Autonomous Vehicle Simulation Using Simplified Scenario Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle testing methods are inefficient due to the time-consuming and costly collection of real sensor data, and simulated sensor data fails to accurately account for noise inherent in driving scenarios, limiting the evaluation of autonomous vehicle systems.
Innovation Solution
A computing system utilizing a machine-learned perception-prediction simulation model processes simplified scenario data to generate simulated perception and prediction outputs, which are then used to optimize motion planning systems, effectively simulating various driving scenarios without the need for real sensor data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real sensor data collection is used for autonomous vehicle testing, then measurement precision and reliability are improved, but loss of time and productivity deteriorate due to time-consuming and costly data collection
Solution Approach 1:
The patent creates simplified scenario data as a copy or abstraction of complex real sensor data. This simplified data captures essential driving scenarios and noise characteristics without requiring actual sensor collections, thereby maintaining measurement precision while dramatically reducing time loss.
Solution Approach 2:
The patent transforms the data representation by changing parameters from raw sensor formats to simplified scenario formats that preserve critical information. This parameter transformation allows efficient processing and testing while maintaining the essential characteristics needed for accurate autonomous vehicle system evaluation.
2Productivity
If simulated sensor data is used for autonomous vehicle testing, then productivity is improved, but measurement precision deteriorates due to failure to account for noise inherent in driving scenarios
Solution Approach 1:
The patent applies local quality by selectively incorporating noise characteristics only where they are most relevant to testing autonomous vehicle systems. Rather than uniformly adding noise throughout all data, the simplified scenario data includes noise factors like occlusion and false detections at specific locations and intensities that mirror real driving conditions, thus maintaining measurement precision while preserving productivity.
3Productivity
If simplified scenario data is processed through machine-learned perception-prediction simulation model, then productivity and measurement precision are improved, but device complexity increases
Solution Approach 1:
The patent segments the autonomous vehicle system into distinct functional components (perception system, prediction system, motion planning system) that can be independently tested using the simplified scenario data. This segmentation allows the complex simulation system to be broken down into manageable modules, reducing overall device complexity while maintaining high productivity and measurement precision.
Data Source
AI summary
Systems and methods of the present disclosure are directed to a method. The method can include obtaining simplified scenario data associated with a simulated scenario. The method can include determining, using a machine-learned perception-prediction simulation model, a simulated perception-prediction output based at least in part on the simplified scenario data. The method can include evaluating a loss function comprising a perception loss term and a prediction loss term. The method can include adjusting one or more parameters of the machine-learned perception-prediction simulation model based at least in part on the loss function.