Lane Curvature Correction Using Vehicle Path Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane curvature estimation systems for autonomous and semi-autonomous vehicles face challenges in accuracy due to varying lane markings, poor lighting, and weather conditions, which can lead to incorrect curvature estimates affecting navigation and safety.
Innovation Solution
A system and method that utilize a lane curvature model trained on actual vehicle path data and sensor data to correct estimated lane curvatures, using a processor-based system with a curvature estimation module and a curvature correction module to generate and refine accurate lane curvature estimates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensor-based lane curvature estimation is used, then the system can operate with existing sensors, but the estimation accuracy deteriorates due to varying lane markings, poor lighting, and weather conditions
Solution Approach 1:
The system collects actual vehicle path data from GPS and sensor systems, compares it with estimated lane curvature data, and uses this feedback to train a machine learning model that continuously improves estimation accuracy by learning from the difference between estimated and actual paths
Solution Approach 2:
The system uses the vehicle's own operational data (actual paths traveled) to train and improve its own curvature estimation model, enabling the system to self-improve without external intervention
2Measurement precision
If a machine learning model is trained on vehicle path data to correct curvature estimates, then estimation accuracy is improved, but the system complexity increases
Solution Approach 1:
The system pre-trains the machine learning model using historical vehicle path data before deployment, so that during actual operation, the model can quickly provide corrections without requiring complex real-time training computations
Solution Approach 2:
The machine learning model acts as an intermediary component that receives both sensor-based curvature estimates and actual path data, processes them, and outputs corrected curvature estimates, thereby managing system complexity through modular architecture
Data Source
AI summary
In one embodiment, example systems and methods relate to a manner of providing lane curvature estimations. As vehicles travel, lane curvature estimates for lane segments as provided by the vehicle lane curvature estimation systems are collected. The estimated lane curvatures and the actual paths of the vehicles as they traveled in the lane segments are used as training data for a model that can both determine if a lane curvature estimate for a lane segment is likely incorrect and provide a correct lane curvature estimate. When a vehicle later travels such a lane segment, the model can be used to provide the vehicle with a correct lane curvature estimate. The correct estimate can be used by the vehicle in place of an estimate generated by the existing lane curvature estimation system of the vehicle.