1
About Block
Objectives
The demand for adaptable robots, particularly those with the capability to interact and autonomously grasp objects, has increased due to the growing necessity for automation. Robotic grasping, a longstanding challenge in robotics, is now being readdressed through advances in artificial intelligence and deep learning. These developments are being actively explored in top laboratories and stimulated the interest of many leading robotics companies worldwide.
This workshop, arising from six years of experience in gripper design and grasp detection at the Human and Robot Interaction Laboratory, University of Tehran, will touch upon various robotic grasping methods. The first section of the workshop will showcase numerous implemented and published projects at the Human and Robot Interaction Laboratory, detailing the challenges faced and solutions found in practical applications using parallel two-fingered grippers attached to a 3-DOF Delta parallel robot. The second section will focus on the application of Transformers in robotic manipulation, a cutting-edge area of research.
In this workshop, diverse grasp detection methods will be presented, such as two or three-dimensional-based methods; geometric approaches, abstracting objects into primitive shapes and skeletonizing objects; graph-based methods and reinforcement learning; methods based on the analysis of point clouds and depth images; and imitative learning methods inspired by human behavior . These techniques have applications where a robot equipped with an appropriate gipper is intended to perform a task ranging from chess and Lego assembly to food packaging, scene arrangement such as tools in a workshop, and dishware on a dining table.
The introduction of Transformers has been a game-changer in robotics, NLP, and computer vision, enhancing model capabilities and input versatility, including text and video. With the emergence of models like GPT-4V and Gemini, researchers in fields such as robotic manipulation are now utilizing these models to perform tasks without prior training, relying only on input prompts. The workshop's second section will delve into the role of Transformers in robotic manipulation and review the latest developments in this domain, such as RT-1 and RT-2.
Speakers
⦿ Dr. Mehdi Tale Masouleh, Associate Professor, Director of Human-Robot Interaction Laboratory, University of Tehran
HomePage: https://taarlab.com/mehdi-tale-masouleh/
Linekdin: https://www.linkedin.com/in/mehdi-tale-masouleh-874a477/
⦿ Hamed Hosseini, PhD Candidate, Human-Robot Interaction Laboratory, University of Tehran
HomePage: https://taarlab.com/member-0025/
Linkedin: https://www.linkedin.com/in/hamed-hosseini/
⦿ Hamed Ghasemi, PhD Candidate, Human-Robot Interaction Laboratory, University of Tehran
HomePage: https://taarlab.com/member-0028/
Linkedin: https://www.linkedin.com/in/hamed-ghasemi-1010a8166/?originalSubdomain=ir