Geospatial Trajectory Alignment for HD Map Geometry Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for 3D road geometry modeling and feature detection in autonomous vehicle navigation are resource-intensive, time-consuming, and costly, often relying on manual or semi-automated data analysis, and suffer from unreliable feature detection that can impact safety and efficiency.
Innovation Solution
A system using an iterative attention model with Gated Recurrent Unit gating patterns and Set Transformer processes geospatial observations from discrete trajectories to generate high-definition maps, aligning and concatenating data for navigational assistance and semi-autonomous vehicle control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional manual or semi-automated methods are used for 3D road geometry modeling and feature detection, then measurement precision can be maintained, but productivity is significantly reduced and loss of time increases
Solution Approach 1:
The patent replaces manual mechanical measurement and analysis processes with automated sensor-based systems and machine learning algorithms. Sensors capture geospatial observations automatically, and neural networks process this data to generate HD maps, eliminating the need for manual measurement while maintaining or improving precision through consistent algorithmic processing.
Solution Approach 2:
The system enables self-service by allowing sensors and processing algorithms to automatically capture, process, and generate map data without human intervention. The neural network models autonomously perform feature detection, trajectory alignment, and map generation tasks that previously required manual analysis, significantly improving productivity while maintaining measurement precision through automated consistency.
2Reliability
If feature detection systems are made more complex to improve reliability, then detection accuracy may improve, but device complexity increases
Solution Approach 1:
The patent employs universal sensor platforms and multi-functional neural network models that can perform multiple detection tasks (road geometry, obstacles, features) with a single integrated system. This approach improves reliability through consistent multi-task performance while avoiding the complexity of separate specialized systems for each detection function.
Solution Approach 2:
The patent introduces trajectory alignment as an intermediary processing step that mediates between raw sensor observations and final feature detection. By aligning trajectories using learned offsets before detection, the system improves detection reliability through standardized reference frames while keeping the overall system architecture manageable through modular processing stages.
3Manufacturing precision
If more data processing is performed to improve map accuracy, then manufacturing precision improves, but use of energy and computational resources increases
Solution Approach 1:
The patent performs preliminary trajectory alignment and feature extraction during data collection phases, preparing processed data for later map generation. This preliminary processing reduces the computational burden during final map creation, maintaining high precision while distributing energy consumption across multiple stages rather than concentrating it all in the final processing step.
Solution Approach 2:
The patent segments the map generation process into distinct stages: trajectory alignment, feature detection, and map construction. Each stage processes specific aspects of the data independently, allowing for optimized computational resource allocation at each step and reducing overall energy consumption compared to monolithic processing approaches.
Data Source
AI summary
A method, apparatus and computer program product are provided for learning to generate maps from raw geospatial observations from sensors traveling within an environment. Methods may include: processing geospatial observations from discrete trajectories through an iterative attention model incorporating a Gated Recurrent Unit gating pattern to obtain a feature summary; determining a drive offset for each of the discrete trajectories based on the feature summary; aligning the discrete trajectories to generate aligned geospatial observations based, at least in part, on the drive offset for a respective discrete trajectory; concatenating the aligned geospatial observations; processing the concatenated, aligned geospatial observations using at least one Set Transformer; generating, from the at least one Set Transformer, map geometries including objects from the geospatial observations; and providing for at least one of navigational assistance or at least semi-autonomous vehicle control based on the map geometries.