ROBOT SYSTEMS, METHODS, CONTROL MODULES, AND COMPUTER PROGRAM PRODUCTS THAT LEVERAGE LARGE LANGUAGE MODELS

Smart overview of the Invention

The patent application discusses innovative systems, methods, control modules, and computer program products designed to enhance robot autonomy by integrating large language models (LLMs). These models allow robots to process natural language (NL) inputs for control parameters and instructions. The LLM generates a task plan in NL, which is then scrutinized for potential errors or faults. If any issues are detected, the LLM can be queried to provide a resolution, ensuring the robot operates efficiently and autonomously.

Technical Field

The invention primarily focuses on improving robot control by leveraging LLMs. It aims to automate processes such as task planning, motion planning, human interaction, and logic reasoning. By using LLMs, robots can interpret and execute complex instructions conveyed in NL, enhancing their ability to perform a wide range of tasks autonomously.

Background

Robots have traditionally been used for various tasks, from simple household chores to complex industrial operations. The integration of LLMs represents a significant advancement in robotic technology. These models, like OpenAI's GPT series, are trained on vast datasets to produce human-like text responses. By incorporating LLMs into robotic systems, the aim is to automate tasks more effectively, allowing humans to focus on other priorities.

Methodology

The methodology involves capturing sensor data from the robot's environment and generating an NL description of this data. This description is used in an NL query to the LLM module, which returns a task plan. The system then checks the task plan for faults; if any are found, it generates an NL description of these faults and queries the LLM for a resolution plan. The robot executes an updated task plan based on this resolution or proceeds with the original plan if no faults are found.

Implementation

The control module includes non-transitory storage media containing processor-executable instructions that enable the robot to interact with its environment using NL descriptions. The system simulates task execution to identify potential faults and adjust steps as necessary. This approach ensures that robots can perform tasks effectively while minimizing errors that could disrupt operations or produce undesirable outcomes.