Real-time detection of lanes and boundaries by autonomous vehicles

Abstract

In various examples, sensor data representative of an image of a field of view of a vehicle sensor may be received and the sensor data may be applied to a machine learning model. The machine learningmodel may compute a segmentation mask representative of portions of the image corresponding to lane markings of the driving surface of the vehicle. Analysis of the segmentation mask may be performed to determine lane marking types, and lane boundaries may be generated by performing curve fitting on the lane markings corresponding to each of the lane marking types. The data representative of the lane boundaries may then be sent to a component of the vehicle for use in navigating the vehicle through the driving surface.

Description

Background technique [0001] For autonomous vehicles to operate safely in all environments, autonomous vehicles must be able to efficiently perform vehicle maneuvers - such as lane keeping, lane changing, lane splitting, turning, stopping and starting at intersections, crosswalks, etc., and / or other vehicles manipulate. For example, for an autonomous vehicle to navigate surface streets (e.g., city streets, sidewalks, neighborhoods, etc.) lanes, intersections, crosswalks, boundaries, etc.) where one or more subdivisions of the road are usually minimally delineated and appear to be difficult even for the most attentive and experienced driver at certain difficult to identify under certain conditions. In other words, an autonomous vehicle must be the functional equivalent of an attentive human driver, utilizing a perception and action system with an incredible ability to recognize and respond to moving and stationary obstacles in complex environments, just is to avoid collisions...

AI Technical Summary

Problems solved by technology

However, performing lane and road boundary detection in this way has proven to be too computationally expensive to run efficiently in real-time, and / or be inaccurate due to shortcuts implemented to reduce computational requirements

In other words, these traditional systems either give up the accuracy of real-time operation, or give up real-time operation to produce acceptable accuracy

Furthermore, even in conventional systems that achieve the level of accuracy required to safely and efficiently operate autonomous vehicles, the accuracy is limited to ideal road and weather conditions

As a result, autonomous vehicles operating using these traditional methods may not operate accurately in real time and / or with precision in all road and weather conditions

Images