React Robotics’ DogBot Pursues Revolution in the Construction Industry

Today, the media often writes about four-legged robots. It is reported what new features and capabilities they received and how close to their prototypes. But one question always remained unanswered until the end: how to endow them with intelligence, teach them to navigate independently in the world around them? How and where can you use such four-legged machines in addition to connecting these robots to the analysis of blockages as a result of various emergencies and disasters, as well as military use? Company Boston dynamics already showed how one of her robots Spotmini conducts inspection at the construction site. And this is not the only such example.


Moving around potentially unsafe areas, cleaning the job site and assessing the progress of construction are many inevitable everyday tasks at many construction sites. And all of them are connected with the problems of ensuring the safety of people working at a construction site. British developers intend to reduce risks and delays affecting the efficiency of construction by transferring these jobs to a new four-legged friend – DogBot. They believe that this robot will be able to significantly change the idea of ​​working in the construction industry, and not only in it, in the future.

Four-legged assistant

The robotic four-legged assistant DogBot uses machine learning algorithms for movement, perception of the environment, a sense of one’s own position and position in space and orientation in the surrounding world.

British artificial intelligence researchers from the company React robotics are exploring the possibilities of using robotics in modern conditions, and their brainchild, DogBot, is one of the first such developments. Created using Autodesk Fusion 360 software, the robotic “assistant” is ready to perform tasks such as 3D scanning to monitor the construction process, manage logistics at construction sites and collect data from various sensors in real time.

Unlike land vehicles on wheeled or tracked tracks, a four-legged robot can move around incredibly difficult rugged terrain, carrying a payload or equipment.

“We wanted to find ways to use artificial intelligence in the real world. The robotic equipment we are creating is the path to a future revolution in its use, ”says Charles Galambos, CTO and co-founder of React Robotics. “DogBot-like robots really“ understand ”the world and interact with it.”

For hazardous working conditions

DogBot can be programmed to perform tasks in an industrial environment with a high level of risk, which gives significant advantages for a construction site where there is a likelihood of injury to employees. To ensure the safety of people, DogBot can be used to clean work areas, as well as to collect tools. Then the next shift construction crews will be able to return to a cleaner and safer work environment.

Robot instead of working

According to the developers, DogBot can also be a solution for those almost 80% of construction companies that, according to a recent survey conducted by Associated General Contractors of America in partnership with Autodesk, lack staff. When creating DogBot, the goal of React Robotics was to offer a “tool” that will allow professionals to work more efficiently, while DogBot robots will fill up the shortage of labor in specific areas.

In addition, with a better understanding of the state of construction, DogBot increases overall productivity at the construction site. Feedback is getting better. You can see exactly what has already been built, and decide what to do next, reducing the time and amount of resources spent on the work of the construction site.

Platform for intelligence

The high level of complexity and the large amount of data needed to teach the DogBot robot everything that it needs to do – from moving around the construction site and navigating to understanding how to perform tasks – requires significant computational resources. For this, React Robotics involved various Lenovo equipmentsuch as the AI-configured ThinkStation P920 workstation and ThinkPad P1 mobile workstation.

The Lenovo ThinkStation P920 workstation is equipped with two Intel Xeon processors, three NVIDIA Quadro RTX graphics cards with AI Tensor cores, as well as a variety of input / output interfaces. It can be used for rendering, modeling, visualization, deep learning systems and AI. In React Robotics, it serves as a data compilation system and is used to perform resource-demanding machine learning algorithms and service deep learning workflows.

Previously, the training process took several days, new equipment allows React Robotics to see results almost instantly, which makes it possible to quickly change settings and conduct tests in a testing environment. Ultimately, reduced time to market. The ThinkPad P1 workstation is used for programming and for deploying the DogBot.

How to train a robot?

“We attach great importance to our work with partners. They share our vision for technology adoption, ”said Gregory Epps, CEO of React Robotics.

“Our robot is really capable of perceiving the surrounding world, interacting with it. Each of his legs has 12 degrees of freedom. It can walk and be equipped with various sensors. Most of its parts are printed on a 3D printer. We use Autodesk Fusion 360, so we can quickly develop something new and test the part in one day. We see that the capabilities of robots are developing very quickly. In robotics and in the field of artificial intelligence, something new happens every day. And partners help us achieve meaningful results, ”says Charles Galambos.

As artificial intelligence, machine learning, and deep learning continue to penetrate all industries, the need for high-performance, hardware solutions also grows. Lenovo Workstations help implement complex AI projects. Lenovo P-Series workstations are designed to meet the demanding performance requirements of today’s AI, machine and deep learning applications. Performance is ensured through the integration of AI platforms for computing on GPUs and software systems for data analysis and processing.

This allows you to speed up the processing of machine and deep learning algorithms, including data preparation, model training and visualization tasks, as well as speed up the acquisition of useful information, reduce the cost of data processing and analysis projects using a solution based on NVIDIA Quadro RTX GPUs with tensor kernels.

Lenovo workstations include a range of solutions – from the ThinkStation P920 for deep learning models based on desktop PCs and the ThinkStation P520 for developing models of artificial intelligence and peripheral computing to the universal ThinkStation P330 Tiny for generating inference based on artificial intelligence.

Change is coming

As examples of robots like DogBot, which help construction professionals, show AI will have an increasingly noticeable impact on jobs, not only in construction, but in a wide variety of industries. AI industry partnerships are expanding.

In August 2019, Lenovo and Intel announced a collaboration aimed at optimizing their technologies for data centers. It aims at bringing HPC and AI closer together.
Lenovo’s cloud services will be adapted to Intel’s development, including Intel Xe computing architecture, Optane memory, oneAPI platform and 2nd generation Xeon Scalable processors with support for Deep Learning Boost technology.

Intel and Lenovo work with software as well. So, Lenovo will finalize its LiCO HPC / AI package with an eye to compatibility with Intel oneAPI and other partner software. In addition, HPC and AI joint development centers will appear in different countries. Companies hope to make these technologies more accessible to universities and organizations involved in solving problems such as genome research, climate change, space exploration, etc.

Platform LiCO Available at Lenovo AI Innovation Labs labs around the world. Third-party companies can test their solution before deploying. Lenovo’s innovation centers are equipped with the necessary hardware and software, and also have specialists in the field of AI.

Concerning mobile robots, then researchers from different countries are currently working on several similar projects in the energy and manufacturing sectors, as well as in construction, agriculture and other more specialized fields – where mobile robots can provide significant assistance and support.

Undoubtedly, over time, new ideas for using such machines will appear. For example, in Russia in early 2020, a new anthropomorphic robot will be tested, which will be used for waste disposal in radioactive canyons. In the United States, Massachusetts police began testing four-legged robots to complete tasks. The range of commercial opportunities will grow along with the capabilities of the robots themselves. The high cost of the final product will decrease, even without taking into account development costs, remains one of the key shortcomings of such “smart” systems.

Similar Posts

Leave a Reply Cancel reply