Awarded author: Ing. Elizaveta Isianova
University / Research Institution: Czech Technical University in Prague, Faculty of Electrical Engineering
Title of the Thesis: Estimating Object Grasp Positions Using Multimodal Models
Robots Gain Smarter Vision
The Werner von Siemens Award for the Best Master’s Thesis addressing Industry 4.0 topics was awarded to Ing. Elizaveta Isianova from the Faculty of Electrical Engineering at the Czech Technical University in Prague (also a researcher at the Czech Institute of Informatics, Robotics and Cybernetics working in the RICAIP Testbed Prague) for her master’s thesis titled Estimating Object Grasp Positions Using Multimodal Models. The thesis was supervised by Ing. Varun Burde from the Czech Institute of Informatics, Robotics and Cybernetics.
Elizaveta Isianova’s thesis focuses on one of the key challenges in robotics: autonomous manipulation of objects in the real world. It introduces a solution that integrates modern vision–language models (VLMs) into the grasp planning process. These models work similarly to ChatGPT but additionally have “eyes.” Using VLMs, Elizaveta Isianova developed a method for generating six-degree-of-freedom (6-DoF) grasp positions for two types of grippers: a parallel gripper and a vacuum gripper.
At this stage, the thesis does not present a final solution to this complex problem. However, by transferring the latest advances in artificial intelligence into practical robotics, it represents an important contribution to a rapidly developing field with enormous future potential. Applications may include industrial manufacturing, logistics and e-commerce operations, as well as household and service robotics.
Vision With Understanding
Elizaveta Isianova’s thesis explores how artificial intelligence can help robots better understand the surrounding world. Industrial robots are extremely precise, but they lack contextual understanding.
“When a robot sees a drill, it doesn’t know that it should be held by the handle. It only sees a cluster of geometric shapes and points. If a human does not explicitly program the grasp coordinates, the robot might grab the drill by the drill bit or the trigger button, which is both dangerous and impractical,” explains Elizaveta Isianova.
She decided to address this challenge using cutting-edge technologies, specifically vision–language models (VLMs). The software she developed allows a robot to look at an unfamiliar object and understand it semantically. The robot can effectively realize that “this is a handle, this is meant for holding,” and then calculate the optimal grasp point in that safe zone.
“I tested this approach on a real robot equipped with a vacuum gripper. The results showed that thanks to AI integration, we no longer need to laboriously reprogram robots for every new object. The robot becomes more autonomous and can operate in changing environments, which is essential for modern automation or future home robots,” adds Elizaveta Isianova.
Abstract Artificial Intelligence Meets Heavy Hardware
Methods based on this research may contribute to the development of flexible industrial production lines in the future. If the need for complex reprogramming for every new component can be removed, robots will be able to quickly adapt to new products on their own, saving companies both time and money.
In logistics, handling a wide variety of goods remains a major challenge. The development of “smart vision,” to which this thesis contributes, is one step toward enabling robots to safely recognize and grasp objects such as fragile vases without breaking them.
“In my work, I also try to look ahead,” says Elizaveta Isianova. “One day I would like a robot to be able to empty a dishwasher at home. To do that, it must understand the semantics of objects — holding a mug by the handle, a knife by the handle. Our research belongs to a field that gradually brings us closer to that goal. What fascinates me most is the moment when a robot stops being just a ‘blind’ machine. I’m fascinated by the connection between two worlds: abstract artificial intelligence that understands images, and heavy industrial hardware.”
Robotics Offers a Broad Perspective
Elizaveta Isianova enjoys challenges and wants to see a real impact from her work, whether scientific or practical.
“Robotics is ideal for that because it offers immediate feedback. You program a system and instantly see how your idea turns into a real physical movement. It’s not just abstract theory, but something that actually works and moves. At the same time, it’s a field that is developing very quickly today. That gives me a huge opportunity not only to keep learning new things but also to actively contribute to technologies that may shape our future.”
What she enjoys most about robotics is its complexity and the way it connects different fields.
“Robotics is not a single discipline. It includes a wide range of topics, from electronics and embedded systems to advanced artificial intelligence and computer vision. I’m fascinated by the broad perspective this field gives me. It has allowed me to build a solid technical foundation across many disciplines and to look for connections where others might not see them. It’s like putting together a puzzle where hardware and software fit neatly together,” she concludes.
