Lane Template Localization for Occluded and Curved Roads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional approaches for vehicle localization and object detection in autonomous systems face challenges due to inaccurately detected lane boundaries, occlusions, and curvatures, leading to failed localization and unsafe navigation.
Innovation Solution
A machine learning model is trained to classify detected lane boundaries into lane templates and localize vehicles and objects based on lane detection and object detection data, generating a machine-readable map for precise localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional lane detection methods are used, then object detection can be performed, but localization accuracy deteriorates due to inaccurately detected lane boundaries, occlusions, and curvatures
Solution Approach 1:
The patent introduces lane templates as an intermediary between lane detection and object localization. The lane template serves as a reference model that mediates the relationship between detected lane boundaries and object positions, enabling accurate localization even when direct boundary detection is inaccurate or occluded. The template matching process acts as a buffer that corrects detection errors.
Solution Approach 2:
The patent performs preliminary lane template generation and selection before object localization. By pre-establishing lane templates from map data and selecting the most appropriate template before localization, the system prepares accurate reference frameworks in advance, avoiding the need to resolve detection uncertainties during the critical localization step.
2Adaptability or versatility
If lane boundaries are detected in complex environments with occlusions and curvatures, then more comprehensive environmental data is obtained, but localization reliability deteriorates due to detection inaccuracies
Solution Approach 1:
The lane template acts as an intermediary that filters and corrects detection data from complex environments. By matching detected lane boundaries against pre-established templates, the system can reliably identify valid boundaries even in occluded or curved scenarios, maintaining localization reliability while handling environmentally complex situations.
Solution Approach 2:
The patent transforms raw lane detection data into standardized template parameters through matching processes. By converting detected boundaries into template-based parameter representations (such as lane curvature, width, and position relative to template models), the system normalizes varied environmental conditions into consistent localization parameters.
Data Source
AI summary
Methods, systems, and non-transitory computer-readable media are configured to perform operations comprising determining lane detection data and object detection data associated with an environment; determining a lane template based on the lane detection data; and generating localization data that identifies a location of an object in the environment based on the lane template and the object detection data.