Road Topography Change Detection for Selective Map Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in efficiently processing and updating map data due to the sheer volume of data required for navigation, which can limit or adversely affect their navigation capabilities.
Innovation Solution
A system and method using a trained model to identify and update road topography changes in a road segment by analyzing drive data packets, determining changes that impact vehicle drivable paths by a threshold amount, and updating the road topography database accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional mapping technology is used to store and update map data, then complete navigation information is available, but the volume of data required poses daunting challenges for storage and processing
Solution Approach 1:
The patent extracts only the essential navigation elements (road centerlines, lane markings, intersections, drivable surfaces) from complete map data, storing merely what is necessary for autonomous vehicle navigation rather than all possible map information
Solution Approach 2:
The patent segments the map into discrete road segments with specific geometric representations, allowing efficient storage and processing by dividing the continuous map into manageable, standardized units that can be processed independently
2Reliability
If comprehensive map updates are performed frequently, then navigation accuracy is maintained, but computational resources and time required for processing increase significantly
Solution Approach 1:
The system employs autonomous vehicles themselves to collect and validate map data during normal operation, eliminating the need for dedicated survey vehicles or manual mapping processes. The vehicles self-report observations that contribute to map updates
Solution Approach 2:
The patent implements selective updating where only road segments with detected changes are updated, rather than performing comprehensive updates across the entire map. This partial action approach reduces processing requirements while maintaining accuracy where needed
3Measurement precision
If all road changes are detected and processed, then complete map accuracy is achieved, but the time and computational resources allocated to map updates increase
Solution Approach 1:
The system extracts and processes only those road changes that affect the drivable surface geometry, filtering out irrelevant changes such as signage or surface texture variations that do not impact navigation
Solution Approach 2:
The patent replaces traditional manual or survey-based map updating mechanisms with automated image processing and machine learning algorithms that analyze camera data from autonomous vehicles to detect and validate road changes
Data Source
AI summary
A system for detecting road topography changes in a road segment using a trained model, the system comprising: at least one processor comprising circuitry and a memory, wherein the memory includes instructions that when executed by the circuitry cause the at least one processor to: receive at least one drive data packet from at least one vehicle that traversed the road segment, wherein the at least one drive data packet includes one or more representations of at least one road topography object associated with the road segment, generate a collected data representation of road topography for the road segment based on the one or more representations of the at least one road topography object included in the at least one drive data packet, obtain, from a road topography database, a mapped data representation for the road segment based on stored road topography information, provide to the trained model an input including the collected data representation and the mapped data representation of road topography for the road segment, wherein the trained model is configured to automatically identify at least one change in road topography of the road segment based on an inferred difference between the collected data representation and the mapped data representation of road topography, and wherein the trained model is also configured to provide an output identifying the at least one change in road topography and indicating whether the at least one change in road topography would impact at least one vehicle drivable path representation stored in the road topography database by more than a threshold amount, and update the road topography database based on the at least one change in road topography after a determination that the at least one change in road topography would impact the at least one vehicle drivable path representation stored in the road topography database by more than a threshold amount.