Hybrid Regression Framework for Trajectory Prediction of Supercavitating Vehicles
Kangjian Wang , Ruoyu Du
Journal of Marine Science and Application ›› : 1 -24.
Trajectory prediction of supercavitating vehicles remains challenging due to complex hydrodynamics and limited observation. To address the restricted field of view in optical measurement and the insufficient accuracy of the adopted numerical model, this study proposes a hybrid regression framework. It integrates time-frequency domain features with physical constraints from impact points. Using multi-view high-speed camera data and simulation results, key features reflecting dynamics such as tail-slap are extracted through piecewise alignment and error frequency-domain analysis. A hybrid feature space is constructed from time-domain polynomials and frequency-domain basis functions, and features are fused via ridge regression. Experimental validation shows that the proposed method achieves high-accuracy trajectory reconstruction. The overall root-mean-square error is reduced from 0.562 m to 0.005 6 m, an improvement of about 99%. Errors in all segments remain stable at the millimeter level. The framework applies to supercavitating vehicles with similar motion characteristics, where impact point constraints and multi-view fragmentary observations are available. This work provides a methodological reference and insights for real-time trajectory prediction of supercavitating and underwater vehicles.
Supercavitating vehicles / Trajectory prediction / Hybrid regression / Terminal accuracy / Data driven correction
| [1] |
|
| [2] |
|
| [3] |
|
| [4] |
|
| [5] |
|
| [6] |
|
| [7] |
|
| [8] |
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
|
| [16] |
|
| [17] |
|
| [18] |
|
| [19] |
|
| [20] |
|
| [21] |
|
| [22] |
|
| [23] |
|
| [24] |
|
| [25] |
|
| [26] |
|
Harbin Engineering University and Springer-Verlag GmbH Germany, part of Springer Nature
/
| 〈 |
|
〉 |