Autonomous Driving Model Validation With Low-Discrepancy Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for training and validating autonomous driving models face challenges in ensuring evenly distributed training data, leading to inadequate coverage of scenarios and potential safety issues due to variations in data density.
Innovation Solution
The use of low-discrepancy sequences to map data corpora into multidimensional spaces, allowing for the selection of evenly distributed training data sets and the generation of synthetic data for comprehensive model training and validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional random sampling methods are used to select training data, then data selection process is simple, but data distribution is uneven leading to inadequate scenario coverage
Solution Approach 1:
The patent transforms the data selection process by changing from random sampling to low-discrepancy sequence sampling. This parameter change in the sampling methodology ensures that training data points are distributed more uniformly across the multidimensional scenario space, improving scenario coverage and model reliability without requiring complex manual intervention
Solution Approach 2:
The patent introduces low-discrepancy sequences as an intermediary mechanism between the training data corpus and the model training process. This intermediary structure systematically organizes and selects data samples to achieve even distribution across scenarios, resolving the contradiction between simple selection processes and comprehensive coverage
2Adaptability or versatility
If more training data is collected to improve coverage, then scenario diversity increases, but data density variations cause some regions to be overrepresented while others are underrepresented
Solution Approach 1:
The patent changes the fundamental parameter of data sampling from random to systematic low-discrepancy sequencing. This ensures that as more training data is collected, the samples are automatically distributed uniformly across the scenario space, achieving both high scenario diversity and uniform data density without manual balancing
Solution Approach 2:
The patent applies low-discrepancy sequences in a multidimensional space where each dimension represents a scenario parameter. This dimensional approach ensures that data points are evenly distributed across all scenario combinations, achieving uniform coverage in high-dimensional scenario spaces that traditional methods cannot handle
3Reliability
If traditional validation methods are used, then validation process is straightforward, but validation coverage is insufficient leading to safety issues
Solution Approach 1:
The patent changes the validation approach by using low-discrepancy sequences to generate test scenarios. This ensures that validation covers the full range of possible scenarios uniformly, improving safety assurance while maintaining efficient automated processing through systematic scenario generation
Data Source
AI summary
Autonomous vehicle model training and validation using low-discrepancy sequences may include: generating a low-discrepancy sequence in a multidimensional space comprising a plurality of multidimensional points; mapping each sample of a plurality of samples of a data corpus to a corresponding entry in the low-discrepancy sequence, wherein each sample of the plurality of samples comprises one or more environmental descriptors for an environment relative to a vehicle and one or more state descriptors describing a state of the vehicle; selecting, from the data corpus, a training data set by selecting, for each multidimensional point of the low-discrepancy sequence having one or more mapped samples, a mapped sample for inclusion in the training data set; and training one or more models used to generate autonomous driving decisions of an autonomous vehicle based on the selected training data set.