SANTA CLARA, Calif. — For service robots to become truly useful, they need artificial intelligence running on the cloud and accessible over 5G networks, according to CloudMinds Technology Inc. The company has taken a cloud-first approach to robotics and AI development.
At Mobile World Congress (MWC) in Los Angeles this week, CloudMinds is displaying its XR-1 humanoid service robot, which it said is the first in the U.S. to use cloud-based AI and 5G connectivity, It first showed the XR-1 at MWC Barcelona this past spring.
XR-1 includes a proprietary Smart Compliant Actuator and proprietary “smart joints” for precise movement and manipulation.
“With vision-controlled grasping and the ability to perform intricate tasks, the XR-1 simply raises the bar and lays the foundation for an even wider range of intelligent compliant cloud service robots from CloudMinds — from wheeled to two-legged form factors,” stated Bill Huang, co-founder and CEO of CloudMinds.
“We weren’t trained roboticists, but we wanted to leverage our backgrounds,” said Robert Zhang, co-founder and president of CloudMinds. “We came up with the idea of taking a telecommunications approach to robotics, in which robots all over the world would communicate with a scalable, high-performance network infrastructure.”
“We want to help every family eventually get a humanoid robot nanny, which everyone has been talking about for years,” he told The Robot Report. “We don’t intend to build all the hardware or software ourselves, but in the next two to three years, we are building end-to-end solutions including robots, connectivity, and the cloud for verticals including retail, hospitality, and residential and office campuses. Our long-term plan is to encourage application developers to come to our platform with AI and network optimization to build robots of different functionalities.”
AI needs humans to learn
CloudMinds’ Human Augmented Robotics Intelligence (HARIX) platform is the “cloud brain” for its cleaning, security, and concierge robots, but it still “keeps humans in the loop,” said Zhang.
“AI is still lagging human intelligence in a big way, but we don’t want to hinder performance for customers,” he said. “We work very closely with state-of-the-art machine learning, deep learning, and imitation learning, but our unique contribution is gathering data in the field to feed the algorithms. For something that the robot can’t handle in the beginning, it can become smarter with supervised learning in HARIX.”
“A few days or weeks later, the robot can handle more tasks and gets smarter as it gathers more data,” Zhang said. “A combination of human annotation and semi-supervised learning with some automation of these processes is the trend.”
“We’re proud to say that we created the phrase ‘cloud robotics,’ and we’re the first company to turn the concept into a product and a service, but an increasing number of companies are getting into this space, like Amazon, Google, and Microsoft,” he acknowledged. “But because we started with the cloud and branched into verticals, our network infrastructure is set up to be more scalable. We could even be partners with the big guys.”
Designing for security
CloudMinds is aware of the security risks of connected devices. “Our approach is end to end,” Zhang said. “We put a lot of effort designing security technology into the robot, the network layer, and the cloud.”
“On the robot side, when we work with a third-party robot maker, we provide a robot control unit that connects to our network and the cloud,” he added. “We spent a lot of time to ensure that it is very secure. We designed the operating system to be divided into a public and an enterprise/private domain.”
CloudMinds’ Nerve Network backbone also incorporates multiple security features. “The network must be hacker-proof, and we’ve combined Blockchain with a software-defined network and perimeter,” said Zhang. “It’s like an invisible network layer — people can’t hack what they can’t see; it’s like not knowing the IP address on a network.”
“On the cloud side, we use virtual private clouds,” he said. “It’s like a hotel with many rooms — each person has access only to specific rooms.”
“While our robots rely heavily on cloud-compute resources, that doesn’t mean they’re dumb,” Zhang explained. “Certain tasks that are time-sensitive and mission-critical are performed locally.”
Among the challenges of managing robotic fleets is making sure that useful data is collected but not sharing it among multiple parties.
“We support multi-tenancy, but the infrastructure we set up does not automatically allow a view into that data,” Zhang said. “We anonymize data. We’re striking a balance between getting data from the robots but not tying it to a specific customer so that we can still use it to improve our algorithms’ performance.”
Thousands of service robots on order
In August, CloudMinds announced that it will provide thousands of service robots to two companies in China. Jin Yu Ao Environmental Technology will use up to 10,000 Cloud Cleaning Robots connected via 5G to clean schools, exhibition halls, large venues, and high-end offices and residences in the Beijing area, according to CloudMinds.
Zhongtai Min’an Security Service Group has ordered 3,700 robots for tasks such as security, cleaning, providing information, and reception at its properties.
How does this compare with SoftBank Robotics’ Whiz cleaning robot? SoftBank is a backer of CloudMinds, and its technology is complementary, Zhang replied.
“Cleaning is relatively new to SoftBank’s portfolio, and it is mostly in Japan,” he said. “Our cleaning robot is actually from China, where the requirements for robots and services are different.”
“With our customers’ agreement, we’re working on the delivery of these robots in batches,” noted Zhang. “In the meantime, our technology will enhance itself. The idea is that in the next two years, the cleaning and patrol robots will be ready to launch in the mass market beyond China.”
CloudMinds ready to scale up
How will CloudMinds handle scaling up management of the robots? “This is where HARIX and the robot operating center [ROC] come in,” Zhang said. “Provisioning combines humans and automated methods. The ROC is like a telecommunications support center.”
“In the beginning, we’ll have one staff member watching two or three robots, but going forward, we’re expect to have one person managing hundreds of robots,” he said. “Think of the massive scale of smartphone deployments; that infrastructure is similar to what we’re building for cloud robotics.”
CloudMinds filed for a $500 million initial public offering in June. “We’re at a point where we need to educate the market and align expectations,” Zhang said.
Earlier this month, CloudMinds opened new offices in Santiago, Chile, to expand its customer support for the Southern Hemisphere. The company has brought Cloud Pepper, based on SoftBank Robotics’ Pepper, to events about climate change throughout Latin America.
The Robot Report is launching the Healthcare Robotics Engineering Forum, which will be on Dec. 9-10 in Santa Clara, Calif. The conference and expo will focus on improving the design, development, and manufacture of next-generation healthcare robots. Learn more about the Healthcare Robotics Engineering Forum, and registration is now open.
CloudMinds ready for 5G
CloudMinds has touted its use of 5G in China, but it also noted that it works with existing wireless networks.
“Reliability is very important, but in the evening, everyone’s streaming, making remote control a challenge,” said Zhang. “We look at 5G as a better network infrastructure to solve latency, bandwidth, and quality of service.”
“We’re taking a phased approach — the products we’ve been working on so far work pretty well in 4G environments,” he said. “We made a conscious decision not to work on drones or self-driving vehicles because these robots are all fast-moving — a split-second miss would result in catastrophic failures. When 5G is more prevalent, our infrastructure will naturally migrate to it.”
XR-1 is serving drinks in the Sprint 5G exhibit, South Hall 1702, at MWC Los Angeles. Sprint is also launching its Curiosity IoT dedicated, virtualized, and distributed Internet of Things network and operating system.