Log-Based Autonomous Driving Simulation for Rare Scenario Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for evaluating autonomous vehicle control software are limited by the scarcity of specific scenarios, such as collisions or near collisions, which hinders thorough testing and requires manual creation of simulations, making the process time-consuming and resource-intensive.
Innovation Solution
A method that identifies initial situations from log data or simulated scenarios and generates new simulations by inserting objects into similar situations, using a neutral frame of reference and interpolating characteristics to create realistic scenarios, thereby expanding the range of test cases without extensive human operator involvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual creation of simulations is used to ensure scenario diversity, then the variety of test cases improves, but the time and resource consumption increases significantly
Solution Approach 1:
The patent creates synthetic simulations by copying and modifying existing real-world log data. Instead of manually creating diverse scenarios, the system takes actual driving logs and synthetically generates new simulation scenarios by inserting objects, modifying parameters, and recombining data elements. This automated copying approach maintains scenario diversity while eliminating manual creation time.
Solution Approach 2:
The system performs self-service by automatically generating its own test scenarios from available log data without requiring manual intervention. The simulation generation process is automated through algorithms that selectively combine log entries, insert objects, and adjust parameters, allowing the system to serve its own testing needs without external human effort.
2Reliability
If extensive manual iteration is performed to create realistic scenarios, then the realism of simulations improves, but the resource expenditure increases
Solution Approach 1:
The patent replaces the mechanical process of manual scenario creation with an automated computational system. Instead of human operators manually designing and adjusting simulation parameters, the system uses algorithms to automatically generate realistic scenarios from log data, substituting human mechanical effort with automated processing that is both faster and more consistent.
Solution Approach 2:
The system performs preliminary actions by pre-processing and organizing log data into reusable components before simulation generation. It identifies and extracts relevant driving scenarios, objects, and parameters from raw logs in advance, preparing them for automated assembly into realistic test simulations. This preliminary organization enables efficient generation without requiring resource-intensive manual iteration during the actual simulation creation phase.
3Productivity
If the same limited set of scenarios is reused for testing, then the resource expenditure decreases, but the thoroughness of software evaluation deteriorates
Solution Approach 1:
The patent introduces dynamics by enabling the scenario set to adapt and expand automatically based on testing needs. Instead of using a static, limited scenario collection, the system dynamically generates new scenarios from available log data, continuously expanding the test coverage. This dynamic approach allows the same resource base (log data) to produce an ever-growing variety of test scenarios without requiring additional manual resources.
Solution Approach 2:
The system achieves universality by creating a multi-functional scenario generation platform that can produce diverse test cases from a single source (log data). The same log database serves multiple purposes: it provides the foundation for generating various scenario types, supplies objects for insertion, and offers parameter ranges for variation. This universal approach allows comprehensive software evaluation using a single resource base rather than requiring separate resources for each scenario type.
Data Source
AI summary
The technology relates to generating simulations in order to evaluate software used to control vehicles in an autonomous driving mode. In one example, an initial situation involving a certain kind of interaction between a vehicle operating in the autonomous driving mode and an object may be identified. A search of log data may be conducted in order to identify one or more similar situations based on characteristics of the initial situation. A new simulation may be generated using the identified one or more similar situations by inserting the object into the one or more similar situations. The new simulation may be run in order to evaluate the software.