Aerial Road View Generation for Long-Range Vehicle Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face inaccuracies in generating aerial views of roads, limiting their ability to identify lanes, vehicles, and objects, especially at longer distances, which affects their control and planning, particularly for braking scenarios.
Innovation Solution
A method using a machine learning model to generate an aerial view from a forward-facing camera view, processing the image to determine the vehicle's trajectory and road layout in a Frenet-Serret coordinate system, allowing for accurate identification of obstacles and lane lines up to 200 meters, enabling improved autonomous control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If sensor data is used to generate aerial view, then autonomous vehicle control is enabled, but measurement precision deteriorates at distances greater than 50 meters
Solution Approach 1:
The patent creates a virtual aerial view by copying and transforming the forward-facing camera image into an aerial perspective using machine learning models. This virtual copy provides accurate road layout and vehicle position information at distances beyond the reliable sensor range, resolving the measurement precision deterioration issue while maintaining autonomous control capability
Solution Approach 2:
The patent transforms the 2D forward-facing camera view into a top-down aerial view representation, changing the dimensional perspective. This dimensionality transformation allows the system to accurately represent road geometry and vehicle positions in the aerial domain, overcoming the distance limitations of the original sensor data
2Device complexity
If forward-facing camera view is used, then device complexity is reduced, but measurement precision of aerial view deteriorates
Solution Approach 1:
The system creates a synthetic aerial view by copying and geometrically transforming the forward-facing camera image. This approach eliminates the need for additional sensors while generating accurate aerial perspective data through machine learning-based image transformation, resolving the contradiction between device simplicity and measurement precision
Solution Approach 2:
The patent replaces the mechanical sensor array required for traditional aerial view generation with a computational image processing system. The machine learning model performs the transformation from forward-facing to aerial view, substituting physical sensors with algorithmic processing to achieve accurate aerial measurements using only the existing camera
3Productivity
If aerial view range is extended to 200 meters, then autonomous control capability is improved, but reliability of object identification deteriorates
Solution Approach 1:
The patent creates a virtual aerial view that copies and extends the visible road layout up to 200 meters ahead. This virtual representation maintains reliable object identification by using the machine learning model to accurately predict road geometry and vehicle positions beyond the immediate sensor range, enabling extended braking distance coverage without sacrificing identification reliability
Data Source
AI summary
This application is directed to aerial view generation for vehicle control. A vehicle obtains a forward-facing view of a road captured by a front-facing camera of a vehicle and applies a machine learning model to process the forward-facing view to predict determine a trajectory of the vehicle and a road layout based on a Frenet-Serret coordinate system of the road for the vehicle. The trajectory of the vehicle is combined with the road layout to predict an aerial view of the road, and the aerial view of the road is used to at least partially autonomously drive the vehicle. In some embodiments, the machine learning model is applied to process the forward-facing view to determine a first location of an obstacle vehicle in the Frenet-Serret coordinate system. The obstacle vehicle is placed on the aerial map based on the first location of the obstacle vehicle.