Traffic Sign Learning via Sensor Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems face challenges in achieving high accuracy and real-time updates of traffic sign data for autonomous driving, as traditional methods rely on specialized mapping vehicles with expensive sensors, limiting their coverage and frequency of updates, and consumer vehicles with lower accuracy sensors introduce variability and uncertainty.
Innovation Solution
A system that collects sensor data from consumer vehicles to cluster traffic sign observations based on location and property data, determining learned sign values through aggregation, and applies rules to ensure accurate and timely updates of traffic sign data in geographic databases, leveraging machine learning and computer vision systems for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialized mapping vehicles with expensive sensors are used, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent merges data from multiple consumer vehicles with standard sensors into a collective dataset, combining their individual low-precision measurements to achieve high-precision traffic sign detection and mapping updates without requiring any single vehicle to have specialized expensive sensors
Solution Approach 2:
The system enables standard consumer vehicle sensors to perform specialized mapping functions by aggregating data across a fleet, making ordinary vehicles capable of contributing to high-precision geographic databases through their participation in the collective data gathering network
2Measurement precision
If specialized mapping vehicles are used, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent combines data collection efforts across a large fleet of consumer vehicles, merging their individual observations to achieve comprehensive coverage and frequent updates that would be impossible for a small number of specialized mapping vehicles to accomplish alone
Solution Approach 2:
Instead of relying on a few specialized vehicles, the system uses many standard consumer vehicles that copy the data collection function, distributing the mapping workload across the fleet to increase overall productivity and update frequency
3Device complexity
If consumer vehicles with lower accuracy sensors are used, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent merges observations from multiple consumer vehicles with standard sensors, combining their individual lower-precision measurements through clustering and aggregation algorithms to produce high-precision traffic sign data that compensates for the limitations of individual vehicle sensors
Solution Approach 2:
The system uses multiple copies of standard consumer vehicles instead of single specialized vehicles, relying on the redundancy and statistical power of numerous identical or similar sensor systems to achieve precision that individual units cannot attain alone
Data Source
AI summary
An approach is provided for traffic sign learning. The approach involves, for example, receiving a plurality of traffic sign observations generated using sensor data collected from a plurality of vehicles. Each of the plurality of traffic sign observations includes location data and sign property data for an observed traffic sign corresponding to said each of the plurality of traffic sign observations. The approach also involves clustering the plurality of traffic speed sign observations into at least one cluster based on the location data and the sign property data. The approach further involves determining a learned sign for the at least one cluster, and determining a learned sign value indicated by the learned sign based on the location data, the sign property data, or a combination of the plurality of traffic sign observations aggregated in the at least one cluster.