In the whole human history, the role that the abilities of our hands do not underestimate. From the prehistoric man who manipulates the earliest instruments to the accuracy of the demonstration of modern surgeons, this dexterity is based on the limb, which included 27 bones and more than 30 muscles, the leadership of perhaps the most people of all organs: brain.
This complexity makes a robotic hand with a highly demanding control. In the world of robotics, there is no higher level than the fine motor skills needed to understand and manipulate objects at exact speed and force.
Meanwhile, companies like Google Deepmind move the boundaries of artificial intelligence (AI) and try to understand what machines can learn, both to expand the spectrum of practical possibilities and research. When Google Deepmind wanted to expand the machine learning in a complex field of robotic hands, they encountered a video of one such model that learned how to quickly finish Rubik’s cube.
The hand of the robot for the real world
It was the Shadow Robot’s Shadow Hand, developed in cooperation with Openi, which impressed the Google Deepmind team. However, this new project still asked something.
“Google Deepmind wanted a robotic hand capable of learning about the real world tasks,” explained Rich Walker, director of Shadow Robot. “Hand should be the most skillful and sensitive but developed, but unlike the other robots they tested, needed it to survive, albeit subject to impact on hard and practical tasks.”
Google Deepmind asked to include a high number of sensors to prefer data collection, so the Shadow Robot has set up to design a hand, as it says, “Much more sensors than would be sensitive in any other context.”
The aim was to create a robotic hand with high dexterity, sensitivity and robustness for real tasks without replicating the appearance of the human hand. To achieve these needs, design recession on three robust fingers and hands about 50% larger than the human hand.
The result is a dex-ee, a robotic hand full of high-speed sensor networks that provide rich data included, strength and inertial measurement. It is widespread by hundreds of tactile finger sensing channels, which optimizes sensitivity to the level of dizzying size, almost similar to the level of the human hand.
Innovation of the drive system
To perform a gentle control over the application of force and activate a number of joints in his hand, the shadow robot had to rely on a highly capable drive system. The key dex-ee innovation is its unique design that has a tendon driver using multiple joints for the joint instead of a typical single-on-board single-pin.
Since five engines control four joints on each of the three fingers, this approach eliminates the will, the “game” that can occur when the direction of movement is turned to optimize controlled movement. With careful control of each engine, each joint can mimic the torque of the zero joint, which gives the Dex-EE beautifully sensitive movement control and the ability to handle fine objects without risk.
To achieve the necessary reliability and performance of Dex-EE, Shadow Robot turned to its original partner partner.
Dex -e Dexterous Robotic Hand, developed Shadow Robot, in collaboration with Robotics Google Deepmind. | Source: Shadow Robot
“Maxon Motors has had a long development of production and the pedigree they bring was essential for an application that would be placed on Dex-EE,” Walker said. “This was especially the case of strictness in the real world that Google Deepmind was looking for.”
Dex-EE integrates a total of 15 Maxon DCX16 DC engines that achieve high torque density necessary for robotic hand to apply through tendons. This allows your hands to move with the desired dynamics and force for actions such as gripping and holding. At the same time, the engines were compact enough to fit into the boundaries of each finger base.
The free engine winding also eliminates the orbits, relative jerk generated by traditional iron design designs. This helps to achieve a smooth, controlled movement necessary for the Dex-EE to achieve the demanding level of accuracy for the most delicate tasks. High tolerance in design and production, along with premium materials, ensures quiet operation and achieves a high life.
Sign up today and save 40% We Conference Pass!
The future of robotic hands
DEX-EE and liabibility have been ensured with more than 1,000 hours of testing. This included the simulation of the process known as political learning, where AI examines how to effectively achieve the task by engaging repeated random movements, which also caused mechanical stress. The Shadow Robot team also subjected to the Dex-EE high level of impact and impact testing, including pistons and various tools.
Google DeepMind has already published research representing Dex-EE, involving a video showing the ability of a robotic hand to manipulate and attach a connector in a limited workspace, sufficiently closed hands of the robot to enforce. This task emphasizes the robustness of Dex-EE and shows how it lasts repeated collisions against the walls of the workspace while still finishing the task.
“Google Deepmind uses Dex-EE as a research platform to study learning in the real world and robustness and hand sensitivity allows it to interact with objects in a way that would damage traditional robots,” Walker said.
DEX-EE is now also an ESO research platform for wider organizations. And while the creation of Shadow Robot has been developed for further understanding of machine learning in everyday environment, Walker said that complex robotic hand technology in the future is increasingly integrating into everyday life. As the technology has become normalized, he said that the “robot” label could begin to disappear as soon as the device becomes common.
“In the future, people working in Robotice will develop development devices that we use every day. On this internship we will no longer call it a “robot”. Our perception may not be exciting for a long time, because our current ideas about what a robot should be, but in fact, these devices could be far from humanity than we imagined first. ”