LIDAR Scenario Generation Using Merged Point Clouds and Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current approaches for autonomous driving training and testing are limited by the lack of sufficient and varied driving scenarios, making it difficult to statistically prove the safety of autonomous vehicles compared to human drivers, and existing simulation methods are cumbersome and limited to previously encountered scenarios.
Innovation Solution
A computer-implemented method that generates simulation scenarios by merging LIDAR point clouds, camera images, and velocity/acceleration data into a common coordinate system, classifying static and dynamic objects, and generating trajectories for dynamic road users, allowing for the creation and export of simulation scenarios that can be used to test autonomous driving functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-world road tests are conducted to prove safety, then statistical proof of safety can be obtained, but the time and cost required are excessively long and high
Solution Approach 1:
The patent creates virtual copies of real-world driving scenarios by processing recorded sensor data (LIDAR point clouds, camera images, velocity/acceleration data) into simulation scenarios. These virtual replicas allow repeated testing without additional real-world time consumption, enabling statistical safety proof through virtual replication rather than physical repetition.
Solution Approach 2:
The system performs preliminary processing of raw sensor data into structured simulation scenarios that can be reused multiple times. By pre-processing and organizing the data into reusable scenario formats during initial data collection, the system eliminates the need to repeat real-world data collection for each test iteration, significantly reducing time loss.
2Adaptability or versatility
If real-world road tests are conducted to gather sufficient training data, then diverse driving scenarios can be captured, but the cost and resources required are excessively high
Solution Approach 1:
The patent creates virtual copies of real-world driving scenarios by processing recorded sensor data (LIDAR point clouds, camera images, velocity/acceleration data) into simulation scenarios. These virtual replicas allow repeated testing without additional real-world time consumption, enabling statistical safety proof through virtual replication rather than physical repetition.
Solution Approach 2:
The system processes multi-sensor data (LIDAR, cameras, velocity/acceleration sensors) into a unified simulation scenario format that can serve multiple purposes: training autonomous driving functions, testing safety, and analyzing diverse driving conditions. This universal processing pipeline eliminates the need for separate data collection campaigns for different testing objectives, reducing overall resource consumption.
3Productivity
If simulations are used to increase the number of driven kilometers, then the variety of training scenarios can be expanded, but the complexity of modeling appropriate driving scenarios increases
Solution Approach 1:
The patent creates virtual copies of real-world driving scenarios by processing recorded sensor data (LIDAR point clouds, camera images, velocity/acceleration data) into simulation scenarios. These virtual replicas allow repeated testing without additional real-world time consumption, enabling statistical safety proof through virtual replication rather than physical repetition.
Solution Approach 2:
The system automatically processes raw sensor data into structured simulation scenarios using automated algorithms for merging LIDAR point clouds, detecting objects, and generating trajectories. This self-service automation eliminates the need for manual scenario modeling, significantly reducing the complexity burden while enabling large-scale scenario generation for increased virtual driving kilometers.
4Reliability
If recorded sensor data is replayed for testing, then previously encountered scenarios can be tested, but the ability to test new or critical situations is limited
Solution Approach 1:
The system performs preliminary processing of raw sensor data into structured simulation scenarios that can be reused multiple times. By pre-processing and organizing the data into reusable scenario formats during initial data collection, the system eliminates the need to repeat real-world data collection for each test iteration, significantly reducing time loss.
Solution Approach 2:
The patent enables dynamic modification of simulation scenarios by allowing users to adjust parameters, combine scenarios, or introduce critical situations that may not have been captured in original recordings. This dynamic flexibility transforms static replay into adaptable scenario generation, expanding scenario coverage beyond what was originally recorded while maintaining the reliability of real sensor data foundations.
Data Source
AI summary
A method for generating a simulation scenario includes: receiving raw data, wherein the raw data comprises a plurality of successive LIDAR point clouds, a plurality of successive camera images, and successive velocity and/or acceleration data; merging the plurality of LIDAR point clouds from a determined region into a common coordinate system to produce a composite point cloud; locating and classifying one or more static objects within the composite point cloud; generating road information based on the composite point cloud, one or more static objects and at least one camera image; locating and classifying one or more dynamic road users within the plurality of successive LIDAR point clouds and generating trajectories for the one or more dynamic road users; creating a simulation scenario based on the one or more static objects, the road information, and the generated trajectories for the one or more dynamic road users; and exporting the simulation scenario.