Road Feature Alignment Neural Network for Drivable Path Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in navigating due to the vast amount of data they need to process and store, which can limit or adversely affect navigation, especially when relying on traditional mapping technology.
Innovation Solution
A system that generates a map for navigating a host vehicle relative to a road segment by aggregating drive information from multiple harvesting vehicles using a trained neural network to determine target drivable paths, refine road signs and lane markings, and distribute this data to host vehicles for navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional mapping technology is used to navigate autonomous vehicles, then navigation coverage can be achieved, but the sheer volume of data needed to store and update the map poses daunting challenges and adversely affects navigation
Solution Approach 1:
The patent extracts only the essential navigation elements (drivable paths, road signs, lane markings, road edges) from complete map data, storing only what is necessary for navigation rather than comprehensive map information. This reduces data volume while maintaining navigation reliability.
Solution Approach 2:
The patent segments map data into discrete road feature elements (drivable paths, signs, lane markings, edges) that can be independently processed and stored. This segmentation allows selective storage of critical navigation elements, reducing overall data volume while preserving navigation accuracy.
2Adaptability or versatility
If vast volumes of information are collected and analyzed by autonomous vehicles, then navigation decisions can be made, but the sheer quantity of data poses challenges that can limit autonomous navigation
Solution Approach 1:
The patent performs preliminary processing of map data by harvesting vehicles that pre-extract and normalize road features before distribution to autonomous vehicles. This preliminary action reduces the data processing complexity for navigating vehicles while maintaining their decision-making capabilities.
Solution Approach 2:
The patent introduces an intermediary processing system that normalizes and aggregates drive information from multiple harvesting vehicles into standardized map data. This intermediary reduces the complexity of data that autonomous vehicles must process while preserving the adaptability needed for navigation decisions.
3Measurement precision
If drive information from multiple harvesting vehicles is aggregated, then more accurate map data can be generated, but data alignment and normalization must be performed
Solution Approach 1:
The patent uses feedback mechanisms where harvested drive information is aggregated and normalized, then distributed back to the network for use by other vehicles. This feedback loop continuously improves map data accuracy through collective learning while the normalization process manages processing complexity.
Solution Approach 2:
The patent transforms drive information from multiple vehicles into a standardized parameter set representing road features (drivable paths, signs, lane markings, edges). This parameter transformation enables accurate map generation from diverse data sources while managing processing complexity through consistent normalization.
Data Source
AI summary
In one implementation, a system for generating a map for use in navigating a host vehicle relative to a road segment includes 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 drive information from each of a plurality of harvesting vehicles that traversed the road segment, wherein the drive information received from each of the plurality of harvesting vehicles includes at least one location indicator associated with an actual trajectory traveled by the harvesting vehicle, as the harvesting vehicle traversed the road segment; provide the drive information received from each of the plurality of harvesting vehicles to a trained model, wherein the trained model is configured to receive the drive information as input and output normalized drive information for each of the plurality of harvesting vehicles, wherein the normalized drive information includes the at least one location indicator aligned relative to a predetermined reference location; aggregate the normalized drive information provided for each of the plurality of harvesting vehicles to determine one or more target drivable paths through the road segment; store in the map the one or more target drivable paths; and distribute the map data to at least one host vehicle navigation system for use in navigating the host vehicle along the road segment relative to the one more mapped target drivable paths.