Edge Driving Data Reduction Through Real-Time Scenario Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The collection of large amounts of data for Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicles (AV) poses computational challenges due to limited storage capacities and the inefficiency of data retention, with significant portions being repetitive, low-quality, or irrelevant, necessitating improved data flow and processing.
Innovation Solution
A system and method for identifying driving scenarios in real-time using a scenario description language, determining signal computation functions, and selectively storing or discarding data based on detected scenarios to reduce computational burden and optimize data retention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If large amounts of data are collected for ADAS and AV validation, then data completeness and validation quality are improved, but storage capacity requirements and computational processing burden increase significantly
Solution Approach 1:
The system performs preliminary classification of driving data into structured categories (scenario-based, component-based, quality-based) during the data collection phase. This preliminary organization enables efficient filtering and retrieval later, reducing the need to process and store all raw data while maintaining validation quality.
Solution Approach 2:
The patent segments the monolithic data collection approach into multiple organized categories including scenario-based data, component-based data, quality metrics, and relevance annotations. This segmentation allows the system to selectively retain and process only relevant data subsets for specific validation tasks.
2Loss of information
If all collected driving data is retained for post-processing and annotation, then data availability for analysis is improved, but storage costs and processing time increase
Solution Approach 1:
The system performs preliminary annotation and metadata generation during data collection, including scenario classification, component identification, and quality assessment. This preliminary processing reduces the need for extensive post-processing while ensuring data availability for immediate analysis.
Solution Approach 2:
The patent introduces an intermediary data structure layer that organizes raw driving data into structured formats with embedded metadata and annotations. This intermediary representation enables efficient querying and analysis without requiring access to all原始数据, reducing processing time while maintaining data availability.
3Adaptability or versatility
If comprehensive data collection is performed to capture all driving scenarios, then scenario coverage is improved, but storage requirements and data management complexity increase
Solution Approach 1:
The patent segments comprehensive data collection into organized categories including scenario-based data, component-based data, quality metrics, and relevance annotations. This structured segmentation enables efficient management and retrieval of specific scenario types without handling all raw data.
Solution Approach 2:
The system implements a universal data structure framework that can accommodate multiple scenario types, components, and quality metrics within a single organized system. This multi-functional framework reduces management complexity by providing consistent handling procedures for diverse data types.
Data Source
AI summary
Methods, computing devices, and software programs for identifying a driving scenario in real-time driving data. A driving scenario is received and translated into an ordered sequence of events that correspond to the driving scenario. A signal computation function is determined for each event in the sequence, which quantifies proximity to the respective event. Driving data is received for a plurality of time steps. Values of each signal computation function are determined, evaluated for the driving data at each of the plurality of time steps. It is determined whether the driving scenario has occurred in the driving data based on the values of the signal computation function. A first portion of the driving data is either modified, discarded, or stored in memory based on the determination whether the driving scenario has occurred.