GMM-searcher: efficient object search in large-scale scenes using large language models

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Science

Background:

• Autonomous object search in large-scale, dynamic environments is challenging.
• Existing methods struggle with environmental complexity and adaptability.

Purpose of the Study:

• To introduce GMM-Searcher, a novel framework for efficient robot-guided object search.
• To enhance robot adaptability and lifelong learning in complex environments.

Main Methods:

• Leveraging large language models (LLMs) for intelligent search guidance.
• Combining adaptive-resolution topological graphs (ARTG) with Gaussian Mixture Models (GMM) for efficient memory usage and environmental fidelity.
• Utilizing GMM to store search experiences for performance enhancement and adaptation.

Main Results:

• GMM-Searcher effectively directs robots to likely object locations.
• The framework optimizes memory usage in large-scale environments.
• Search performance improves with repeated tasks and environmental changes.

Conclusions:

• GMM-Searcher significantly enhances autonomous object search efficiency.
• The framework provides robots with lifelong learning and adaptability.
• LLM-guided search strategies are effective in dynamic and complex environments.