Multi-Sensor Lane Graph Generation for Autonomous Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional lane detection methods for autonomous vehicles are limited by inaccuracies and inconsistencies due to reliance on single sensor data types and online processing, leading to ineffective navigation in complex environments.
Innovation Solution
Utilizing a deep neural network (DNN), particularly a transformer architecture, to infer lane data from sensor data collected by a fleet of ego-machines, generating lane graphs that include cross-section and connection indicators to model the environment accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional lane detection methods use single sensor data types and online processing, then device complexity is reduced, but measurement precision and reliability of lane detection deteriorate
Solution Approach 1:
The patent combines multiple sensor data types (camera images, LiDAR point clouds, radar data) from multiple ego-machines to infer lane information. This merging of diverse data sources improves measurement precision and reliability by compensating for individual sensor limitations and providing redundant verification, directly resolving the contradiction between system complexity and detection precision.
Solution Approach 2:
The system performs offline processing to pre-infer lane information and generate lane graphs before actual navigation. By conducting complex computations in advance, the system reduces real-time processing requirements, maintaining low device complexity while achieving high measurement precision through thorough pre-analysis of multiple sensor datasets.
2Measurement precision
If multiple sensor types and offline processing are used, then measurement precision and reliability of lane detection improve, but computing resource usage increases
Solution Approach 1:
The patent performs computationally intensive lane inference operations offline, processing multiple sensor types and generating lane graphs in advance. This preliminary action shifts computing resource consumption from the moving object during operation to a stationary system beforehand, improving measurement precision while reducing the energy burden on autonomous vehicles during actual navigation.
Solution Approach 2:
The system generates lane graphs as simplified representations or copies of the actual lane structure, which can be stored and reused for multiple navigation tasks. This copying approach allows precise lane detection to be performed once offline, with the resulting lane graphs serving as efficient references that reduce repeated computing resource usage while maintaining high measurement precision.
3Ease of operation
If lane detection uses trajectory data from vehicles, then ease of data collection is improved, but measurement precision deteriorates due to inability to detect lane changes and merges
Solution Approach 1:
The patent merges trajectory data from multiple ego-machines with direct sensor observations (camera images, LiDAR point clouds). This combination maintains the ease of data collection from vehicle trajectories while improving measurement precision by supplementing trajectory information with direct visual and spatial data that can detect lane changes, merges, and other dynamic road features that trajectories alone cannot identify.
4Speed
If path perception is performed online using camera images, then response time is improved, but measurement precision deteriorates due to perspective dependency and missed dividers
Solution Approach 1:
The system performs lane inference and generates lane graphs in advance during offline processing, creating a pre-computed reference framework. This preliminary action allows the system to quickly match real-time camera observations against the pre-established lane graph, maintaining fast response speed while improving measurement precision by reducing reliance on real-time perspective-based detection alone.
Data Source
AI summary
In various examples, various types of sensor data from multiple ego-machines are used to infer lanes and/or generate lane graphs for use in autonomous systems and applications. In some embodiments, one or more DNNs may be used to infer lane data indicating a representation of a lane shape using sensor data from various vehicles to represent a 3D environment. The inferred lane data may include cross-section indicators that indicate cross-sections of a lane and/or connection indicators that indicate a lane channel connecting two locations (e.g., two lane portions). The inferred lane data may be used to generate a lane graph that represents lanes on a road and, in some cases, lane dividers (e.g., polyline represented as a solid line, a dashed line, a double line, etc.). A lane graph may be used, for example, to model the environment around a vehicle, facilitate localization, provide guidance for autonomous driving, etc.