Whether to grasp, hold, turn, touch, type, or press – we use our hands for a wide variety of tasks in our everyday lives without thinking twice about it. And yet the human hand is one of nature’s most miraculous tools. So what could be more natural than equipping robots in collaborative workspaces with a gripper inspired by this natural model? Robots that can then learn to solve a wide variety of gripping and turning tasks through the use of artificial intelligence?
The reinforcement learning paradigm, a system of learning based on rewards, is used with the BionicSoftHand. This means that instead of defining a specific action that the robot must imitate, the hand is merely given an goal. It then attempts to achieve this through trial and error. Based on the feedback it receives – both positive and negative – it gradually optimizes its actions until it finally completes the task successfully.
Specifically, the BionicSoftHand is tasked with rotating a twelve-sided cube so that a previously defined side points up in the end. Learning the necessary motion strategy takes place in a virtual environment using a digital twin, which is created using data from a depth camera and artificial intelligence algorithms.
The digital simulation model accelerates learning significantly, especially if you replicate it. Through Massive Parallel Learning, the acquired knowledge is shared with all of the virtual hands, which then continue to work with the new state of knowledge – meaning each mistake is made only once. Successful actions are immediately available to all of the models.
After the controller has been trained in the simulation, it is transferred to the real BionicSoftHand. Using the movement strategy acquired in the virtual environment, it can turn the cube to the desired side and move other objects correspondingly in the future. As a result, building blocks of knowledge and new skills that have been learned can be shared with other robot hands and made available globally.
Unlike the human hand, the BionicSoftHand does not have any bones. It controls its movements via the pneumatic bellows structures in its fingers. When the chambers are filled with air, the fingers bend. If the air chambers are empty, the fingers remain straight. The thumb and index finger are additionally equipped with a swivel module, which allows these two fingers to also move laterally. This gives the bionic robot hand a total of twelve degrees of freedom.
The bellows in the fingers are enclosed in a special 3D fabric sheath knitted from both elastic and high-strength threads. As a result, the fabric can be used to determine exactly where the structure expands and thus develops force, and where it is prevented from expanding.
In order to keep the amount of tubing required for the BionicSoftHand as low as possible, the developers designed a small, digitally controlled valve manifold that is mounted directly below the hand. This means that the tubing used to control the fingers does not have to be pulled through the entire robot arm. As a result, the BionicSoftHand can be quickly and easily connected and brought into operation with just one pneumatic tube each for supply air and exhaust air. The proportional piezo valves used allow the movements of the fingers to be precisely controlled.
Its flexible, pneumatic kinematics and the use of elastic materials and lightweight components make the BionicSoftHand stand out from electric or rope-operated robot hands and make its low-cost production possible. Thanks to its modular design, gripper models with three or four fingers can also be used.
Combined with lightweight pneumatic robots – such as the BionicCobot or the BionicSoftArm – the robot hand can be used for safe and direct human-robot collaboration. Both robots are designed to yield and do not have to be isolated from workers like conventional factory robots.
This makes the BionicSoftHand the perfect choice for applications in the collaborative workspaces of tomorrow’s factories. Since the flexible robot hand can grip both powerfully and delicately, its use as a helping third hand in assembly is just as conceivable as its use in service robotics.