Digital Map Creation Using Generative Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating digital maps for automated vehicles are limited by the need for extensive data collection and are prone to errors caused by sensor noise and algorithmic processing.
Innovation Solution
A method using generative neural networks to create perturbed images of surrounding areas, which are then merged with original images using a further neural network to generate highly accurate digital maps suitable for automated vehicle navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive real data collection is performed to create digital maps, then map accuracy and completeness are improved, but data collection effort and time are significantly increased
Solution Approach 1:
The system performs preliminary actions by collecting and storing diverse surrounding area images (from multiple vehicles, times, weather conditions) in advance. These pre-collected images form a training dataset that can be used to train neural networks for map creation, eliminating the need for extensive real-time data collection when actual digital maps need to be generated.
Solution Approach 2:
The system creates synthetic copies of real surrounding area images by generating perturbed versions through neural networks. These synthetic images simulate various sensor noise conditions and environmental variations without requiring actual physical data collection under those specific conditions, thus reducing data collection effort while maintaining map accuracy.
2Speed
If traditional scan matching methods are used to merge surrounding area images, then processing speed is maintained, but errors from sensor noise and algorithmic processing increase
Solution Approach 1:
The system replaces traditional mechanical scan matching algorithms with a neural network-based approach. The neural network learns optimal merging strategies from training data, substituting deterministic algorithmic processing with learned probabilistic processing that can better handle sensor noise and produce more reliable digital maps while maintaining processing speed.
Solution Approach 2:
The system implements feedback by training the neural network on pairs of original and perturbed images with known ground truth relationships. The network learns to compensate for various types of disturbances by comparing predicted merged results with actual outcomes, continuously improving its ability to produce accurate digital maps despite sensor noise.
3Quantity of substance
If perturbed images are generated using generative neural networks, then dataset size is increased and model robustness is improved, but computational complexity and training time are increased
Solution Approach 1:
The generative neural network serves multiple functions: it generates perturbed images for data augmentation, simulates various sensor noise conditions, and creates training samples for different environmental scenarios. This single multi-functional component replaces what would otherwise require multiple separate data collection campaigns under different conditions, reducing overall computational complexity despite the added capability.
Data Source
AI summary
A method and device for creating a digital map. The method includes: receiving first training data sets; generating second training data sets by means of a generative neural network by determining a second, perturbed image of the same surrounding area in each case for each first image, wherein each second training data set represents the second image of the corresponding first image; training a further neural network; receiving surrounding area data sets, which in each case represent a surrounding area image of a vehicle surrounding area, wherein these surrounding area data sets comprise a position description of the corresponding vehicle surrounding area; creating the digital map by merging the surrounding area images, depending on the position description, by means of a scan matching method, wherein the scan matching method includes at least the further neural network, and a step of providing the digital map.