Cloud robotics

Few doubt that in the near future, robotics will fundamentally change production systems and dramatically increase their level of automation. To do so, robots will have to figure out for themselves how to solve problems and adapt to dynamic environments. This can be achieved by exploiting the IoT to facilitate the creation of new technology involving large-scale data propagation, stream analytics and machine learning.

It is predicted that the use of robotics in manufacturing and automation will increase significantly in the near future and that this growth will drive a major expansion of the industrial robot market¹. These expectations are predicated on industrial robots finding their way into many more automation scenarios than is currently the case.

Today, industrial robots can tirelessly repeat complex tasks with high precision — for example, welding, painting, automobile production and certain types of assembly. However, there are many other manufacturing or assembly scenarios that would benefit from robotic automation but that are challenging to automate. This can be due to, for example, short production runs or environments that are not well enough controlled. In many of these cases, humans currently play an important role. If the use of robots is to be extended to these challenging scenarios, robots have to become more flexible, easier to program and more autonomous. Further, at the same time as robots need to more intelligently use information provided by humans and the environment, robots also need to channel information to humans in a more intelligent way. They can do this by analysing known information, extracting knowledge from it and making that knowledge easily accessible also to non-experts.

The importance of IoT and cloud technologies

With commercial Internet of Things (IoT) and cloud technologies, it is currently possible to transport large amounts of sensor data and other information from devices to data centres. Within the data centre, stream analytics can be used to process the device information in real time for filtering, selection and aggregation.

The processed information can be fed into different cloud services such as business intelligence (BI) tools that turn raw data into tables and graphs — giving instant insight into production situations. The information can also be used by machine-learning packages to make predictions — for process optimisation or predictive maintenance, for example. Many such highly scalable and cost-effective services that can analyse large quantities of data in data centres are already available.

It is, of course, imperative that such analysis is done safely, securely and with full data integrity. Also, reliability and availability levels must be maintained.

By increasing robot capabilities using IoT and cloud technologies, and by locating most storage, analysis and large-scale computation in data centres, future requirements for robot intelligence can most likely be met without any increase in the cost or physical size of controllers.

Motivation

The ways in which the IoT can help improve operational performance in robotic production scenarios can be illustrated by considering an example: In a small-part assembly cell, two robots are working collaboratively. Small parts come in on two separate feeders. The robots pick parts from their respective feeders, assemble them and put the assembly on a conveyor belt. An operator or a production manager can use a mobile device to monitor the production status and obtain information about the devices in the production cell at any time and from any location. Device predictive KPIs can also be checked so that maintenance decisions can be made.

In the case of a sudden disturbance — such as one feeder slowing down due to an assembly part supply problem — information is exchanged between the robots, feeders and conveyor belt, which all adapt their rate of operation to accommodate the new circumstances. The operator is notified of the situation via their mobile device. If operational performance is within a certain tolerance, they may decide not to interrupt the production process. Or, in the case of a faulty feeder, they may check the KPIs of the devices and find out that a service technician is shortly due to replace some parts on that feeder. This may mean the system can be run in its current state until the service occurs, meaning a possibly costly immediate production shutdown can be avoided.

Solution strategy

The scenario just described involves industrial robotics control, and networks of sensors and actuators that demand real-time and predictable temporal behaviour of the robot control system. A further requirement is a set of intelligent robot service features that can be deployed using IoT technologies to improve operational performance on the factory floor. One way to make this constellation of requirements a reality is to:

• enable data sharing among connected robots and other devices within a production cell;
• host real-time robot applications that require very low and predictable latency at the network edge or in the robot controllers;
• connect to a remote data centre for large-scale BI and data analytic capabilities.

Figure 1: Industrial IoT from edge to cloud. For a larger image click here.

In this way, additional cloud-based service solutions can be offered to customers, such as easy access to, and visualisation of, production data in the cloud. Moreover, by utilising cloud infrastructures that can provide elastic computation resources and storage, new intelligent robot services centralised on BI and data analytics can be developed (Figure 1). Examples of these are machine learning and advanced analysis of large datasets of robot information collected during operation life cycles.

End-to-end concept and technical solution

To realise the solution strategy just described, an example of a scalable collaboration platform that enables information sharing between connected industrial robots, other industrial devices in a production cell and people is shown in Figure 2. Such a platform, when it becomes a final product, will offer ease of use with respect to configuration, for example, discovery of robots, connecting robots for collaboration and service provision.

In the platform’s automation layer, real-time data exchange between robots is enabled through publish-subscribe middleware technology, such as the data distribution service (DDS) framework. One device publishes information on a topic and other interested devices can subscribe to receive it. Subscriber devices do not need to know where information comes from as context data is also provided to tell the subscriber devices what to do with the information.

Figure 2: System architecture for a scalable collaboration platform. For a larger image click here.

The devices exchange information through a virtual global data space. The robots and the feeder mentioned in the example above could, for instance, exchange information (current position, speed, etc) through this global data space.

Not all devices in a production cell may be suitable for participation in a publish/subscribe framework. This can be due to, for example, accessibility limitations of third-party devices or finite computing power. Such devices can, however, interact with robots and other devices through a lightweight RESTful interface, which is provided by a collaborative agent in the IoT layer. RESTful interfaces are based on REST (representational state transfer) — a web architecture that takes up less bandwidth than other equivalent architectures and that simplifies connection of diverse clients. The collaborative agent can be deployed on any device (including the robot controller) on which the published/subscribe framework can be installed. The RESTful interface is also employed by the different mobile devices that are used for production cell monitoring, as well as by a cloud agent. The cloud agent, deployed on a robot controller or some other device in the production cell, uses AMQP (advanced message queuing protocol) and HTTP as an interface to send data to or interact with the cloud layer.

The proposed cloud layer in the architecture enables increased service opportunities by connecting the devices in the production cell, or the production cell itself, to the cloud. A cloud platform is used such as Microsoft Azure IoT Suite², which offers a broad range of capabilities — for example, data collection from devices, stream analytics, machine learning, storage and data presentation. In particular, such a solution would consist of an IoT client, an event hub (which acts as an event ingester), stream analytics and a self-service BI solution. The cloud agent sends robot data to the event hub. The stream analytics service consumes that data and enables stream processing logic (in a simple SQL-like language) to be run. The results of this processing are sent to the BI application, which carries out the monitoring and visualisation of the production data.

In the engineering layer, two types of applications are distinguished: web-based simplified configuration applications and desktop applications for advanced configuration of the robots and the rest of production (such as ABB RobotStudio).

Grasping the future

Using IoT technology to connect things, services and people will change the everyday life of users and enable intelligent industrial operations. Imagine that the small parts in the example scenario described earlier have attached to them smart tags that allow, via wireless communication, the transmission of certain types of information — for example, CAD drawings, item description and handling instructions — to robots and operators. The dissemination of such information could, for example, allow the adjustment of robot grasp planning with the available grippers when there are changes in the types of small parts. At present, this is an offline and manual task.

The key idea of utilising the IoT in this way is to obtain information about devices and the environment, analyse data from the physical and virtual world for optimised operations, and provide enhanced services to users. By delivering new end-customer software services and experiences that are based on information extracted from multiple connected devices, the IoT is providing a new way of realising business agility and a faster pace of innovation.

References

1. “Industrial robotics market expected to reach $41 billion by 2020”, Modern Materials Handling, 28 October 2015, <http://www.mmh.com/article/industrial_robotics_market_expected_to_reach_41_billion_by_20202>
2. Microsoft Azure IoT Suite, <https://www.microsoft.com/en-us/server-cloud/internet-of-things/azure-iot-suite.aspx>

This article is based on an article previously published in ABB Review.