Robot control method and system based on dual system, training method and system
By having the decision-making and execution systems work together in a dual-system architecture, the decision-making system breaks down complex tasks and outputs control conditions, while the execution system adaptively generates multimodal thought chains. This solves the problems of insufficient logical reasoning and lack of fine granularity in long-term robot tasks, thereby improving task completion and success rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XINGHAITU (SUZHOU) ARTIFICIAL INTELLIGENCE TECHNOLOGY CO LTD
- Filing Date
- 2026-06-12
- Publication Date
- 2026-07-17
AI Technical Summary
Existing robot control systems suffer from insufficient high-level logical reasoning and long-term task decomposition capabilities in long-term tasks, and lack fine-grained directionality, resulting in poor task completion and catastrophic forgetting problems.
A robot control method based on a dual system is adopted. Through the collaborative work of the decision system and the execution system, the decision system performs cognitive decomposition of complex tasks and outputs control conditions, while the execution system adaptively generates a multimodal thought chain based on the control conditions, thereby realizing hierarchical collaboration between cognition and execution.
It improves the completion rate and success rate of long-term tasks, solves the problem of the lack of fine-grained directionality in the execution system of long-term tasks, and significantly improves the success rate of multi-stage operations.
Smart Images
Figure CN122401441A_ABST