Synthetic Radar and Lidar Point Cloud Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for testing and training driving assistance systems and semi-automated driving systems require extensive real-world data collection to account for the stochastic nature of radar and LIDAR measurements, which is time-consuming and inefficient.
Innovation Solution
A method for synthetically generating radar or LIDAR point clouds using distribution functions that mimic the stochastic properties of real measurements, allowing for the creation of diverse and realistic simulation data that replicates the dynamics of radar and LIDAR reflections, including the use of multiple distribution functions and kernel density estimation to ensure accuracy and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-world data collection is used to account for stochastic measurements, then measurement reliability is improved, but time consumption and efficiency deteriorate
Solution Approach 1:
The patent creates synthetic copies of real radar and LIDAR measurements by generating artificial point clouds that replicate the stochastic characteristics of actual sensor data. Instead of collecting extensive real-world data, the system uses distribution functions to model and reproduce measurement variations, thereby maintaining reliability while eliminating time-consuming data collection campaigns
Solution Approach 2:
The patent performs preliminary modeling of measurement stochasticity by creating distribution functions from limited real data before actual testing. These pre-established distribution functions are then used to generate synthetic data that already incorporates realistic measurement variations, eliminating the need for extensive real-world data collection during the testing phase
2Adaptability or versatility
If extensive real-world data collection is performed to cover variance of test scenarios, then testing completeness is improved, but productivity deteriorates
Solution Approach 1:
The patent introduces dynamic variability into synthetic data generation by using multiple distribution functions that can be stochastically selected and combined. This allows the system to dynamically adjust the characteristics of generated point clouds to cover diverse test scenarios including different objects, positions, and environmental conditions, achieving comprehensive testing coverage without physical test drives
Solution Approach 2:
The patent systematically varies parameters such as object positions, velocities, signal strengths, and reflection characteristics by sampling from different distribution functions. This parameter variation approach enables comprehensive coverage of test scenario diversity while maintaining high productivity, as all variations can be generated computationally without physical reconfiguration
3Measurement precision
If multiple distribution functions are used to model stochastic properties, then measurement accuracy is improved, but system complexity increases
Solution Approach 1:
The patent segments the complex stochastic behavior of radar and LIDAR measurements into multiple independent distribution functions, each modeling specific aspects such as position, velocity, or signal strength. This segmentation allows accurate representation of different measurement characteristics while keeping each individual distribution function relatively simple and manageable
Data Source
AI summary
A method for synthetically generating a point cloud of radar or LIDAR reflections, a reflection indicating at least one location at which radar or LIDAR interrogating radiation has been reflected. In the method, distribution functions which according to a random distribution provide samples in each case for at least one of the variables contained in the radar or LIDAR reflections are provided; synthetic reflections are generated by drawing samples in each case from the distribution functions for variables contained in the radar or LIDAR reflections, one of multiple distribution functions being selected according to at least one selection random distribution in order to draw each sample; the synthetic reflections are combined to form the sought point cloud.