Neural Networks for high accuracy short term ship motion predictions with applications to autonomous UAVs
Bibliographic record
Abstract
• Computationally efficient models which can be used by UAV mounted computers • Comparison of common Neural Network model and new model • Robust Neural Network models that can now be used without noise filtering • Large augmented dataset to reflect real data measured from UAV An Autonomous Uncrewed Aerial Vehicle (UAV) attempting to perform a vertical landing on a moving ship’s deck must be capable of predicting the ship’s motion in order determine the most opportune landing time. If the UAV is acting independent of the ship additional constraints are introduced; computation resources are limited to what can be mounted on the drone and the UAV must predict motion from noisy UAV-mounted sensors. The work presented proposes a Gated Recurrent Unit based Autoencoder (GRU-A) Neural Network (NN) model for predicting future ship motion with the aforementioned constraints. The GRU-A model is compared to a more typical Multi-Layered Perceptron, Nonlinear Auto-Regressive (MLP-NAR) NN model. Both NN models are tested for their ability to minimize error over a 5 s prediction horizon composed of 50 separate time-steps, their ability to predict through noisy inputs and mitigate the introduced error, and their computation costs. Furthermore, a large dataset made from a high fidelity simulation is transformed to reflect data that would be encountered in-situ, improving the applicability of the work. It was found that the proposed GRU-A model has superior signal prediction capabilities, achieving approximately 30 times lower error than the MLP-NAR model when predicting over a 5 s period, suitable of a vertical landing time horizon. In addition, the proposed GRU-A model was more resilient to input noise and, when trained with noise it outperformed the MLP-NAR. It was also found that the memory required to compute predictions with both models is approximately equal and that the computation time of the GRU-A model is similar to the MLP-NAR model with both models being capable of making predictions within 100 ms so long as they are not chosen to be too large. Overall, the proposed GRU-A model is demonstrated as a superior alternative to the more typical MLP-NAR model when predicting full signals in all cases for use with a small UAV.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".