networks and the like. It is the collection of storage devices, storage systems and storage services, and is used to store and manage data.

7.2.3 Network infrastructure. The network infrastructure includes network devices, routers, switches and the like. Network resources are integrated into a resource pool by virtualisation technology, and network connection and transmission services are provided, so that data are transmitted quickly and communication is efficient.

7.3 Platform layer

7.3.1 Service robot and device management. The cloud platform can manage the types, instances and commands of service robots and devices, and comprises the following four parts. a) Type management: it manages the type information of service robots and devices, including the definition of service robot types and device types; new service robot types and device types can be defined and their attributes and functions configured. b) Device instance management: it manages the instance information of registered devices; the information of registered device instances can be viewed and instances can be added, modified and deleted. c) Service robot instance management: it manages the state and control of service robot instances, including the online state, the running state and the task scheduling of the service robot; the state information of a service robot can be viewed and control and task scheduling carried out. d) Command management: it manages the commands of service robots and devices, including the definition, issue and execution of commands; new commands can be defined and issued to service robots and devices for execution.

7.3.2 Data communication. The cloud platform can carry out real-time data transmission and command exchange with service robots and devices, so that they are managed and controlled, and comprises the following three parts. a) Data relay: the data uploaded by service robots and devices are received, stored and forwarded. For example, the data a service robot obtains through its sensors are uploaded to the platform layer, which receives and stores them and can at the same time forward them to other systems or applications that need them. b) Command issue: commands are issued to service robots and devices to make them carry out particular tasks or operations, for example moving a service robot to a given position or making it perform a given action. c) Message push: messages are pushed to the display end of service robots and devices for notification and reminding.

7.3.3 Resource management. The cloud platform can monitor and manage the resources of service robots and devices so that performance is optimised and resources allocated, and comprises the following four parts. a) Resource monitoring: the use of the various resources of service robots and devices is monitored, including the central processing unit, memory and storage; the resource utilisation, performance and other indicators of service robots and devices can be viewed in real time so that performance can be optimised and resources allocated. b) Resource alarm: resource alarm rules are set, and when the resource use of a service robot or a device exceeds the set threshold an alarm notification is triggered, so that notice of an abnormal resource condition is received in time and measures can be taken. c) Hardware resource management: the hardware resources of service robots and devices are managed, including sensors, actuators and cameras; registered hardware resource information can be viewed and resources added, modified and deleted. d) Information resource management: the information resources of service robots and devices are managed, including map data, task data and log data.

7.3.4 Data storage. The cloud platform can store, access and query the data of service robots and devices, providing data support for the service robot, and comprises the following four parts. a) Relational data storage: it stores the relational data of service robots and devices, for example user information, device configuration and task records. b) Cache data storage: it stores the temporary data of service robots and devices, for example sensor data and calculation results. c) Distributed data storage: it stores large-scale data of service robots and devices, for example log data and image data. d) Indexed data storage: it provides fast data retrieval and query.

7.3.5 Perception. The cloud platform can provide the service robots connected to it with several intelligent perception algorithm services, supporting the cognition, decision and control of the service robot, and comprises the following four parts. a) Visual perception and algorithms: visual information such as images and video streams is obtained through vision sensors; with visual perception the service robot can carry out visual tasks such as object recognition, face recognition and pose estimation. b) Auditory perception and algorithms: auditory information such as sound and speech is obtained through hearing sensors; with auditory perception the service robot can carry out auditory tasks such as speech recognition, speech synthesis and sound source localisation. c) Tactile perception and algorithms: tactile information such as contact force and temperature is obtained through tactile sensors; with tactile perception the service robot can carry out tactile tasks such as grasping objects and force-controlled operation. d) Fused perception and algorithms: the data of the different perception modules are fused to provide more accurate and more complete perception information; with fused perception the service robot can carry out complex perception tasks such as scene understanding and environment modelling.

7.3.6 Cognition. The cloud platform can provide the service robot with intelligent and individualised cognitive functions, and comprises the following four parts. a) Knowledge base building: a knowledge base is built to store and organise the knowledge and information the service robot needs. The knowledge base includes facts, rules and concepts and the relations between them. b) Knowledge base query: the service robot can obtain the relevant knowledge by querying the knowledge base. c) Intent understanding: natural language processing, machine learning and similar techniques are used to understand the intent of the user input. By analysing the way the user expresses himself the platform can recognise the user's intent, for example a question, a need or a command. d) Knowledge reasoning: inference engines, logical reasoning and similar techniques are used to reason and draw inferences from the knowledge in the knowledge base. By reasoning over the logical relations and rules of the knowledge, the platform can generate new knowledge or infer an answer.

7.3.7 Decision. The cloud platform can give the service robot an intelligent decision capability. It can plan tasks and adjust strategies automatically according to the requirements of the task and the conditions of the environment, so that tasks are carried out efficiently and flexibly. It can also recognise and deal with abnormal situations, keeping the robot safe and stable, and comprises the following two parts. a) Task planning: planning algorithms and optimisation techniques are used to plan and arrange the tasks of the service robot. Task planning includes task decomposition, path planning and resource allocation, so that the robot can complete its tasks efficiently. b) Abnormal decision: abnormal situations are recognised and dealt with by monitoring and analysing the running state of the service robot. When the service robot meets an unexpected situation or cannot complete a task, the platform can take the corresponding decision, for example replanning the task or adjusting the execution strategy, so that the service robot keeps running stably.

7.3.8 Control. The cloud platform can control the service robot flexibly and efficiently, raising the level of intelligence and automation of the service robot, and comprises the following four parts. a) Service robot control: through the communication interface with the service robot the cloud platform controls the service robot remotely. For example, the platform can send commands to the service robot to control its movement, make it carry out a particular task, or make it interact with the environment. b) Service robot state monitoring: the cloud platform can monitor the state of the service robot in real time, including its position, its sensor data and its battery charge. By monitoring the state of the service robot the running situation is understood and the corresponding control decisions are made. c) Service robot task and priority management: the cloud platform manages the task queue of the service robot and schedules and allocates tasks according to their priority. The platform can arrange the order in which the service robot carries out its tasks according to how urgent they are, the resources they need and other factors. d) Handling of abnormal behaviour of the service robot: the cloud platform can detect and deal with abnormal behaviour of the service robot. When the robot breaks down, meets an obstacle or fails to complete a task, the corresponding measures can be taken, for example replanning the task, sending an alarm or notifying maintenance staff.

7.3.9 Collaborative computing. Through collaborative computing the cloud platform can schedule cloud, edge and device heterogeneous computing units together, and deploys and runs each computing task in a distributed way according to the current resource load of the system and the demand for computing resources, producing a reliable output.

7.4 Application layer

The application layer abstracts and encapsulates the resources of the infrastructure layer and the platform layer into application services or complete solutions, and offers them outwards in the form of application services or software interfaces, for example by providing a visual interface or an API to serve the service robot. It is the topmost layer of the cloud platform. What the user reaches is the application layer, and through it the service components of the platform layer and the data stored in the infrastructure layer can be reached. The cloud resource provider can deploy and maintain the relevant software on the service robot cloud platform, and users can reach the application services deployed on the cloud platform through either a browser/server or a client/server architecture.

The application layer is an important part of the whole system. It is responsible for the interaction between the service robot application services on the cloud platform and the service robots, and provides the management window for user services. It mainly provides the following five functions.

a) Service robot interfaces: once the resources related to the service robot have been turned into services, the application layer provides abstract interfaces to business applications and offers services of different kinds. There are two main classes of user of a service robot platform: one is the service robot producer, who analyses data through the service robot cloud platform as a reference for later development and who can also be offered remote control and maintenance of the service robots; the other is the service robot user, who uses the platform services offered by connecting to the service platform.

b) Visual interaction: the application layer should provide visual interaction that is friendly and easy to operate, including the simulation interface of the service robot and the various visual interaction and control interfaces that service robot developers and users build according to their own needs.

c) Remote control and parameter configuration: the application layer should support the control and parameter configuration of the service robot by remote means such as remote control software and remote configuration tools.

d) Collaborative control: the application layer should allow several users to carry out service design, operation management and application deployment on the same service robot or the same batch of service robots.

e) Customisation: the application layer should allow users of the cloud platform to abstract and encapsulate the resources of the platform layer and the infrastructure layer into application services or complete solutions themselves, according to their particular needs.

7.5 Security management

7.5.1 Infrastructure security. Infrastructure security covers not only the traditional protection of hardware such as servers, switches and storage, but also the techniques that protect the virtualisation monitor, including tamper protection of components, isolation of virtual machines, memory isolation and protection of hypervisor metadata.

7.5.2 IaaS security. IaaS security mainly covers the following three aspects. a) Virtual machine security: it includes port control inside the virtual machine, security of peripherals, virus protection, vulnerability scanning and intrusion detection. b) Virtual network security: it includes east-west network protection on the cloud platform, and covers traffic auditing, access control, intrusion detection and prevention, protection against address resolution protocol attacks, and bandwidth and traffic management. c) Virtual storage security: it includes storage encryption, integrity protection, migration encryption and access control for virtual machine disks, images and snapshots.

7.5.3 PaaS security. PaaS security mainly covers the following three aspects. a) Service security: it includes access control of services, identity authentication, web application protection, compliant configuration and operation auditing. b) Interface security: it includes access authentication and authorisation, encryption of interface communication, integrity of transmitted data, and protection against distributed denial of service attacks. c) Data security: it includes access control of data, authentication and authorisation, data classification and data masking.

7.5.4 SaaS security. SaaS security mainly covers protection of application security, application identity authentication, application access control, network application protection, application data encryption and auditing of application behaviour.

7.5.5 Cloud security management. Cloud security management covers the cloud security posture, the orchestration of cloud security services, the elastic scaling of security services and the unified management of security resources. The main difference between security protection in the cloud and traditional security protection lies in the disappearance of the protection boundary and in the dynamic change of the resources in the cloud. The service chain orchestration of cloud security services includes resource orchestration and service life cycle management; the elastic scaling of cloud security services includes high availability of services, horizontal scaling of services and load balancing of services. Cloud security management also covers a unified cloud security posture, cloud security operations and log auditing.

7.5.6 Cloud security supervision. Cloud security supervision applies security supervision to the cloud platform, the cloud services, the cloud applications and the cloud security services, and covers cloud security auditing, checking of cloud security policies, cloud security compliance checking and cloud security state checking. For cloud security management to achieve an adaptive cloud security architecture, the cloud security policy has to be adaptive, which includes obtaining the security policies of the cloud security services in a unified way, adjusting the security policy adaptively and checking policy compliance in a unified way. Cloud security state checking includes carrying out security checks on cloud components, cloud networks and cloud assets, detecting whether the security state meets the security requirements and putting forward proposals for improvement. Cloud security compliance checking is the checking of whether the cloud platform and the cloud services conform to the relevant security standards and specifications, analysing their security state and carrying out a security baseline check through a self-defined compliance policy.

7.5.7 Process security between the service robot and the cloud service platform. It mainly covers the following six aspects. a) Security of the communication protocol between the service robot and the cloud service platform, including encryption and compression of the data at the service robot terminal, authorisation of the terminal and identity authentication of the cloud service platform, and optimisation of the communication data when packets are lost on the terminal network. b) Setting up a dedicated communication tunnel between the cloud service platform and the service robot terminal. c) Security of the control rights of the service robot terminal. d) Security against data spoofing and deception at the service robot terminal. e) Monitoring of the state of the service robot as it moves in physical space (physical position, path of movement, collision envelope). f) Security during the interaction between the service robot and the user, including the security and privacy protection of speech recognition and face recognition and the judgement of whether touch interaction is appropriate.

7.6 Operation and maintenance management

Operation and maintenance management of a service robot cloud platform covers the following four aspects.

a) System monitoring and fault finding: the running state of the cloud platform is monitored regularly, including the load on the servers, the state of the network connections and the response time of the services.

b) The data on the cloud platform are backed up regularly, so that they are safe and recoverable. At the same time a data recovery mechanism is established to deal with unexpected data loss or system failure.

c) Resource management and optimisation: the resources of the cloud platform are managed and optimised, including the configuration and expansion of servers, the management of storage resources and the optimisation of network bandwidth, so that the cloud platform has enough resources behind it and the performance and stability of the system are improved.

d) Performance monitoring and optimisation: the performance indicators of the cloud platform are monitored, including response time, throughput and concurrency. Performance is optimised on the basis of the monitoring results, improving the performance of the cloud platform and the user experience.

7.7 Evaluation and assessment

Through the evaluation and assessment function the service robot cloud platform can evaluate and analyse the behaviour and the performance of the robot objectively, so as to provide a better service experience and to guide the improvement of the robot. It covers the following two aspects.

a) Evaluation of the execution results and the interaction of the service robot. For example, for a task carried out by a service robot the cloud platform can evaluate the execution according to how far and how well the task was completed. The evaluation results can be used to monitor the behaviour of the service robot and to give the user feedback and proposals for improvement.

b) Assessment of the performance of the service robot. For example, for the navigation capability of a service robot the cloud platform can assess the accuracy and the efficiency of navigation in different environments. The assessment results can be used to understand the performance of the service robot and as a reference for improving and optimising the robot.