Steering control for vehicles

Patent metadata

Field	Value
Publication number	US-11958554-B2
Application number	US-202016949642-A
Country	US
Kind code	B2
Filing date	Nov 9, 2020
Priority date	Mar 1, 2017
Publication date	Apr 16, 2024
Grant date	Apr 16, 2024

Abstract

Model-based control of dynamical systems typically requires accurate domain-specific knowledge and specifications system components. Generally, steering actuator dynamics can be difficult to model due to, for example, an integrated power steering control module, proprietary black box controls, etc. Further, it is difficult to capture the complex interplay of non-linear interactions, such as power steering, tire forces, etc. with sufficient accuracy. To overcome this limitation, a recurring neural network can be employed to model the steering dynamics of an autonomous vehicle. The resulting model can be used to generate feedforward steering commands for embedded control. Such a neural network model can be automatically generated with less domain-specific knowledge, can predict steering dynamics more accurately, and perform comparably to a high-fidelity first principle model when used for controlling the steering system of a self-driving vehicle.

First claim

What is claimed is:

1. A method comprising:
receiving a reference trajectory associated with a vehicle, wherein the vehicle is an autonomous vehicle;
receiving sensor data from a sensor associated with the vehicle;
determining, based at least in part on the sensor data and the reference trajectory, vehicle state data comprising one or more of a yaw, a steering rate, a position, an orientation, a speed, or an acceleration of the vehicle;
determining, based at least in part on the sensor data, the reference trajectory, and the vehicle state data, vehicle control data comprising one or more commands for controlling the vehicle;
inputting the vehicle state data and the vehicle control data into a first model comprising a neural network modeling a steering system associated with the vehicle;
inputting the vehicle state data and the vehicle control data into a deterministic and repeatable second model representing a first principle model of the steering system associated with the vehicle;
determining, based at least in part on a first output from the neural network and a second output from the second model, a predicted vehicle state comprising one or more of a predicted yaw, a predicted steering rate, a predicted position, a predicted orientation, a predicted speed, or a predicted acceleration of the vehicle;
determining, based at least in part on the predicted vehicle state, a control signal for controlling the vehicle; and
controlling, based at least in part on the control signal, a steering actuator associated with the steering system to control steering of the vehicle to follow the reference trajectory.

2. The method of claim 1 , wherein:
the neural network and the second model form a neural network block, and
the first model comprises a plurality of neural network blocks.

3. The method of claim 2 , wherein a number of the plurality of neural network blocks is determined based, at least in part, on a time window for a receding horizon.

4. The method of claim 2 , wherein the neural network comprises a recurrent neural network.

5. The method of claim 1 , wherein the vehicle state data comprises an indication of one or more of a steering angle or a lateral velocity of the vehicle; and
the vehicle state data represent dynamics of the vehicle.

6. The method of claim 1 , wherein the second output is based at least in part on a model of steering dynamics of the vehicle.

7. The method of claim 1 , wherein the control signal comprises control torque data and the method further comprises:
determining, based at least in part on the control torque data and tracking control torque data, command torque data;
determining, based at least in part on the command torque data, control command data for the steering actuator; and
wherein the steering actuator actuates based at least in part on the control command data.

8. One or more non-transitory computer-readable media storing instructions that, when executed, cause one or more processors to perform operations comprising:
receiving a reference trajectory associated with a vehicle, wherein the vehicle is an autonomous vehicle;
receiving sensor data from a sensor associated with the vehicle;
determining, based at least in part on the sensor data and the reference trajectory, state data comprising one or more of a yaw, a steering rate, a position, an orientation, a speed, or an acceleration of the vehicle;
determining, based at least in part on the sensor data, the reference trajectory, and the state data, control data comprising one or more commands for controlling the vehicle
inputting the state data and the control data into a first model, wherein the first model includes a neural network modeling a steering system of the vehicle, and a deterministic and repeatable second model representing a first principle model of the steering system of the vehicle;
determining, based at least in part on a first output from the neural network and a second output from the second model, a predicted state comprising one or more of a predicted yaw, a predicted steering rate, a predicted location, a predicted orientation, a predicted speed or a predicted acceleration of the vehicle;
determining, based at least in part on the predicted state, a control signal for controlling the vehicle; and
causing, based at least in part on the control signal, a steering actuator associated with the steering system of the vehicle to actuate and assist in controlling steering of the vehicle to follow the reference trajectory.

9. The one or more non-transitory computer-readable media of claim 8 , wherein:
the neural network and the second model form a neural network block, and
the first model comprises a plurality of neural network blocks.

10. The one or more non-transitory computer-readable media of claim 9 , wherein a number of the plurality of neural network blocks is determined based, at least in part, on a time window for a receding horizon.

11. The one or more non-transitory computer-readable media of claim 9 , wherein
the second output is based at least in part on a model of steering motion of the steering system of the vehicle, and
the predicted state is fed back to the neural network.

12. The one or more non-transitory computer-readable media of claim 9 , wherein:
the neural network is trained by a cost function comprising a sum of weighted squares of the state data, layer weights, and layer biases, and
at least one layer of the neural network comprises a hyperbolic tangent.

13. The one or more non-transitory computer-readable media of claim 8 , wherein the state data includes an indication of one or more of a steering angle, or a lateral velocity of the vehicle.

14. A vehicle comprising:
a steering assembly;
one or more sensors;
one or more processors; and
one or more non-transitory computer-readable media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
receiving a reference trajectory associated with the vehicle, wherein the vehicle is an autonomous vehicle;
receiving sensor data from the one or more sensors;
determining, based at least in part on the sensor data and the reference trajectory, state data associated with a steering system, the state data comprising one or more of a yaw, a steering rate, a position, an orientation, a speed, or an acceleration of the vehicle;
determining, based at least in part on the sensor data, the reference trajectory, and the state data, control data comprising one or more commands for controlling the vehicle;
inputting the reference trajectory, the state data, and the control data into a first model, the first model comprising a neural network modeling the steering system, and a deterministic and repeatable second model representing a first principle model of the steering system of the vehicle;
determining, based at least in part on a first output from the neural network and a second output from the second model , a predicted state comprising one or more of a predicted yaw, a predicted steering rate, a predicted position, a predicted orientation, a predicted speed, or a predicted acceleration associated with the vehicle;
determining a control signal for controlling the vehicle, based at least in part on the predicted state; and
controlling, based at least in part on the control signal, a steering actuator associated with the steering system to cause the steering actuator to exert a torque on the steering assembly to follow the reference trajectory.

15. The vehicle of claim 14 , wherein the one or more commands comprise one or mor

Assignees

• Zoox Inc

Inventors

• Funke Joseph
• Garimella Gowtham
• Kobilarov Marin
• Wang Chuang

Classifications

• B62D6/003 Primary

  in order to control vehicle yaw movement, i.e. around a vertical axis (B62D6/007 takes precedence) · CPC title

• B62D5/046

  Controlling the motor · CPC title

• B62D6/001

  the torque NOT being among the input parameters · CPC title

• G05D1/0088

  characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours (using knowledge based models G06N5/00) · CPC title

• G05D1/0221

  involving a learning process · CPC title