Map Data Noise Filtering for Autonomous Vehicle Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mapping service providers face challenges in creating accurate road geometry representations that exclude special areas and noise, while also needing to transition to three-dimensional representations to support autonomous vehicles, requiring efficient probe data processing and noise filtering in a distributed computing environment.
Innovation Solution
A method involving the retrieval of probe data from sensors, division into grid cells, detection of directional groupings, determination of density maxima locations, and creation of continuous road paths to build a base map, with additional processing to identify and exclude special areas and unconstrained regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If probe data from special areas (e.g., construction zones, tree-trimming vehicles) is included in road geometry creation, then more comprehensive coverage is achieved, but map accuracy deteriorates due to noise from non-permanent road segments
Solution Approach 1:
The patent extracts and removes probe data from special areas (construction zones, tree-trimming vehicles, etc.) from the overall dataset. By identifying these special areas through spatial analysis and excluding their probe data, the system prevents noise from contaminating the road geometry representation while retaining comprehensive coverage from normal road areas.
Solution Approach 2:
The patent applies different processing quality standards to different geographic regions. Normal road areas receive standard processing, while special areas are identified and handled differently (excluded or separately processed). This local differentiation ensures high accuracy for permanent road segments while managing the complexity of special areas appropriately.
2Device complexity
If traditional centralized processing methods are used for probe data, then algorithm simplicity is maintained, but processing efficiency deteriorates in large-scale distributed environments
Solution Approach 1:
The patent segments the probe data processing into independent, parallelizable units that can be distributed across multiple computing nodes. By dividing the geographic area into regions and assigning different regions to different processing units, the system achieves scalable efficiency while maintaining relatively simple individual processing algorithms at each node.
Solution Approach 2:
The patent creates a universal processing framework that can handle multiple types of probe data (from different sensors, different vehicle types, different geographic areas) through a common distributed processing architecture. This multi-functional approach maintains algorithmic simplicity while achieving high processing efficiency across diverse data types.
3Device complexity
If two-dimensional road geometry is used for mapping, then data processing simplicity is maintained, but applicability to autonomous vehicles deteriorates due to lack of three-dimensional information
Solution Approach 1:
The patent extends the traditional two-dimensional road geometry representation by incorporating elevation data to create three-dimensional road models. This additional vertical dimension provides the depth information required for autonomous vehicle navigation (e.g., detecting bridges, overpasses, and terrain variations) while building upon the existing two-dimensional foundation rather than completely replacing it.
Data Source
AI summary
An approach is provided for identifying special areas with significant vehicle activities that do not represent established roads (e.g., construction areas). The approach involves, for example, retrieving probe data collected from sensors of probe devices traveling within a geographic area including geographic partitions. The approach also involves dividing each of the geographic partitions into grid cells. The approach further involves, for each grid cell, detecting directional groupings of the probe data that share a respective common direction of travel. The approach further involves detecting a set of grid cells with a count of the directional groupings that is above a count threshold. The approach further involves designating an area associated with the set of grid cells as an unconstrained area or a sub-area of the unconstrained area. The approach further involves providing the unconstrained area and/or the sub-area as a mapping data output.