Neural Network Terrain Database Encoding for Aircraft
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contemporary terrain databases used in aircraft systems are large and require significant storage space, leading to inefficiencies in loading and retrieval, especially in legacy systems.
Innovation Solution
A system and method utilizing a neural network to efficiently generate and retrieve terrain elevation data by inputting queries and base resolution elevation data, allowing for high-speed loading and reduced storage needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional compression methodologies are used to store terrain database, then storage capacity is reduced, but data loading speed and retrieval efficiency remain insufficient
Solution Approach 1:
The patent replaces conventional mechanical compression methodologies with a neural network-based encoding system. The neural network learns compressed representations of terrain data during training, enabling rapid decomposition and high-speed loading during operation without requiring traditional compression/decompression processes.
Solution Approach 2:
The patent transforms terrain database storage from conventional compressed formats to a neural network parameter space. By encoding terrain data as neural network weights and activations, the system achieves both compact storage and rapid retrieval through the network's inherent parallel processing capabilities.
2Quantity of substance
If conventional compression methodologies are used for terrain database, then storage space is reduced, but dedicated hardware is still required for data storage and retrieval
Solution Approach 1:
The neural network serves multiple functions: it acts as both the compression engine during data preparation and the retrieval system during operation. This eliminates the need for separate dedicated hardware components, as standard processors can execute the neural network for both encoding and decoding terrain data.
Solution Approach 2:
The patent replaces dedicated hardware retrieval systems with a software-based neural network approach. The neural network can be implemented on general-purpose processors, eliminating the need for specialized hardware while maintaining efficient data access.
3Loss of information
If evenly-gridded array of elevation posts is used, then terrain data is complete, but storage capacity consumes multiple gigabytes
Solution Approach 1:
The neural network extracts the essential features and patterns from the complete evenly-gridded terrain data during training. By learning the underlying structure of terrain elevation patterns, the network can reconstruct accurate terrain information from a compressed representation, retaining only the most critical data elements.
Solution Approach 2:
The patent performs preliminary compression and pattern recognition during the neural network training phase. By pre-processing the complete terrain data to create optimized network weights and structures, the system achieves both data completeness and storage efficiency when the trained network is deployed.
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
A system is disclosed. The system may include a display (112) and one or more controllers (102) communicatively coupled to the display. The one or more controllers may include one or more processors (104) configured to execute a set of program instructions stored in a memory (106). The set of program instructions may be configured to cause the one or more processors to receive a neural network (300) configured to output elevation data based on a plurality of queries. Each query (204) may correspond to a queried location of a terrain area. The one or more processors may be configured to input each query to the neural network to output the elevation data corresponding to the queried location. The one or more processors may be configured to direct the elevation data (206) to be displayed on the display.