Vehicle Onboard Storage Control Using ML-Based Data Retention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vehicles, such as aircraft, face challenges in storing and retaining operational data due to limited onboard storage resources, where data is often overwritten or lost due to dynamic operational conditions and the difficulty in determining what data is noteworthy, leading to overinclusion or underinclusion issues with traditional threshold-based storage schemes.
Innovation Solution
A vehicle monitoring system utilizing machine-learning techniques to monitor real-time environmental conditions, determine expected values of environmental variables, and generate intelligent storage decisions based on these evaluations, ensuring that more significant data is retained while optimizing onboard storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional threshold-based storage schemes are used, then storage decisions are simple to implement, but data overinclusion or underinclusion occurs leading to loss of important information
Solution Approach 1:
The patent replaces traditional mechanical threshold-based storage decision systems with a machine-learning-based intelligent system. The machine-learning model analyzes multiple environmental variables and their correlations to dynamically determine storage decisions, substituting simple threshold comparisons with sophisticated predictive analytics that adapt to changing operational conditions, thereby reducing information loss without requiring complex manual configuration
Solution Approach 2:
The machine-learning model continuously learns from incoming sensor data and automatically adjusts its predictions and storage decisions without external intervention. The system self-optimizes by identifying patterns in the data stream and autonomously determining which data points are noteworthy based on learned correlations between environmental variables, eliminating the need for manual threshold tuning and adapting to new operational scenarios automatically
2Reliability
If all sensor data is retained, then complete operational record is maintained, but onboard storage resources are quickly exhausted
Solution Approach 1:
Instead of retaining all sensor data or none, the patent applies partial action by selectively storing only those data points that the machine-learning model predicts are noteworthy or anomalous. The system performs a subset of storage actions based on predictive analysis, storing approximately 20% of data points that deviate from expected patterns while discarding redundant normal operations, thereby maintaining reliability for critical events while conserving storage capacity
Solution Approach 2:
The patent dynamically changes the storage parameter from fixed threshold-based retention to adaptive machine-learning-based selection. The system transforms the storage decision parameter from a static value to a dynamic prediction that changes based on learned patterns in environmental variable correlations, allowing the storage system to adapt its retention criteria based on the actual operational context and significance of data points
3Measurement precision
If machine-learning model is used for storage decisions, then data retention accuracy is improved, but computational resources and processing time increase
Solution Approach 1:
The machine-learning model is pre-trained offline using historical sensor data to learn correlations between environmental variables and noteworthy events. This preliminary action transfers computational burden from the resource-constrained vehicle system to the training environment, allowing the deployed model to make accurate predictions with minimal real-time computation, thereby achieving high measurement precision while consuming limited onboard energy
Solution Approach 2:
The patent introduces a machine-learning model as an intermediary layer between raw sensor data and storage decisions. This intermediary processes data using learned patterns rather than raw computational power, acting as a smart filter that translates environmental variable correlations into storage decisions. The intermediary reduces the computational energy required for real-time analysis while maintaining high accuracy in identifying noteworthy data points
4Productivity
If dynamic storage decisions are implemented, then important data is prioritized for retention, but system complexity and difficulty of implementation increase
Solution Approach 1:
The machine-learning model serves multiple functions simultaneously: it predicts noteworthy data points, identifies anomalies, learns from historical patterns, and generates storage decisions. This universal approach consolidates what would otherwise require multiple separate systems into a single multi-functional model, improving storage utilization efficiency while managing complexity through functional integration rather than proliferation of separate components
Data Source
AI summary
In an embodiment, a method includes monitoring environmental conditions via a plurality of sensors positioned in proximity to a vehicle. The method also includes receiving, from one or more of the plurality of sensors, a first measured value of a first environmental variable and at least one additional measured value of at least one additional environmental variable. In addition, the method includes automatically determining, via a machine-learning model, a first expected value of the first environmental variable given the at least one additional measured value of the at least one additional environmental variable. Furthermore, the method includes automatically generating a storage decision in relation to the first measured value based, at least in part, on an evaluation of the first measured value relative to the first expected value. Additionally, the method includes causing execution of the storage decision at least with respect to onboard storage of the vehicle.