Hybrid planning system for autonomous vehicles

Patent metadata

Field	Value
Publication number	US-11628858-B2
Application number	US-202017021207-A
Country	US
Kind code	B2
Filing date	Sep 15, 2020
Priority date	Sep 15, 2020
Publication date	Apr 18, 2023
Grant date	Apr 18, 2023

Abstract

In one embodiment, a system/method generates a driving trajectory for an autonomous driving vehicle (ADV). The system perceives an environment of an autonomous driving vehicle (ADV). The system determines one or more bounding conditions based on the perceived environment. The system generates a first trajectory using a neural network model, wherein the neural network model is trained to generate a driving trajectory. The system evaluates/determines if the first trajectory satisfies the one or more bounding conditions. If the first trajectory satisfies the one or more bounding conditions, the system controls the ADV autonomously according to the first trajectory. Otherwise, the system controls the ADV autonomously according to a second trajectory, where the second trajectory is generated based on an objective function, where the objective function is determined based on at least the one or more bounding conditions.

First claim

What is claimed is:

1. A computer-implemented method to generate a driving trajectory for an autonomous driving vehicle (ADV), the method comprising:
determining one or more bounding conditions based on a perceived environment of an ADV, wherein the one or more bounding conditions include a path bound and a speed bound;
generating a first trajectory with a deep learning models layer, the first trajectory generated using a neural network model trained to generate a driving trajectory;
generating a second trajectory with a rules-based models layer, the second trajectory generated based on an objective function and the one or more bounding conditions such that the second trajectory satisfies the one or more bounding conditions;
determining if the first trajectory satisfies the one or more bounding conditions;
if the first trajectory satisfies the one or more bounding conditions, controlling the ADV autonomously according to the first trajectory; and
otherwise, controlling the ADV autonomously according to the second trajectory.

2. The method of claim 1 , wherein generating the second trajectory based on at least an objective function comprises:
generating a path profile based on traffic rules and one or more obstacles perceived by the ADV;
generating a speed profile based on the path profile, wherein the speed profile includes, for each of the one or more obstacles, a decision to yield or overtake the obstacle; and
generating the second trajectory based on the path profile, the speed profile, and the objective function using dynamic programming such that the ADV can be controlled autonomously based on the second trajectory.

3. The method of claim 1 , further comprising smoothing the first or the second trajectory based on a smoothing function, wherein the smoothing function is determined based on the one or more bounding conditions.

4. The method of claim 1 , wherein the one or more bounding conditions includes a lane bound, an obstacle bound, or a traffic light bound.

5. The method of claim 1 , wherein the first trajectory is generated using the neural network model based on a capability of the ADV and the perceived environment of the ADV.

6. The method of claim 1 , further comprising determining the one or more bounding conditions based on map information, wherein the map information is retrieved from a local or a remote database of the ADV.

7. The method of claim 1 , wherein generating the second trajectory based on at least an objective function comprises:
generating a plurality of trajectory candidates;
determining a trajectory cost based on the objective function for each of the plurality of trajectory candidates, the objective function having a safety factor, a comfort factor, and/or a progress factor; and
selecting one of the plurality of trajectory candidates as the second trajectory, wherein the trajectory selected as the second trajectory has a lowest trajectory cost.

8. A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations, the operations comprising:
determining one or more bounding conditions based on a perceived environment of an autonomous driving vehicle (ADV), wherein the one or more bounding conditions include a path bound and a speed bound;
generating a first trajectory with a deep learning models layer, the first trajectory generated using a neural network model trained to generate a driving trajectory;
generating a second trajectory with a rules-based models layer, the second trajectory generated based on an objective function and the one or more bounding conditions such that the second trajectory satisfies the one or more bounding conditions;
determining if the first trajectory satisfies the one or more bounding conditions;
if the first trajectory satisfies the one or more bounding conditions, controlling the ADV autonomously according to the first trajectory; and
otherwise, controlling the ADV autonomously according to the second trajectory.

9. The non-transitory machine-readable medium of claim 8 , wherein generating the second trajectory based on at least an objective function comprises:
generating a path profile based on traffic rules and one or more obstacles perceived by the ADV;
generating a speed profile based on the path profile, wherein the speed profile includes, for each of the one or more obstacles, a decision to yield or overtake the obstacle; and
generating the second trajectory based on the path profile, the speed profile, and the objective function using dynamic programming such that the ADV can be controlled autonomously based on the second trajectory.

10. The non-transitory machine-readable medium of claim 8 , wherein the operations further comprise smoothing the first or the second trajectory based on a smoothing function, wherein the smoothing function is determined based on the one or more bounding conditions.

11. The non-transitory machine-readable medium of claim 8 , wherein the one or more bounding conditions includes a lane bound, an obstacle bound, or a traffic light bound.

12. The non-transitory machine-readable medium of claim 8 , wherein the first trajectory is generated using the neural network model based on a capability of the ADV and the perceived environment of the ADV.

13. The non-transitory machine-readable medium of claim 8 , wherein the operations further comprise determining the one or more bounding conditions based on map information, wherein the map information is retrieved from a local or a remote database of the ADV.

14. The non-transitory machine-readable medium of claim 8 , wherein generating the second trajectory based on at least an objective function comprises:
generating a plurality of trajectory candidates;
determining a trajectory cost based on the objective function for each of the plurality of trajectory candidates, the objective function having a safety factor, a comfort factor, and/or a progress factor; and
selecting one of the plurality of trajectory candidates as the second trajectory, wherein the trajectory selected as the second trajectory has a lowest trajectory cost.

15. A data processing system, comprising:
a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations, the operations including:
determining one or more bounding conditions based on a perceived environment of an autonomous driving vehicle (ADV), wherein the one or more bounding conditions include a path bound and a speed bound;
generating a first trajectory with a deep learning models layer, the first trajectory generated using a neural network model trained to generate a driving trajectory;
generating a second trajectory with a rules-based models layer, the second trajectory generated based on an objective function and the one or more bounding conditions such that the second trajectory satisfies the one or more bounding conditions;
determining if the first trajectory satisfies the one or more bounding conditions;
if the first trajectory satisfies the one or more bounding conditions, controlling the ADV autonomously according to the first trajectory; and
otherwise, controlling the ADV autonomously according to the second trajectory.

16. The system of claim 15 , wherein generating the second trajectory based on at least an objective function comprises:
generating a path profile based on traffic rules and one or more obstacles perceived by the ADV;
generating a speed profile based on the path profile, wherein the speed p

Assignees

• Baidu Usa Llc

Inventors

• Jiang Yifei
• Zhou Jinyun
• Tao Jiaming
• Jiang Shu
• Hu Jiangtao
• Miao Jinghao
• Song Shiyu

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/08

  Learning methods · CPC title

• B60W60/0015

  specially adapted for safety · CPC title