Integration of Large Language Models for Autonomous Navigation of a Mobile Robot

The paper considers the integration of large language models (LLMs) to improve autonomous navigation of mobile robots. Traditional navigation methods are often limited by rigid algorithms that cannot adapt to dynamic environments. In contrast, LLMs enable robots to interpret complex spatial scenarios and make more informed decisions. The study demonstrates a new approach to robot control using the ROS with Gazebo simulator, where LLMs transform visual and sensory data into navigation commands. Experimental results demonstrate the feasibility of using LLMs to navigate in a pre-undefined environment improving the adaptability and autonomy of robots. The most successful model was the GPT -4o, and the best camera field of view was <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$90^{\circ}$</tex>. The paper highlights the potential of LLMs in creating more intelligent and intuitive robot control systems.