Lane Edge Association Using Cost Matrices for Vehicle Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicles face challenges in accurately identifying lane edges due to perspective distortions and noise in sensor and map data, leading to incorrect lane identification.
Innovation Solution
A system that calculates a cost matrix to associate perceived lane edges with corresponding map data using a combination of distance and type mismatch costs, where the type mismatch cost is determined by the negative log likelihood of lane marking types representing the same underlying lane edge, and the distance cost is calculated using average squared Mahalanobis distance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data and map data are used directly for lane edge identification, then the process is simple, but the accuracy is reduced due to noise and perspective distortions
Solution Approach 1:
The patent introduces a cost matrix as an intermediary structure that mediates between sensor data and map data. The cost matrix computes association costs between perceived lane edges from sensors and mapped lane edges from maps, using metrics such as geometric distance, lane marking type compatibility, and perspective transformation consistency. This intermediary computation filters out noisy matches and identifies correct lane edge associations, thereby improving identification accuracy without requiring direct raw data comparison.
Solution Approach 2:
The system performs preliminary transformations and computations on sensor data and map data before association. This includes applying perspective transformations to correct distortion, pre-computing lane edge candidates from both sensor and map data, and preparing cost matrices in advance. By performing these preliminary actions, the system reduces the impact of noise and distortions before the actual lane edge identification, improving accuracy while managing complexity through structured preprocessing.
2Measurement precision
If multiple cost factors are computed to improve association accuracy, then the identification accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent segments the association cost computation into multiple independent cost factors, each evaluating a specific aspect of lane edge matching. These include geometric distance cost (spatial proximity), type mismatch cost (lane marking category compatibility), and perspective transformation cost (geometric consistency). By segmenting the computation, the system can evaluate each factor separately and sum them to obtain the total association cost, improving accuracy through comprehensive evaluation while managing computational power through modular, independent calculations.
Data Source
AI summary
A system for associating perceived and mapped lane edges can include a processor and a memory. The memory includes instructions such that the processor is configured to receive a sensor data representing a perceived object; receive map data representing a map object; determine a cost matrix a cost matrix indicative of an association cost for associating the map object to the perceived object; compare the association cost with an association cost threshold; and associate the perceived object with the map object based on the association cost.