Neural Network HD Map Learning for Road User Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost and complexity of creating highly accurate high-definition maps limit their availability and practical use for predicting the future positions of road users, especially for pedestrians and cyclists, making it difficult to effectively warn vulnerable road users.
Innovation Solution
A system and method that uses historical sensor data to learn a sufficiently accurate high-definition map through machine learning, specifically by training an artificial neural network with historical movement data, allowing for the prediction of future road user positions without the need for expensive manual map creation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to predict road user positions, then prediction quality can be achieved, but high-definition maps are required which are expensive and manually created
Solution Approach 1:
The patent creates a simplified representation (copy) of the complex HD map by training a neural network on historical sensor data. Instead of using expensive manually created HD maps, the system learns map information implicitly from sensor data, creating a digital model that captures essential road geometry and lane configuration without requiring manual annotation.
Solution Approach 2:
The system uses historical sensor data from sensors already deployed in the environment to train the neural network. The neural network learns map information autonomously from the sensor data patterns, eliminating the need for external manual map creation services and making the system self-sufficient.
2Measurement precision
If HD maps are made available to autonomous vehicles, then navigation accuracy is improved, but the maps are not available for pedestrians and cyclists
Solution Approach 1:
The neural network model serves multiple functions: it predicts positions for different types of road users (vehicles, pedestrians, cyclists), generates HD map data, and provides lane configuration information. The single trained model adapts to different road user types by processing their specific movement patterns from sensor data, making the system universally applicable.
3Manufacturing precision
If manual HD map creation is used, then centimeter accuracy is achieved, but the process is time-consuming and expensive
Solution Approach 1:
The patent replaces the manual mechanical process of creating HD maps with an automated computational approach. Instead of manually annotating maps, the system uses machine learning algorithms to automatically extract map information from historical sensor data, significantly increasing productivity while maintaining precision.
Solution Approach 2:
The system performs preliminary processing by training the neural network on historical sensor data before actual prediction is needed. This pre-training phase allows the model to learn map patterns and road geometry in advance, enabling fast and accurate predictions without requiring manual map creation for each new scenario.
Data Source
AI summary
The present disclosure relates to a system for predicting a future position of road users in a predefined road section, including a storage unit in which at least historical sensor data relating to captured road users in the predefined road section are stored. A processor is provided and extracts the historical movement data relating to the respective road users from the historical sensor data and is further designed to learn a sufficiently accurate high-definition map of the predefined road section by means of a machine learning method based on the historical movement data. The processor is further designed to determine, based on an input of current movement data relating to a road user, a prediction of at least one future position of this road user in this road section, at least implicitly using the learned high-definition map.