Here Comes Cloud Robotics

Earlier this week ARC was invited to the offices of Singtel, the largest telecoms company in Singapore, to witness the first live demonstration of 5G cellular network technology in Southeast Asia. While 5G is still some four to five years away from commercial deployment, intensive development is well under way at companies like Ericsson, Singtel’s 5G technology partner and provider of the 5G Radio Prototypes used for the demonstration of what was revealed to be cellular network speeds of 27.3 Gbps and latency of 2 ms. In contrast, today’s 4G LTE networks typically offer speeds of around 10 Mbps and latency of 25 ms.

For a telecoms company like Singtel, which is faced with declining growth in smartphone subscriptions, at least in developed markets, 5G’s stellar network performance offers the opportunity to enter IoT related sectors such as remote surgery, immersive augmented reality, intelligent transportation systems, and cloud robotics, all of which it mentioned during the 5G demonstration. In particular, it is 5G’s very low latency ­– the time delay between signal transmission and receive on a network – that is key to satisfying real-time needs, and this is expected to get down to 1 ms or below by the time of commercial roll-out in 2020-21.

Taking a closer look at cloud robotics, where (the future) 5G enters the picture is that it enables the effective equipment-to-cloud-to-equipment communications required for time-critical robot applications. While the majority of robots are still fixed, such as those found on automotive assembly lines, there is growing demand for the type of mobile robots used in applications like warehouse logistics (see Amazon’s KIVA robots) for which a wireless connection is naturally a better fit.

But just what is cloud robotics? The term was actually coined in 2010 by James Kuffner, Robotics Director at Google and adjunct professor at The Robotics Institute, Carnegie Mellon University. (He is now CTO at Toyota Research Institute in California.) Essentially, it denotes separation of the thinking and doing parts of a robot, such that while the sensors and actuators remain with the equipment, the bulk of the computational processing and storage takes place remotely, in the cloud.

With this configuration, robots, especially those engaged in tasks that require heavy processing such as image recognition, can be made less complex on a unitized basis, consume less power and be produced at a lower cost. And the virtually limitless nature of the cloud in terms of processing power and storage means robot sophistication is not limited by local hardware constraints.

Here in Asia, a Chinese cloud robotics start-up by the name of CloudMinds recently announced $30 million of seed funding from a consortium including SoftBank and Hon Hai Precision Industry. The company, which has grown quickly from 20 to 100 people and now has offices in the US, Japan as well as China, believes that "only with the cloud can we build intelligent robots."

And SoftBank itself, through its SoftBank Robotics subsidiary, announced in March a strategic collaboration with Microsoft for cloud robotics. The companies will work to create a next-generation cloud-enabled robot using Pepper (SoftBank Robotics' humanoid robot) and Microsoft's cloud-based Azure IoT Suite.

As for traditional robot manufacturers in the industrial realm, while the wholesale devolution of processing power to the cloud is unlikely to happen just yet, there is a trend of using cloud and IIoT technologies to improve robot performance and capabilities.

Last October, Fanuc, the world’s largest robot manufacturer, announced the successful conclusion of a 12-month pilot project featuring the FANUC ZDT (Zero Downtime) system implemented at General Motors’ plants. With FANUC ZDT, robots are connected through a Cisco network and into a Cisco edge compute data collector in the plant. Data relevant for maintenance is sent to the Cisco Cloud where an analytics engine captures the exceptions and predicts the maintenance need.

Once that need is identified, an alert is sent from the cloud application to Fanuc service personnel and to the customer about the need for a replacement part, which can then be shipped in time for the next scheduled maintenance. It’s not hard to see how this proactive approach helps to avoid costly unexpected downtime in manufacturing.

A few months ago, in April, Fanuc revealed an enhancement to FANUC ZDT in the form of the FANUC Intelligent Edge Link and Drive (FIELD) system, which it is developing this time through a collaboration with Cisco, Rockwell Automation, and Preferred Networks (a Japanese company specializing in deep learning technology and which Fanuc has an equity stake in).

Among other aspects, FIELD brings advanced machine learning techniques to the robot world, enabling robots to learn new tasks far more quickly than with conventional programming, and even sharing what they have learnt with each other through so-called “distributed learning.”

Tellingly, in an echo of Jeff Immelt’s quotes on GE’s business transformation initiatives with Industrial IoT, Tetsuaki Kato, GM of Fanuc’s software development laboratory, said earlier this year: “Our future growth potential is in network and software.”

Another indication of traditional robot suppliers looking to information and communications technologies as a platform for growth came at this year’s CeBIT when Germany’s Kuka signed an MoU with China’s Huawei to develop smart manufacturing solutions for industrial markets. Under the agreement, Huawei and Kuka will collaborate in areas including cloud computing, big data, wireless and 5G, and robot programming using imitative deep learning. Notably, Huawei’s infrastructure-as-a-service (IaaS) solutions will be used to develop cloud-hosted smart manufacturing services.