一种面向机器人换电表作业的多模态行为数据构建方法及系统
By constructing a multimodal behavior data system, the problems of single perception modality, insufficient time synchronization, and insufficient scene coverage in robot meter swapping operations have been solved. This system enables high-quality data acquisition and structured processing, and improves the ability to train models and analyze anomalies.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- FUDAN UNIVERSITY
- Filing Date
- 2026-05-12
- Publication Date
- 2026-07-17
AI Technical Summary
Existing robot-based meter swapping datasets suffer from limited perception modalities, insufficient time synchronization accuracy, and a lack of semantic annotations, making it difficult to cover complex scenarios and abnormal operating conditions. Consequently, the models lack generalization ability when deployed in real power fields.
The system is built using multimodal behavioral data. By uniformly collecting visual, depth, robotic arm status, end effector force and pose data, it achieves time synchronization and structured encapsulation, performs semantic annotation, expands the scene coverage to complex working conditions, and combines quality control and versioned data storage.
It improves the integrity, temporal consistency, structuring degree and semantic density of multimodal behavioral data, enhances the interpretability and reusability of data, covers complex scenarios and abnormal working conditions, and ensures data quality stability and long-term maintainability.
Smart Images
Figure CN122185242B_ABST