An intelligent weight prediction method for dynamic scenes
By using a hierarchical attention mechanism network, the problem of modeling nonlinear correlations between targets in unmanned surface vessel (USV) formation combat scenarios is solved. This achieves efficient and accurate weight prediction, adapts to complex and ever-changing dynamic scenarios, and has good real-time performance and computational efficiency.
Patent Information
- Application Number
- CN202511059319.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-30
- Publication Date
- 2026-02-17
- Estimated Expiration
- 2045-07-30
AI Technical Summary
Existing dynamic methods struggle to effectively model nonlinear dependencies and complex relationships between targets in unmanned surface vessel (USV) formation combat scenarios. They also have a strong dependence on training data, limited generalization ability, and are unable to adapt to new environments, failing to meet the requirements of real-time performance and accuracy.
A hierarchical attention mechanism combining local and global attention networks is adopted. By acquiring unmanned surface vessel (USV) data, the index vectors of each USV are calculated, and attention weights are used for weight prediction. The final weight prediction results are generated by combining a fully connected feedforward layer, and the network parameters are adjusted using a stochastic gradient descent algorithm.
It achieves adaptive adjustment of multiple indicator systems in dynamic scenarios, improving robustness and accuracy. The error is less than 10-3, the standard deviation does not exceed 3×10-3, and the single inference time is stable at about 1 millisecond, meeting the requirements of efficient and real-time weight evaluation.