Granted Patent B2

US 12,554,257 · App. 18/631,700 · Granted Feb 17, 2026

Inventors: Wutu Lin (San Diego, CA); Liu Liu (San Diego, CA); Xing Sun (San Diego, CA); Kai-Chieh Ma (San Diego, CA); Zijie Xuan (San Diego, CA); Yufei Zhao (San Diego, CA)

Assignee: CreateAI, Inc.

G05D1/0088B60W40/09G05D1/81B60W2720/103

Abstract

Claims (40)

1 . A method, comprising:

receiving a vehicle control command prior to controlling an autonomous vehicle to perform the vehicle control command;

comparing vehicle operational data to simulated vehicle data and updating a driving behavior model according to the comparing;

validating or modifying the vehicle control command based on the updated driving behavior model; and

causing the autonomous vehicle to perform the vehicle control command according to the validated or modified vehicle control command.

2 . The method of claim 1 , wherein the driving behavior model is trained by a reinforcement learning process comprising training by simulation to generate data that is compared to data corresponding to human driving behaviors during the simulation.

3 . The method of claim 1 , wherein the driving behavior model is trained by a reinforcement learning process comprising training by comparing data captured while the autonomous vehicle is in operation on a road to data corresponding to human driving behaviors captured by sensors of the autonomous vehicle.

4 . The method of claim 1 , wherein the comparing comprises determining a deviation between the vehicle operational data at time steps and the simulated vehicle data at corresponding time steps.

5 . The method of claim 3 , wherein updating the driving behavior model comprises performing the reinforcement learning process in iterations that modify parameters of the driving behavior model.

6 . The method of claim 1 , wherein the driving behavior model is based on driving parameters comprising:

a speed parameter,

a braking parameter, or

a steering angle parameter of the autonomous vehicle.

7 . The method of claim 5 , wherein the comparing comprises determining a deviation between the vehicle operational data and the simulated vehicle data relative to a threshold, wherein the vehicle operational data comprises a trajectory of the autonomous vehicle during a first iteration of the reinforcement learning process.

8 . The method of claim 1 , wherein the vehicle operational data is aggregated from data collected from a population of vehicles and drivers.

9 . An apparatus, comprising:

at least one processor; and

at least one memory including executable instructions that, when executed, cause the at least one processor to perform operations comprising:

receiving a vehicle control command prior to controlling an autonomous vehicle to perform the vehicle control command;

comparing vehicle operational data to simulated vehicle data and updating a driving behavior model according to the comparing;

validating or modifying the vehicle control command based on the updated driving behavior model; and

causing the autonomous vehicle to perform the vehicle control command according to the validated or modified vehicle control command.

10 . The apparatus of claim 9 , wherein the driving behavior model is trained by a reinforcement learning process comprising training by simulation to generate data that is compared to data corresponding to human driving behaviors during the simulation.

11 . The apparatus of claim 9 , wherein the driving behavior model is trained by a reinforcement learning process comprising training by comparing data captured while the autonomous vehicle is in operation on a road to data corresponding to human driving behaviors captured by sensors of the autonomous vehicle.

12 . The apparatus of claim 9 , wherein the driving behavior model is trained to identify standards of driving behavior determined by training a neural network or generating a rules set.

13 . The apparatus of claim 9 , wherein the modifying the vehicle control command modifies driving parameters in the vehicle control command.

14 . The apparatus of claim 13 , wherein the driving behavior model is based on driving parameters comprising:

a speed parameter,

a braking parameter, or

a steering angle parameter of the autonomous vehicle.

15 . The apparatus of claim 10 , wherein updating the driving behavior model comprises performing the reinforcement learning process in iterations that modify parameters of the driving behavior model.

16 . The apparatus of claim 15 , wherein the vehicle operational data is aggregated from data collected from a population of vehicles and drivers.

17 . A non-transitory machine-readable storage medium including instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:

receiving a vehicle control command prior to controlling an autonomous vehicle to perform the vehicle control command;

comparing vehicle operational data to simulated vehicle data and updating a driving behavior model according to the comparing;

validating or modifying the vehicle control command based on the updated driving behavior model; and

causing the autonomous vehicle to perform the vehicle control command according to the validated or modified vehicle control command.

18 . The non-transitory machine-readable storage medium of claim 17 , wherein the driving behavior model is trained by a reinforcement learning process comprising training by simulation to generate data that is compared to data corresponding to human driving behaviors during the simulation.

19 . The non-transitory machine-readable storage medium of claim 17 , wherein the driving behavior model is trained by a reinforcement learning process comprising training by comparing data captured while the autonomous vehicle is in operation on a road to data corresponding to human driving behaviors captured by sensors of the autonomous vehicle.

20 . The non-transitory machine-readable storage medium of claim 17 , wherein the comparing comprises determining a deviation between the vehicle operational data at time steps and the simulated vehicle data at corresponding time steps.

Assignments (3)

Continuity (4)

References Cited (189)