Computational Model Emulation for Memory-Efficient Air Combat Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional software models for predicting air-to-air combat scenarios require large amounts of digital memory and have low throughput on aircraft hardware, necessitating a more efficient model for accurate predictions.
Innovation Solution
Training a first computational model to emulate a second computational model by generating outputs, selecting rewards based on output differences, and updating the model using these rewards to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional software models are used to predict air-to-air combat scenarios, then prediction accuracy is maintained, but digital memory requirements increase and throughput decreases
Solution Approach 1:
The patent creates a simplified copy (surrogate model) of the complex conventional software model. This copy replicates the essential predictive functionality of the original model but uses significantly fewer digital memory resources. The surrogate model is trained to emulate the input-output behavior of the conventional model, providing accurate predictions while reducing memory requirements from megabytes to kilobytes.
Solution Approach 2:
The patent extracts only the essential predictive functionality from the complex conventional model, separating the core prediction capability from the computational overhead. By isolating and implementing only the critical prediction function in a simplified form, the system achieves accurate combat scenario predictions without requiring the full computational infrastructure of the original model.
2Measurement precision
If conventional software models are used to predict air-to-air combat scenarios, then prediction accuracy is maintained, but throughput on aircraft hardware decreases
Solution Approach 1:
The surrogate model serves as a lightweight copy that replicates the predictive accuracy of the conventional model while executing much faster on aircraft hardware. The simplified architecture enables real-time predictions during combat scenarios, whereas the conventional model would be too slow for operational use.
Solution Approach 2:
The patent changes the computational parameters of the model by reducing complexity, number of calculations, and computational operations. This parameter optimization transforms the model from a computationally intensive system into a lightweight system that can process predictions rapidly on embedded aircraft hardware with limited processing power.
3Quantity of substance
If a simplified model is used to reduce memory and computing resources, then resource efficiency improves, but prediction accuracy may deteriorate
Solution Approach 1:
The training process uses feedback from the conventional model's predictions to continuously improve the surrogate model's accuracy. During training, the surrogate model's outputs are compared against the conventional model's outputs, and adjustment algorithms use this feedback to minimize prediction errors, ensuring the simplified model maintains high accuracy despite reduced complexity.
Solution Approach 2:
The patent performs preliminary training of the surrogate model using extensive datasets and the conventional model as a reference before deployment. This preliminary action ensures the simplified model is pre-calibrated to match the conventional model's predictive behavior across various combat scenarios, so no accuracy is lost when the simplified model is used in operational conditions.
Data Source
AI summary
An example includes a method of training a first computational model to emulate a second computational model. The method includes using the first computational model to generate a first output in response to receiving an input and selecting a reward based on whether a difference between the first output and a second output is less than a threshold. The second output is generated by the second computational model in response to receiving the input. The method further includes updating the first computational model using the reward.