Efficient safety aware path selection and planning for autonomous machine applications

Patent metadata

Field	Value
Publication number	US-12077190-B2
Application number	US-202016877127-A
Country	US
Kind code	B2
Filing date	May 18, 2020
Priority date	May 18, 2020
Publication date	Sep 3, 2024
Grant date	Sep 3, 2024

Abstract

In various examples, systems and methods are disclosed for weighting one or more optional paths based on obstacle avoidance or other safety considerations. In some embodiments, the obstacle avoidance considerations may be computed using a comparison of trajectories representative of safety procedures at present and future projected time steps of an ego-vehicle and other actors to ensure that each actor is capable of implementing their respective safety procedure while avoiding collisions at any point along the trajectory. This comparison may include filtering out a path(s) of an actor at a time step(s)—e.g., using a one-dimensional lookup—based on spatial relationships between the actor and the ego-vehicle at the time step(s). Where a particular path—or point along the path—does not satisfy a collision-free standard, the path may be penalized more negatively with respect to the obstacle avoidance considerations, or may be removed from consideration as a potential path.

First claim

What is claimed is:

1. A method comprising:
generating, based at least on a first safety procedure that is associated with one or more first operations for controlling an ego-vehicle to avoid a collision, a first vehicle-occupied projection corresponding to a first boundary associated with the ego-vehicle in an environment if the ego-vehicle were to implement the first safety procedure at a first time instance along a proposed path and a second vehicle-occupied projection corresponding to a second boundary associated with the ego-vehicle in the environment if the ego-vehicle were to implement the first safety procedure at a second time instance along the proposed path;
generating, based at least on a second safety procedure that is associated with one or more second operations for controlling an actor to avoid a collision, a first actor-occupied projection corresponding to a third boundary associated with the actor in the environment if the actor were to implement a second safety procedure at the first time instance along an estimated path and a second actor-occupied projection corresponding to a fourth boundary associated with the actor in the environment if the actor were to implement the second safety procedure at the second time instance along the estimated path;
determining whether the first boundary intersects with the third boundary;
determining whether the second boundary intersects with the fourth boundary;
determining, based at least on whether the first boundary intersects with the third boundary and whether the second boundary intersects with the fourth boundary, a weight associated with the proposed path;
selecting, based at least on the weight, the proposed path for the ego-vehicle through the environment; and
causing, based at least on the proposed path, performance of one or more third operations associated with control of the ego-vehicle.

2. The method of claim 1 , further comprising:
determining at least one other estimated path of another actor in the environment;
comparing at least a first point along the proposed path to at least a second point along the at least one other estimated path;
determining, based at least in part on the comparing, that the first point and the second point satisfy a spatial relationship such that the first point for the first time instance is a threshold distance away from the second point for the first time instance; and
determining that, over a period of time represented by the second point, there are no intersections between the ego-vehicle and the another actor.

3. The method of claim 2 , wherein the determining that the first point and the second point satisfy the spatial relationship comprises:
determining a distance between the ego-vehicle at the first point and the another actor at the second point;
determining a speed of the another actor;
executing, based at least on the distance and the speed, a lookup in a lookup table to determine an interval of speeds for the ego-vehicle;
comparing an actual speed of the ego-vehicle to the interval of speeds; and
determining, based at least in part on the comparing the actual speed to the interval of speeds, that the actual speed is within the interval of speeds.

4. The method of claim 1 , wherein the generating the first vehicle-occupied projection comprises:
computing a subscribing shape for the ego-vehicle at the first time instance;
determining an updated subscribing shape for the ego-vehicle based at least on the first safety procedure and state information of the ego-vehicle; and
connecting one or more vertices of the subscribing shape and the updated subscribing shape.

5. The method of claim 4 , wherein the computing the subscribing shape comprises:
determining an object fence corresponding to the ego-vehicle; and
computing the subscribing shape based at least on one or more extremal values of the object fence and an orientation of the ego-vehicle.

6. The method of claim 1 , wherein the generating the first vehicle-occupied projection comprises:
generating, in three-dimensional space corresponding to space and time, the first vehicle-occupied projection representative of one or more points of the ego-vehicle; and
projecting the first vehicle-occupied projection into two-dimensional space.

7. The method of claim 1 , further comprising determining whether the first actor-occupied projection is fully contained within a first vehicle-occupied trajectory by at least casting a semi-infinite ray from a vertex of the actor-occupied projection and counting a number of intersections of the semi-infinite ray with the first boundary.

8. The method of claim 1 , further comprising:
generating a shape of a predefined radius and centered at a location on the ego-vehicle;
determining at least one additional path corresponding to at least one additional actor in the environment is entirely outside of the predefined radius; and
determining that, over a period of time represented by the at least one additional path, there are no intersections between the ego-vehicle and the additional actor.

9. The method of claim 1 , further comprising generating data representative of the proposed path of the ego-vehicle based at least on sensor data and using one or more components of a planning layer of an autonomous driving software stack of the ego-vehicle.

10. The method of claim 1 , wherein:
the one or more first operations are associated with controlling the ego-vehicle to at least one of:
begin slowing using a first deceleration along the proposed path and to a stop; or
navigate to a first new path to avoid the collision; and

the one or more second operations are associated with controlling the actor to at least one of:
begin slowing using a second deceleration along the estimated path and to a stop; or
navigate along a second new path to avoid the collision.

11. A system comprising:
one or more processing units to:
generate, based at least on a first safety procedure associated with an ego-vehicle, a first vehicle-occupied projection delineated by a first boundary associated with the ego-vehicle if the ego-vehicle were to implement the first safety procedure at a first time instance along a proposed path and a second vehicle-occupied projection delineated by a second boundary associated with the ego-vehicle if the ego-vehicle were to implement the first safety procedure at a second time instance along the proposed path;
generate, based at least on the first boundary and the second boundary, a third boundary associated with the ego-vehicle;
generate, based at least on a second safety procedure associated with an actor, an actor-occupied projection delineated by a fourth boundary associated with the actor if the actor were to implement the second safety procedure at the first time instance or the second time instance along an estimated path;
select, based at least on the third boundary and the fourth boundary, the proposed path for the ego-vehicle through the environment; and
control the ego-vehicle to navigate based at least on the proposed path.

12. The system of claim 11 , wherein the one or more processing units are further to determine one or more vehicle controls for controlling the ego-vehicle according to the proposed path.

13. The system of claim 11 , wherein the one or more processing units are further to:
determine at least one other estimated path of another actor in the environment;
compare at least a first point along at least one of the proposed path to at least a second point along the at least one other estimated path;
determine, based at least on the comparison, that the first point and the second point sa

Assignees

• Nvidia Corp

Inventors

• Ng Julia
• Nister David
• Zhang Zhenyi
• Wang Yizhou

Classifications

• B60W60/0011 Primary

  involving control alternatives for a single driving scenario, e.g. planning several paths to avoid obstacles · CPC title

• B60W2554/4042

  Longitudinal speed · CPC title

• B60W2554/80

  Spatial relation or speed relative to objects · CPC title

• B60W2554/4041

  Position · CPC title

• B60W30/0953

  the prediction being responsive to vehicle dynamic parameters · CPC title