Autonomous Vehicle Simulation Data from Logged Driving Scenarios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicle technology faces challenges in acquiring sufficient and high-quality training data to accurately represent various driving conditions and scenarios, as existing methods, such as simulation data from video game-like simulators, fail to provide realistic representations of real-world driving.
Innovation Solution
The method involves receiving logged data from autonomous vehicles, generating augmented data that describes actors in the vehicle's environment, and creating simulation scenarios based on this augmented data to train machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If simulation data from video game-like simulators is used to train machine learning models, then the quantity of training data is increased, but the quality and realism of the training data deteriorates
Solution Approach 1:
The patent uses real-world logged sensor data from autonomous vehicles as a template to copy and recreate realistic driving scenarios in simulation. Instead of using fictional video game data, the system captures actual sensor readings, actor positions, and environmental conditions from real vehicles and uses this authentic data to populate simulation environments, thereby maintaining high realism while generating abundant training data
Solution Approach 2:
The system performs preliminary data collection and processing in the real world by logging sensor data, actor information, and driving conditions from autonomous vehicles before deploying these scenarios to simulation. This preliminary action of capturing real-world data upfront allows the simulation to inherit authentic characteristics without requiring continuous real-world data collection during simulation operations
2Measurement precision
If more training data is collected from real autonomous vehicle operation, then the quality and realism of training data is improved, but the quantity of data remains insufficient
Solution Approach 1:
The system dynamically generates multiple variations of training scenarios by modifying parameters such as actor types, environmental conditions, weather, and traffic patterns based on the core real-world logged data. This dynamic generation allows a single real-world driving episode to produce numerous diverse training examples, exponentially increasing data quantity while preserving the authenticity of the underlying real-world scenario
Solution Approach 2:
The simulation platform serves multiple functions: it acts as both a data generation engine and a test environment. The same simulation infrastructure that recreates real-world scenarios can also systematically vary parameters to generate edge cases and rare events that would be difficult to capture in normal operation, making the system universally applicable for generating diverse training data
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Logged data from an autonomous vehicle is processed to generate augmented data. The augmented data describes an actor in an environment of the autonomous vehicle, the actor having an associated actor type and an actor motion behavior characteristic. The augmented data may be varied to create different sets of augmented data. The sets of augmented data can be used to create one or more simulation scenarios that in turn are used to produce machine learning models to control the operation of autonomous vehicles.