With Daystar and DogBot Lenovo is building next generation of high technology robots

Lenovo unveiled its first in-house developed industrial robot, the Lenovo Daystar Robot at Lenovo Tech World in November 2020. Powered by Lenovo’s industry-leading intelligent edge computing technologies, and mixed reality and computer vision technologies, the Daystar is a spray-painting robot, for industrial use, working autonomously or with a remote human operator.

The Daystar Robot uses 3D scanning to model its workspace, and simultaneous localisation and mapping technology to navigate it. It provides a 3D Video Stream that an operator can view through an AR headset. The worker can then control the robot, as if they were in the workroom spray painting in person.

The robot can remember each paint job, so in future, if the same work is required again, the robot can carry out the task autonomously. As a final step, the robot will perform painting quality inspection automatically by leveraging computer vision technology.

The robot is intended to replace human workers in production lines, restricted workspace, dangerous, radioactive and other toxic environments, thus greatly reducing the risk posed to human operators and improving operational efficiency.

Typical application scenarios include assembly lines, painting workshops, power line patrol and inspection scenarios, hot chambers for radioactive tests or operations at nuclear power plants, handling and treatment of nuclear waste, high-temperature and high-pressure working environments etc.

Lenovo also has a range of robot vacuum cleaners, meant for consumer use in the home. The robot cleaners can be programmed via a dedicated app, so it can be set to clean different parts of the house on different days, and it can even empty its own waste bag. One of the main benefits, as shown by Daystar, is the ability to operate in restricted or dangerous workspaces.

Lenovo is a partner for UK-based AI researcher React Robotics, which is developing the DogBot, an autonomous robot that can execute tasks in high-risk industrial environments. DogBot, is a quadruped robotic helper which is being tested in construction environments, among others, where its advanced sensor allows it to carry out tasks such as build progress verification, managing site logistics and real-time sensor-data acquisition.

DogBot can support humans in high level tasks like inspections, but it can also carry out basic tasks, and even clean-up worksites, meaning human workers can focus on more complex jobs.

Lenovo has packaged computer vision and robotics technologies into its Machine Vision Solution for Smart Manufacturing to inspect assembly line defects. It has developed Esense, an automatic optical inspection solution using computer vision technologies.

Used in its China manufacturing plants, it makes it possible to accomplish the difficult job of detecting inner defects in mobile phones. These robotics solutions are not subject to fatigue or human error, so it is able to increase product quality by using them for quality control.

In the past, some of the limitations of industrial robots was that they were not mobile, had limited sensor capabilities, and were only suited for one task. Now with technology advances such as computer vision, and AI-enabled deep-learning, robots can perceive their environment, learn new tasks, and work much more effectively with humans.

The Daystar robot provides a 3D video stream to the human operator, from the Daystar augmented reality AR headset using a 5G network for real-time transmission of data. The robot’s arm and camera are closely mapped to the movement of the human operator, to make this experience dynamic and life-like with a fast response time.

This level of advanced robotics requires intensive computing resources and data processing. Daystar for example, utilises Lenovo’s Intelligent Edge Computing technology to enable its high-end capabilities.

As part of the spray-painting process, virtual machines are needed to support AR model rendering, while containers are needed to support painting, quality inspection, and other applications.

To meet these needs, multiple machines were traditionally put in place to run different virtualisation stacks. Lenovo’s Hybrid Light-Weight Hypervisor technology, however, is capable of providing light-weight virtual machines and secure containers to host these applications on a single machine. This effectively increases system resource utilisation rate, while maximising the model-rendering performance, and message processing performance among different tasks of the robot.

Furthermore, painting quality inspection requires the deep learning network models to be compressed and deployed to the edge. Lenovo’s model optimisation technology can realise fine-grained matching between model accuracy and computing capability.

Besides common defects, there are new types of defects that the robot has not encountered previously. The pre-trained models deployed from the cloud to the edge cannot deal with such problems. But through learning technology, these models are updated at the edge, so that the robot can learn to address new problems and expand new capabilities.

With the evolution and acceptance of new technologies like 5G and IoT, specifically industrial and commercial IoT, the opportunities for leverage of such offerings are continually growing. As 5G infrastructure proliferates in accessibility, so too will the use cases, and these in turn will be made available to SMEs and not just Large Enterprise clients.

This will be particularly relevant as the price points of the connectivity and the infrastructure solutions are reduced

Key takeaways

• Daystar Robot uses 3D scanning to model its workspace, with localisation and mapping technology to navigate.
• Daystar Robot provides a 3D video stream so that an operator can view through an AR headset.
• Daystar Robot is intended to replace human workers in production lines, restricted workspace, toxic environments.
• Lenovo also has a range of robot vacuum cleaners, meant for consumer use in the home.
• Lenovo is a partner for UK-based AI researcher React Robotics, which is developing the DogBot.
• DogBot is a quadruped robotic helper which is being tested in construction environments.
• DogBot can support humans in high level tasks like inspections, and it can also carry out basic tasks.
• Lenovo has packaged computer vision into its Machine Vision Solution for Smart Manufacturing to inspect assembly line defects.
• Used in Lenovo’s China manufacturing plants, it makes it possible to detect inner defects in mobile phones.
• Robotics solutions are not subject to fatigue or human error, so it is able to increase product quality.

Previous limitations of industrial robots were limited mobility, sensors, and single task, which is changing with computer vision and deep-learning.