Apple has Won their third Project Titan Patent today that covers a Lane Changing Generator System for Autonomous Vehicles
Today the U.S. Patent and Trademark Office officially granted Apple their third Project Titan patent that relates to lane changing generation for autonomous vehicle control.
One aspect of Apple's granted patent is a vehicle that includes an automated vehicle control system, a sensor, and a lane generator that is configured to receive information that describes a first lane portion and a second lane portion and determine that a discontinuity is present between the first lane portion and the second lane portion.
The lane generator is further configured to obtain a sensor output from the sensor, detect presence of a nearby vehicle in the first lane portion using the sensor output, identify a classification for the nearby vehicle by analyzing the sensor output, and select a vehicle kinematics model for the nearby vehicle in accordance with the classification.
The lane generator is further configured to determine one or more paths for a simulated vehicle from the first lane portion to the second lane portion using the vehicle kinematics model, and determine a third lane portion based on the one or more paths from the first lane portion to the second lane portion such that the third lane portion defines a traversable route from the first lane portion to the second lane portion in accordance with the vehicle kinematics model. The automated vehicle control system is configured to generate control outputs based in part on the third lane portion.
The discontinuity may be in an intersection that is located between the first lane portion and the second lane portion, and the third lane portion represents a travel path through the intersection.
In some implementations, the information that describes the first lane portion and the second lane portion is received from stored mapping information. In some implementations, the information that describes the first lane portion and the second lane portion is received by determining boundaries for the first lane portion and the second lane portion based on sensor outputs.
The discontinuity may be in an intersection that is located between the first lane portion and the second lane portion, and the third lane portion represents a travel path through the intersection.
Some implementations of the method also include identifying a classification for a nearby vehicle and selecting the vehicle kinematics model for the simulated vehicle in accordance with the classification. Identifying the classification for the nearby vehicle may include obtaining a sensor output, detecting presence of the nearby vehicle in the first lane portion using the sensor output, and analyzing the sensor output to determine the classification for the nearby vehicle.
Apple's patent FIG. 1 below is a block diagram of a lane generation system; FIG. 2 is a block diagram that shows a first implementation in which the lane generation system is used to process stored map data; (3) FIG. 3 is a block diagram that shows a second implementation in which the lane generation system is used to process lane segments that are output by a machine vision-based lane generator.
Apple's patent FIG. 4 above we see a block diagram that shows a third implementation in which the lane generation system is used to process lane segments that are output by a machine vision-based lane generator using object classification information output by an object classifier; Apple's patent FIG. 7 is an illustration that shows lane segments that are defined in correspondence to the lanes of the roadway.
More specifically, Apple's patent FIG. 7 is an illustration that shows lane segments #706a-706h that are defined in correspondence to the lanes of the roadway #620. A discontinuity #707 is located in the intersection area of the roadway and corresponds to an area in which there is no explicit information describing lane geometry for use when travelling through the intersection area of the roadway.
Arrows show legally permissible movements by which vehicular traffic may travel between pairs of the lane segments 706a-706h. The term “legally permissible movements” refers to those movements (typically one of a left turn, a right turn, or a straight through movement) that may be performed by a vehicle to travel from one of the lane segments 706a-706h to another one of the lane segments 706a-706h in accordance with applicable laws and regulations. In the illustrated example, the lane segment 706b is subject to a left turn only regulation, and the only legally permissible movement is from the lane segment 706b is a left turn to the lane segment 706h.
Connectivity between lane segments via the legally permissible movements can be determined using data from multiple sources. Sources of data include roadway geometry (e.g., relative angles between lanes), street signs, the numbers of approach lanes and receiving lanes at an intersection, neighboring lanes and their features, the direction of travel, records describing previous vehicle trips at a location, sensor outputs showing movement of other vehicles at a location, and/or manual data annotation by a person. These features can be used in a decision-making model of any suitable type, such as a rules-based model or a trained machine learning model.
In some implementations, the legally permissible movements are predetermined, for example. For example, legally permissible movements may be encoded in the map information for each of the lane segments by listing all of the other ones of the lane segments to which travel is permitted. In such an implementation the lane generation system #100 may be configured to determine the legally permissible movements by accessing information describing the legally permissible movements from the map information.
For more details, review Apple's granted patent 11465620. To review more Project Titan patents, click here.
Comments