The history of modern robotics
General Motors deployed the first mechanical-arm robot to operate one of its assembly lines as early as 1959. Since that time, robots have been employed to perform numerous manufacturing tasks such as welding, riveting, and painting. This first generation of robots was inflexible, could not respond simply to errors, and required individual programming specific to the tasks they were designed to perform. These robots were governed and inspired by logic—a series of programs coded into their operating systems. Now, the next wave of intelligent robotics is taking advantage of a different kind of learning, predicated on experience rather than logical instruction, to learn how to perform tasks in much the same way that a child would.
Transforming data into experience
Industrial robots are capable of extreme precision and speed, but normally require careful programming to perform simple functions such as grabbing an object. By contrast, for humans, recognizing and interacting with an object is simple and requires almost no thought whatsoever, since we already have the data we need and immediately apply it to the situation. This is because humans have already been through a massive data collection process ourselves: childhood.
For robots, similar routine tasks require access to massive amounts of data regarding how objects are differentiated from one another and how to best handle or interact with them. This data needs to be collected, stored, analyzed, and processed.
Modern deep learning methodologies backed by high-performance computing have enabled a new training method empowering robots to generate their own experiential data and take it into account when performing tasks.
Training the next generation of robotics
A technique called reinforcement learning is enabling cognitive robots to train themselves on new tasks over time. In one example, a robot might try to pick up objects while capturing video footage and sensor data regarding its progress and subsequent result. Each time the robot succeeds or fails, it records data and uses it to refine a deep learning model, or a large neural network, that informs future actions. This deep learning approach enables robots to recognize patterns and perform actions in response.
Reinforcement learning saves tremendous time by eliminating the need for each action to be programmed by a human expert. Alternatively, engineers can give the robot a task, such as picking widgets out of one box and putting them into another, and it will spend the night figuring out how to perform the task. Come morning, the robot has mastered the job as well as if it were programmed by an expert. In the past, this task would be completed by matching image data to prepared CAD data and returning the target position to fetch. Now, the task is conducted entirely by the robots themselves, who, if they fail to pick up an object the first time, will keep the image and its depth data as a failure and update its algorithm according to the data acquired.
Deep learning has also enabled robots to simulate real-world conditions prior to being deployed. Using these technologies, developers can set up extensive test scenarios and simulate them in minutes before finally transferring the information to real-world robots.
Delivering the future of robotics
In addition to decreased training time, one of the major potential benefits of the reinforcement learning approach is that it can be accelerated exponentially when several robots work in parallel and share what they have learned among themselves. Fanuc, a Japanese robotics company that is training robots using neural networks, is focused on how teams of robots can work together cooperatively. Rather than storing data in a centralized location, these robots use edge-heavy computing to process their sensor data while collaborating with one another. The accelerated pace is proportional to the number of robots participating. For example, eight robots can learn in an hour what a single robot would take eight hours to learn by itself. This form of distributed learning, sometimes called “cloud robotics,” is shaping up to be a big trend both in research and industry, and is key to the future of industrial IoT.
More traditional robots have also benefited from the advantages afforded by modern technology. Sarcos Robotics, a leader in dexterous, tele-operated robots designed for use in unpredictable or unstructured settings, has recently announced a mobile IoT robot designed to inspect otherwise inaccessible or dangerous environments. The Sarcos robot, while man-operated, leverages the Microsoft Azure cloud computing platform and Microsoft Azure IoT to enable secure storage and retrieval of environmental video, audio, and sensor data. The combination of IoT sensors and Microsoft Azure services helps these robots better evaluate the performance of the various industrial machines in which they are placed, while also analyzing cloud data to better predict required maintenance.
Deep learning and IoT connected devices are working together through the cloud to drive the next revolution in robotics, and modern manufacturers are beginning to reap the benefits. With Microsoft technology, you can quickly begin to take advantage of these future-facing solutions to optimize your manufacturing processes and improve worker safety.
Are you ready to unlock the full value of Big Compute for your business? Learn more about our Big Compute solution portfolio today.