Neural Network Encoder-Decoder for Robust Guideway Marker Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication train based control (CTBC) systems in guideway mounted vehicles face issues with robustness and integrity of instruction communication due to interruptions, incorrect information, or rejection of instructions, leading to vehicle braking and loss of control.
Innovation Solution
A neural network encoder-decoder system trained with an environmental filter to encode and decode data embedded in markers along the guideway, simulating various environmental conditions to enhance accuracy and integrity of data transmission in harsh conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional CTBC communication systems are used to transmit movement instructions, then the system can provide basic control functionality, but the robustness and integrity of instruction communication deteriorates due to interruptions, incorrect information, or rejection of instructions
Solution Approach 1:
The system performs preliminary encoding of movement instructions using a neural network encoder before transmission. This pre-encoding process transforms the original instructions into a robust encoded format that can withstand communication interruptions and corruption, allowing the decoder to accurately reconstruct the original instructions even when transmission conditions are adverse.
Solution Approach 2:
The patent introduces an intermediary encoding-decoding mechanism using neural networks between the control system and the vehicle. This intermediary layer processes the communication data through trained neural network models that can handle noise, interruptions, and corruption, thereby protecting the integrity of the original movement instructions during transmission.
2Reliability
If neural network encoding is implemented to improve data transmission robustness, then communication reliability improves, but system complexity increases due to the addition of encoder and decoder components
Solution Approach 1:
The patent uses neural network encoders and decoders that can be trained and deployed as software models on existing hardware platforms. Rather than requiring completely new hardware systems, the solution copies the intelligence of neural networks into trainable models that run on standard computing devices, thereby reducing the physical complexity increase while maintaining the reliability benefits.
3Measurement precision
If environmental filtering is applied during training to simulate harsh conditions, then decoding accuracy in real-world environments improves, but training time and computational resources increase
Solution Approach 1:
The system performs preliminary training with environmental filters that simulate various harsh conditions (noise, interruptions, corruption) before deployment. This pre-training process prepares the neural network decoder to handle real-world adverse conditions, reducing the need for extensive fine-tuning after deployment and thereby offsetting the initial training time investment with long-term operational efficiency.
Data Source
AI summary
A system includes a neural network encoder, an environmental filter and a neural network decoder. The neural network encoder is configured to generate encoded data from input data. The environmental filter is communicably connected with the encoder and configured to combine the encoded data with at least one randomized image to generate signature data corresponding to the input data. The neural network decoder is configured to be trained together with the encoder and the environmental filter to decode the signature data to generate decoded data corresponding to the input data.