Aircraft Control Authority Transfer Using a Trained Classifier
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Aircraft control systems face challenges in transitioning from established control modules to new trained classifiers, requiring assurance of safety and performance, as pilots and regulatory bodies are hesitant to adopt unproven AI systems due to concerns over reliability and safety.
Innovation Solution
An aircraft control system that gradually transfers control from a traditional control module to a trained classifier by using an authority parameter, which is updated based on performance evaluation, allowing incremental authority transfer and ensuring the system's reliability and safety through weighted averages of control outputs from both modules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a new trained classifier is introduced to control the aircraft, then the productivity and adaptability of the control system is improved, but the reliability and safety are compromised due to lack of proven track record
Solution Approach 1:
The system dynamically adjusts the authority parameter over time, transitioning from a static control system to one where the controller's authority is progressively increased based on performance evaluation. The authority parameter evolves from an initial low value to potentially full authority, allowing the system to adapt its reliability characteristics as the new controller proves itself through operation.
Solution Approach 2:
The system implements continuous feedback through performance monitoring and evaluation mechanisms. The authority parameter is updated based on evaluated performance data, creating a closed-loop system where the controller's authority is adjusted according to its demonstrated reliability. This feedback mechanism allows the system to maintain safety while progressively incorporating the new trained classifier.
2Adaptability or versatility
If control is fully transferred to the trained classifier, then the adaptability and performance are improved, but the risk and uncertainty increase due to pilot hesitation and regulatory concerns
Solution Approach 1:
The authority parameter provides dynamic control flexibility, allowing the system to adapt between fully manual and fully automated operation modes. This dynamic adjustment capability addresses pilot hesitation by providing a gradual transition path rather than an abrupt change, while still achieving full adaptability when the trained classifier proves reliable.
Solution Approach 2:
The system changes the authority parameter based on performance evaluation results, transitioning from conservative parameter values to more aggressive ones as the trained classifier demonstrates reliability. This parameter change approach systematically reduces safety concerns by requiring proven performance before increasing automation authority.
3Adaptability or versatility
If the authority parameter is updated frequently based on performance evaluation, then the adaptability and responsiveness are improved, but the system complexity increases
Solution Approach 1:
The system implements dynamic updating of the authority parameter through performance evaluation, allowing responsive adaptation without requiring complex reconfiguration. The evaluation mechanism processes performance data and adjusts the authority parameter accordingly, providing responsiveness while maintaining relatively simple system architecture through standardized evaluation procedures.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is an aircraft control system (100) comprising an aircraft control module (110), a trained classifier module (120), and an aircraft control processing engine (13). The aircraft control module (110) generates first control outputs (104a to 104c) based on received aircraft operating inputs (102a to 102d). The trained classifier module receives the aircraft operating inputs (102a to 102d) and generates second control outputs (104d to 104f). The aircraft control processing engine (130) receives the first control outputs (104a to 104c) and the second control outputs (104d to 104f) and generates operating control outputs (106a to 106c), based on the received first control outputs and second control outputs. The aircraft control processing engine (130) then controls the aircraft using the operating control outputs (106a to 106c). The aircraft control system (100) may be comprised in a test aircraft (500). Also disclosed is a test system for an aircraft (500) comprising a plurality of interfaces (510, 520, 530, 540), a processor (550), and at least one memory (560).