Abstract

Generative Artificial Intelligence (GenAI) is fundamentally reshaping robotics, moving the eld beyond rigid, pre-programmed systems toward flexible, adaptive, and creative machines. Traditional robotics has long relied on precise control systems, detailed planning, and narrow task definitions, but GenAI through technologies such as large vision-language models, diffusion models, and imitation learning enables robots to learn from demonstrations, natural language, and online data. These advances are further amplified by collaborative efforts like Open X-Embodiment, which pool data from diverse robots to build scalable, generalist AI models. Despite these breakthroughs, significant challenges remain before robots can be fully integrated into everyday life. Issues such as safety, interpretability, data efficiency, and real-time performance continue to limit deployment in high- stakes or consumer-facing contexts. Moreover, robots still lack the general- purpose commonsense needed for complex, multi-step tasks in unstructured environments. Nonetheless, the future of robotics is being rapidly transformed by GenAI, with promising directions including open-ended skill acquisition, personalized user interactions, and integration with emerging technologies. Accordingly, this review discusses recent research, challenges and applications of GenAI and robotics and its impact in real-life applications.

Keywords:Generative artificial intelligence; Intelligent robotics; Large vision language models; LLM-driven robotics