The robotics team composed of Alessandro Palleschi, Franco Angelini, Chiara Gabellieri, Do Won Park, Lucia Pallottino, Antonio Bicchi and Manolo Garabini has been awarded the Honorable mention "IEEE Transactions on Robotics King-Sun Fu Memorial Best Paper Award" at ICRA 2024. The awarded paper, "Grasp It Like a Pro 2.0: A Data-Driven Approach Exploiting Basic Shape Decomposition and Human Data for Grasping Unknown Objects" proposes an algorithm to train a robotic system to grasp and manipulate different objects.

"Thanks to improvements in their computational and physical intelligence, robots are now able to operate in real-world environments," explains first author Alessandro Palleschi, a PhD student in Smart Industry at DII. "However, their manipulation and grasping capabilities still require significant improvements. To address this problem, we developed a new algorithm called "Grasp it Like a Pro 2.0," which is based on several examples of how a human operator grasps simple shapes such as cuboids and boxes. With these examples, we can teach a robot how to grasp objects of different basic shapes. Next, advanced algorithms and artificial intelligence allow the robot to map any object to a combination of these basic shapes, approximating its geometry, and generate different ways a human could grasp them. This allows the robot to understand and predict the best way to grasp the object. Using this approach, the algorithm can generate and select the best possible grasp, mimicking human dexterity and decision-making. This innovative method significantly improves the ability of robotic systems to grasp and interact with a wide range of objects, making them more versatile and capable in real-world applications, bringing robotic manipulation closer to human performance and taking a decisive step towards the use of robots in real-world environments.
The results demonstrate the effectiveness of our method in generating and selecting reliable and high-quality grasps. With a soft and underactuated robotic hand, our algorithm achieves a success rate of 94.0% in 150 grasps on 30 different objects. Similarly, with a rigid gripper, a success rate of 85.0% is achieved in 80 grasps on 16 different objects."
