Aircraft Dynamics Prediction With Compressive State Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for predicting the future trajectory of aircraft rely on physics-based dynamic equations, which are not suitable for aircraft autonomy due to the requirement of a pre-existing model and unavailable state variables, and existing learning-based methods fail to make accurate predictions, especially in adversarial scenarios.
Innovation Solution
A machine learning method using a neural network with a feedforward skip connection and compressive encoding scheme that learns to predict aircraft trajectories based on current and past state observations, handling angular variables with a special residual method and applying a loss function only to the residual output to address numerical conditioning challenges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based dynamic equations are used to predict aircraft trajectory, then prediction accuracy is improved, but the method requires pre-existing models and unavailable state variables which reduces adaptability
Solution Approach 1:
The patent creates a virtual copy of the aircraft dynamics through a trained neural network model that replicates the behavior of physics-based models without requiring their complex equations. This copied model can predict trajectories accurately while adapting to different aircraft types through training data, resolving the contradiction between accuracy and adaptability
Solution Approach 2:
The patent replaces the mechanical physics-based dynamic equations with a data-driven neural network system. This substitution eliminates the need for pre-existing models and unavailable state variables while maintaining prediction accuracy, as the neural network learns dynamics patterns from training data without requiring explicit physical equations
2Device complexity
If compressive encoding is applied to reduce state representation, then model complexity is reduced, but information loss may occur
Solution Approach 1:
The patent introduces an intermediary compressive encoding layer that transforms the full state representation into a compressed form while preserving essential dynamics information. This intermediary representation serves as a bridge between the original state space and the prediction model, reducing complexity while maintaining predictive accuracy through learned feature extraction
Solution Approach 2:
The patent extracts only the most relevant features from the full state representation through compressive encoding, removing redundant information while retaining critical dynamics. This extraction process reduces model complexity by focusing on essential state variables that most influence aircraft behavior
Data Source
AI summary
Training a compressive encoder is provided. The method comprises calculating a difference between a current state of an aircraft and a previous state. The current state comprises a reduced state wherein the value of a specified parameter is missing. The difference is input into compressive layers of a neural network comprising an encoder. The compressive layers learn, according to the difference, a value for the missing parameter. The current state and learned value are concurrently fed into hidden layers of a fully connected neural network comprising a decoder. An action applied to the aircraft is input into the hidden layers concurrently with the current state and learned value. The hidden layers, according to the current state, learned value, and current action, determine a residual output that comprises an incremental difference in the state of the aircraft resulting from the current action.