Multi-modality Data Augmentation Engine for Rare Driving Scenario Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Autonomic Driving Assistant Systems (ADAS) face challenges in accurately detecting rare driving scenarios due to limited training data, leading to under-fitting issues and high false positives/negatives, as collecting and labeling such data is costly and time-consuming.
Innovation Solution
A Multi-modality Data Augmentation Engine (MDAE) is introduced to simulate vehicle data, generating high-quality labeled data for both common and rare scenarios by using real common scenario data to train an adjuster network, which refines the simulated rare scenario data, thereby improving the accuracy of scenario detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real rare driving scenario data is collected and labeled for training, then detection accuracy for rare scenarios is improved, but time consumption and cost increase significantly
Solution Approach 1:
The patent uses simulation to create virtual copies of rare driving scenarios instead of collecting real data. The simulation environment generates synthetic training data that replicates rare scenario characteristics without requiring actual field collection, thereby eliminating the time-consuming data gathering and manual labeling processes while maintaining detection accuracy.
Solution Approach 2:
The system performs preliminary action by pre-training the detection model using simulated data before deployment. This allows the model to learn from extensive synthetic rare scenario data in advance, so that when deployed in real conditions, it already possesses the detection capabilities without requiring time-consuming real-world data collection and labeling during deployment phases.
2Loss of time
If simulated data is used for training, then time and cost are reduced, but detection accuracy decreases due to differences between simulated and real data
Solution Approach 1:
The patent implements a feedback mechanism where the detection model is first trained on simulated data, then evaluated on real data, and the performance gap is used to iteratively refine both the simulation parameters and the model. This feedback loop continuously improves the alignment between simulated and real data characteristics, maintaining detection accuracy while preserving the time and cost benefits of simulation-based training.
3Reliability
If more training data is collected to avoid under-fitting, then model performance is improved, but data storage requirements and processing complexity increase
Solution Approach 1:
The patent changes the parameters of the training data by generating diverse synthetic scenarios with varying conditions, environments, and edge cases through simulation. This approach provides rich parameter variation without proportionally increasing data volume, as the simulation can efficiently generate multiple parameter combinations from a single base scenario template, improving model reliability without excessive data storage requirements.
Data Source
AI summary
A computer-implemented method for simulating vehicle data and improving driving scenario detection is provided. The method includes retrieving, from vehicle sensors, key parameters from real data of validation scenarios to generate corresponding scenario configurations and descriptions, transferring target scenario descriptions and validation scenario descriptions to target scenario scripts and validation scenario scripts, respectively, to create first raw simulation data pertaining to target scenario descriptions and second raw simulation data pertaining to validation scenario descriptions, training, by an adjuster network, a deep neural network model to minimize differences between the first raw simulation data and the second raw simulation data, refining the first and second raw simulation data of rare driving scenarios to generate rare driving scenario training data, and outputting the rare driving scenario training data to a display screen of a computing device to enable a user to train a scenario detector for an autonomic driving assistant system.