Multi-modal dynamic weight adaptive fusion acquisition method and system for body-equipped robot
By real-time evaluation of multimodal factors and incremental dynamic weight generation, combined with inverse reinforcement learning and factor graph feedback, the problems of single dimensionality of weight evaluation and unbalanced response in existing technologies are solved, enabling embodied robots to perform tasks with high precision and safety in complex environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YIBO INTELLIGENT TECH (GUANGZHOU) CO LTD
- Filing Date
- 2026-06-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multimodal fusion technologies cannot comprehensively reflect a robot's motion operation capabilities, the degree of environmental change, and the urgency of the task. Furthermore, the weighting evaluation dimensions are singular, making it impossible to make optimal decisions in complex environments, and they lack the ability to balance historical information and instantaneous response.
An adaptive fusion method with multimodal dynamic weights for embodied robots is adopted. By evaluating the operational margin factor, environmental roughness factor, and task urgency factor in real time, and combining an incremental dynamic weight function and an experience-guided inverse reinforcement learning algorithm, adaptive fusion weights are generated, and the planning information is optimized by using a factor graph feedback mechanism.
It achieves smooth response of fusion weights and coordination of planning information, improves the robot's task safety and fusion accuracy in complex environments, enables agile response to environmental changes or sudden disturbances, and ensures the stability and rationality of weight adjustment.
Smart Images
Figure CN122401431A_ABST