Autonomous Driving Data Pipeline Using Filtered Sensor Components
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning systems for autonomous driving face challenges in maximizing signal information from sensor data, as traditional conversion processes can reduce signal fidelity and require new conversion methods for different sensors, leading to a need for a customized data pipeline that enhances signal information for deep learning analysis.
Innovation Solution
A data pipeline that extracts and processes sensor data into separate components, such as feature and global data, using filters like high-pass, low-pass, and band-pass, and provides these components to different layers of a deep learning network, optimizing signal information and computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional conversion processes are used to make sensor data compatible with deep learning systems, then data format compatibility is achieved, but signal fidelity is reduced
Solution Approach 1:
The patent segments the conversion process into multiple specialized components: a first conversion process handles first sensor data with first signal fidelity, while a second conversion process handles second sensor data with second signal fidelity. This segmentation allows each conversion process to be optimized for its specific sensor type, preventing universal conversion losses.
Solution Approach 2:
The patent applies local quality by making conversion processes sensor-specific rather than universal. Each sensor type receives a customized conversion process tailored to its characteristics, ensuring that the conversion maintains the specific signal qualities and features relevant to that sensor type while achieving format compatibility.
2Productivity
If sensor data is compressed and down-sampled for deep learning input, then computational efficiency is improved, but signal information is lost
Solution Approach 1:
The patent performs preliminary action by extracting relevant features and converting sensor data to appropriate representations before the deep learning processing stage. This pre-processing prepares the data in an optimized format that maintains essential signal information while reducing unnecessary data, thereby improving computational efficiency without significant information loss.
3Loss of information
If a customized data pipeline is implemented to maximize signal information, then signal fidelity is improved, but system complexity increases
Solution Approach 1:
The patent achieves universality by creating a multi-functional data pipeline that handles multiple sensor types through a unified architecture. The system can process first sensor data, second sensor data, and other sensor data through coordinated conversion processes, maintaining signal fidelity across different sensor types while providing a standardized interface to the deep learning system.
4Measurement precision
If different conversion processes are used for different sensors, then sensor-specific signal quality is maintained, but processing complexity increases
Solution Approach 1:
The patent segments the conversion system into distinct, specialized processes for different sensor types. Each sensor type has its own optimized conversion process that maintains sensor-specific signal quality, while the overall system coordinates these segmented processes through a unified data pipeline architecture.
Data Source
AI summary
An image captured using a sensor on a vehicle is received and decomposed into a plurality of component images. Each component image of the plurality of component images is provided as a different input to a different layer of a plurality of layers of an artificial neural network to determine a result. The result of the artificial neural network is used to at least in part autonomously operate the vehicle.