A Large Model-based Embodied Intelligent Robotic Arm System

With the rapid advancement of artificial intelligence, particularly large models, robotic

arms have evolved beyond executing pre-programmed tasks. They can now learn and adapt to

new functions through close interaction with their environment, demonstrating capabilities that

surpass traditional disembodied intelligence. This paper explores recent developments in large--

model-driven robotic arms with embodied intelligence and their potential applications. The study

first outlines the development history of large models and embodied intelligence, followed by an

overview of the structure and working principles of embodied intelligent robotic arms. It then examines

their ability to integrate multiple sensory modalities (e.g., vision, hearing) for comprehensive

environmental understanding and the processing of complex tasks. Further analysis covers

multimodal perception, autonomous decision-making, path planning, motion control, and human--

computer interaction feedback mechanisms, highlighting key findings and implementation methods.

A concise overview of current research efforts in multimodal perception is essential. This includes

the development of algorithms for autonomous decision-making, improvements in path

planning and motion control, and the implementation of human-computer interaction feedback

mechanisms. Additionally, a detailed analysis of the associated research methods is necessary. A

synopsis of key advancements in these domains and their implications for large models and robotic

arm technology is essential. Additionally, a discussion of current research challenges is required,

including limitations in environment perception and bottlenecks in inference speed. Ultimately,

it is essential to consider the future implications of these advancements, which have the potential

to revolutionize various fields, including industry, healthcare, and services.