Application of generative AI for robotics learning and perceptions

Abstract

© 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse.

This paper investigates the application of Generative Artificial Intelligence (GenAI) to improve the learning and perception capabilities of robots. Rapid progress in machine learning and neural generative algorithms is expected to bring about major changes in how robots gather information, adjust to their surroundings, and understand what they sense. This paper examines how generative models like Generative Adversarial Networks (GANs), Large Language Models (LLMs), Variational Autoencoders (VAEs), and diffusion models can help robots learn more effectively for tasks such as applying knowledge from simulations to real life, understanding their environment, improving sensor data, and interpreting various types of information. These models significantly reduce data needs, facilitate unsupervised and autonomous learning, and enhance resilient adaptability in dynamic and unpredictable environments. This study shows that GenAI has enormous potential to accelerate the development of smart, adaptable, and perception-based robotic systems capable of operating independently in complex real-world situations. Moreover, this study provides concise information on the application of GenAI models in robotics.

This work was financially supported by the MERIT Project of the European Union under grant agreement number 101083531.

Peer Reviewed

Postprint (author's final draft)