ADS Scenario Description Generation Under Bandwidth and Privacy Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ADS solutions face challenges with data collection from a fleet of vehicles due to bandwidth, storage, and data privacy constraints, making it difficult to maintain adequate performance over time and comply with regulatory frameworks.
Innovation Solution
A method and system for generating textual scenario descriptions from sensor data using embedding and description generator networks, allowing efficient data collection and transmission while preserving privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If sensor data from an entire fleet of vehicles is collected and transmitted to a centralized server, then the amount of data available for training ADS models increases, but storage limitations and bandwidth limitations are exceeded
Solution Approach 1:
The patent extracts only the essential information from raw sensor data by using scenario triggers to identify critical events and generating concise textual descriptions of these scenarios. This extraction process removes unnecessary data while preserving the valuable information needed for ADS development, thereby reducing storage and bandwidth requirements while maintaining data quality.
Solution Approach 2:
Instead of transmitting original sensor data, the system creates textual copies or representations of the scenarios. These textual descriptions serve as simplified replicas that capture the essential information needed for model training without requiring the full original data, thus reducing the data transmission burden while preserving training value.
2Quantity of substance
If sensor data from fleet vehicles is collected and stored, then more training data is available for ADS development, but data privacy concerns and regulatory restrictions are violated
Solution Approach 1:
The system extracts only the necessary scenario information needed for ADS training while leaving behind sensitive personal data. By focusing on extracting scenario triggers and generating textual descriptions of driving events, the system obtains training data without capturing identifiable personal information, thus resolving the privacy conflict.
Solution Approach 2:
The patent uses disposable textual representations of scenarios instead of retaining original sensor data. These textual descriptions serve their purpose for training purposes and can be discarded afterward, whereas original sensor data containing personal information would require long-term secure storage and protection. This approach minimizes privacy risks while maintaining training effectiveness.
3Loss of information
If all sensor data is transmitted to a centralized server, then complete scenario information is available for analysis, but bandwidth limitations and transmission costs increase
Solution Approach 1:
The system extracts only the critical scenario information needed for ADS development by monitoring scenario triggers and generating concise textual descriptions. This extraction eliminates unnecessary data transmission while preserving the essential scenario information required for model training and analysis, thereby reducing bandwidth consumption without sacrificing information quality.
4Adaptability or versatility
If sensor data from production vehicles is collected, then real-world scenario diversity increases for ADS training, but post-processing requirements and computational resources increase
Solution Approach 1:
The system performs preliminary processing at the vehicle端 by monitoring scenario triggers and generating textual descriptions before data leaves the vehicle. This preliminary action filters and structures the data in advance, reducing the computational burden on centralized servers and simplifying subsequent post-processing steps while maintaining scenario diversity from the production fleet.
Data Source
AI summary
A method for providing data-insight for development of an automated driving system (ADS) of a vehicle. The method includes: obtaining sensor data pertaining to a driving scenario, the sensor data being captured by one or more sensors of the vehicle and depicting at least part of a surrounding environment of the vehicle; monitoring a fulfillment of one or more scenario triggers of the driving scenario, wherein the fulfillment is indicative of the driving scenario being a driving scenario of interest; in response to determining at least one of the one or more scenario triggers being fulfilled: generating a scenario description, based on the obtained sensor data pertaining to the driving scenario and/or based on ADS data outputted from the ADS having processed the sensor data, wherein the scenario description includes textual data about the driving scenario to which the obtained sensor data pertains; and storing the generated scenario description.