Large language models (LLMs), such as the model underpinning the functioning of OpenAI's platform ChatGPT, are now widely used to tackle a wide range of tasks, ranging from sourcing information to the generation of texts in different languages and even code. Many scientists and engineers also started using these models to conduct research or advance other technologies.
In the context of robotics, LLMs have been found to be promising for the creation of robot policies derived from a user's instructions. Policies are essentially "rules" that a robot needs to follow to correctly perform desired actions.
Researchers at NYU Tandon School of Engineering recently introduced a new algorithm called BrainBody-LLM, which leverages LLMs to plan and refine the execution of a robot's actions. The new algorithm, presented in a paper published in Advanced Robotics Research, draws inspiration from how the human brain plans actions and fine-tunes the body's movements over time.
"LLMs have demonstrated a strong understanding of human interactions within real-world environments," Vineet Bhat, co-first author of the paper, told Tech Xplore. "In this work, we aim to evaluate this capability in the context of robotics by granting the LLM partial access to a fixed set of robot control commands. To ensure safe deployment and controlled environment testing, this access is deliberately constrained."
To read more, click here