Autonomous Driving Vehicle Data Collection Automation System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The inefficiency of current methods for collecting training data for autonomous driving vehicles, which requires professional knowledge and human intervention, leading to a cumbersome and inefficient data collection process.
Innovation Solution
A computer-implemented method in an autonomous driving vehicle that receives instructions to collect driving data for specific categories, logs and transmits data to a server for machine learning, and updates a user interface to indicate progress, allowing the vehicle to autonomously collect and upload data for training self-driving models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an engineer accompanies the human driver to guide data collection, then the quality and completeness of collected data is improved, but the complexity and time consumption of the data collection process increases
Solution Approach 1:
The system enables the autonomous vehicle to automatically identify and collect required training data without engineer intervention. The vehicle self-determines what data is needed based on its autonomous driving model requirements and autonomously collects data during manual driving modes, eliminating the need for engineer guidance while maintaining data quality.
Solution Approach 2:
The system pre-configures data collection parameters and requirements before the actual data collection process. The autonomous driving model预先 determines what types of data are needed for training, and the system prepares the necessary sensors and logging mechanisms in advance, so that data collection can proceed automatically without real-time engineer intervention.
2Reliability
If an engineer accompanies the human driver to guide data collection, then the completeness of collected data scenarios is improved, but the time consumption and efficiency of data collection worsens
Solution Approach 1:
The autonomous vehicle automatically identifies which driving scenarios require data collection and guides the manual driving process to capture necessary scenarios. The system monitors collected data completeness and automatically determines when sufficient data has been gathered, eliminating the time-consuming back-and-forth between engineer and driver.
Solution Approach 2:
The system continuously monitors the data collection process and provides real-time feedback to both the vehicle system and the human driver about what data is being collected and what is still needed. This feedback mechanism ensures complete scenario coverage while allowing parallel data collection across multiple vehicles, significantly improving overall efficiency.
3Reliability
If professional knowledge of machine learning and ADV control systems is required for data collection, then the quality of training data is improved, but the ease of operation of data collection deteriorates
Solution Approach 1:
The system encapsulates all professional knowledge about data collection requirements within the autonomous driving model and system architecture. The vehicle automatically determines what data is needed based on its own model requirements, eliminating the need for operators to possess specialized machine learning knowledge. Any person can operate the vehicle for data collection following simple system instructions.
4Measurement precision
If manual driving mode data collection is guided by an engineer, then the accuracy of data labeling is improved, but the cost and complexity of the system increases
Solution Approach 1:
The system automatically labels collected data using the autonomous driving model's perception and interpretation of the driving environment. The vehicle's existing sensors and processing systems generate labels for training data without requiring separate manual annotation processes or expert reviewers, maintaining labeling accuracy while dramatically reducing system complexity.
Data Source
AI summary
An autonomous driving vehicle (ADV) receives instructions for a human test driver to drive the ADV in manual mode and to collect a specified amount of driving data for one or more specified driving categories. As the user drivers the ADV in manual mode, driving data corresponding to the one or more driving categories is logged. A user interface of the ADV displays the one or more driving categories that the human driver is instructed collect data upon, and a progress indicator for each of these categories as the human driving progresses. The driving data is uploaded to a server for machine learning. If the server machine learning achieves a threshold grading amount of the uploaded data to variables of a dynamic self-driving model, then the server generates an ADV self-driving model, and distributes the model to one or more ADVs that are navigated in the self-driving mode.