On-Device AI Robot Control Using sLLM to Cut Network Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robot applications using generative AI face issues with slow processing speed, network overload, high costs, and inadequate personal information security due to reliance on remote Large Language Models (LLMs), which are not optimized for user-specific learning and intimacy in conversation.
Innovation Solution
Implementing a smaller Large Language Model (sLLM) on-device for direct interaction between a user's terminal and a companion robot, enabling over-the-air programming (OTA) for distribution and automated conversation, thereby optimizing learning capabilities and enhancing user intimacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If remote LLM is used for robot applications, then generative AI capabilities are provided, but processing speed becomes slow and network overload occurs
Solution Approach 1:
The patent segments the LLM functionality by distributing smaller language models (sLLM) to individual robot terminals through OTA updates. This divides the centralized processing burden into distributed local processing units, enabling faster response times while reducing network traffic for inference operations.
Solution Approach 2:
The patent transitions from a single centralized LLM dimension to a multi-dimensional architecture where multiple sLLM instances run distributed across multiple robot terminals. This dimensional shift enables parallel processing and reduces network overload by localizing AI inference at the edge devices.
2Adaptability or versatility
If ChatGPT is used for robot applications, then conversational capabilities are enhanced, but costs increase exponentially and personal information security is reduced
Solution Approach 1:
The patent enables robots to self-serve conversational AI capabilities by hosting sLLM locally on their terminals. This eliminates dependency on external paid APIs like ChatGPT, reducing operational costs while maintaining conversational functionality through locally hosted language models.
Solution Approach 2:
The patent introduces sLLM as an intermediary between the robot and users, processing conversations locally rather than transmitting sensitive user data to external servers. This intermediary architecture preserves personal information security by keeping data processing within the trusted robot terminal environment.
3Productivity
If remote LLM platform is used, then AI capabilities are provided, but the number of servers must continuously increase
Solution Approach 1:
The patent extracts the language model inference capability from the centralized server environment and places it directly on robot terminals. This extraction eliminates the need for continuously expanding server infrastructure while maintaining AI processing capabilities at the edge devices.
Solution Approach 2:
The patent inverts the traditional client-server AI architecture by making the robot terminal the host of the language model rather than the server. This inversion shifts computational responsibilities from centralized servers to distributed terminals, reducing server infrastructure requirements.
Data Source
AI summary
A robot control system using on-device AI includes an on-device AI control server to distribute a smaller Large Language Model (sLLM) for on-device AI execution; an on-device AI terminal to load and execute the sLLM distributed from the on-device AI control server; and a companion robot to carry out question asking and answering interaction with the sLLM running on the on-device AI terminal and make automated conversation with a user based on the question and answer. According to the robot control system and method using on-device AI, it may be possible to enable direct questioning/answering between the terminal and the robot by using the sLLM installed on each user's terminal without using LLM equipped on a platform, thereby solving the problem with network load caused by questions/answers, enhancing the optimized learning capabilities for each user, and increasing intimacy in conversation with the user.