Method and system for human-like vehicle control prediction in autonomous driving vehicles

1. A method implemented on a computer having at least one processor, a storage, and a communication platform for human-like vehicle control for an autonomous vehicle, comprising:

receiving information related to a target motion to be achieved by the autonomous vehicle, wherein the information includes a current vehicle state of the autonomous vehicle;

generating a first vehicle control signal with respect to the target motion and the current vehicle state in accordance with a vehicle kinematic model;

generating a second vehicle control signal, in accordance with a human-like vehicle control model, with respect to the first vehicle control signal and contextual information associated with the autonomous vehicle, wherein

the second vehicle control signal adjusts the first vehicle control signal to achieve human-like vehicle control behavior, and

wherein the human-like vehicle control model comprises a plurality of human-like vehicle control sub-models, each sub-model corresponding to a group of multiple drivers having similar driving characteristics, and wherein the human-like vehicle control model is established by:

recording human driving data of multiple drivers, each record comprising training data including at least corresponding vehicle state, vehicle control data, and environment data that characterizes a surrounding condition under which the vehicle control data yielded the corresponding vehicle state;

generating, for each record of human driving data, a vehicle kinematic model based vehicle control signal based on the corresponding vehicle state and the vehicle control data of the record in accordance with the vehicle kinematic model;

updating, via machine learning, one of the plurality of human-like vehicle control sub-models based on the training data of each record and the vehicle kinematic model based vehicle control signal, such that at least the one of the plurality of human-like vehicle control sub-models is used to provide the achieved human-like vehicle control behavior.

2. The method of claim 1 , wherein the recorded human driving data is based on manual driving data.

3. The method of claim 2 , wherein the human driving data comprises lane following and lane changing behaviors.

4. The method of claim 3 , wherein the environment data provides context information related to a driving scenario.

5. The method of claim 4 , wherein the context information specifies at least one of road condition and weather condition associated with the autonomous vehicle.

6. The method of claim 1 , wherein the contextual information associated with the vehicle includes at least one of environmental data of the vehicle and information related to passengers of the vehicle.

7. The method of claim 1 , further comprising:

controlling a degree of adjusting the first vehicle control signal based on one or more fusion constraints.

8. A machine readable and non-transitory medium having information stored thereon for human-like vehicle control for an autonomous vehicle, wherein the information, when read by the machine, causes the machine to perform the following:

receiving information related to a target motion to be achieved by the autonomous vehicle, wherein the information includes a current vehicle state of the autonomous vehicle;

generating a first vehicle control signal with respect to the target motion and the current vehicle state in accordance with a vehicle kinematic model;

generating a second vehicle control signal, in accordance with a human-like vehicle control model, with respect to the first vehicle control signal and contextual information associated with the autonomous vehicle, wherein

the second vehicle control signal adjusts the first vehicle control signal to achieve human-like vehicle control behavior, and

wherein the human-like vehicle control model comprises a plurality of human-like vehicle control sub-models, each sub-model corresponding to a group of multiple drivers having similar driving characteristics, and wherein the human-like vehicle control model is established by:

recording human driving data of multiple drivers, each record comprising training data including at least corresponding vehicle state, vehicle control data, and environment data that characterizes a surrounding condition under which the vehicle control data yielded the corresponding vehicle state;

generating, for each record of human driving data, a vehicle kinematic model based vehicle control signal based on the corresponding vehicle state and the vehicle control data of the record in accordance with the vehicle kinematic model;

updating, via machine learning, one of the plurality of human-like vehicle control sub-models based on the training data of each record and the vehicle kinematic model based vehicle control signal, such that at least the one of the plurality of human-like vehicle control sub-models is used to provide the achieved human-like vehicle control behavior.

9. The medium of claim 8 , wherein the recorded human driving data is based on manual driving data.

10. The medium of claim 9 , wherein the human driving data comprises lane following and lane changing behaviors.

11. The medium of claim 10 , wherein the environment data provides context information related to a driving scenario.

12. The medium of claim 11 , wherein the context information specifies at least one of road condition and weather condition associated with the autonomous vehicle.

13. A system for human-like vehicle control for an autonomous vehicle, comprising:

a human-like vehicle control signal generator implemented by a processor and configured for receiving information for a target motion to be achieved by a vehicle, wherein the information for the target motion includes a current vehicle state of the vehicle;

a kinematic vehicle control signal inference engine implemented by the processor and configured for generating a first vehicle control signal with respect to the target motion and the current vehicle state in accordance with a vehicle kinematic model;

a human-like vehicle control model based fusion unit implemented by the processor and configured for generating a second vehicle control signal, in accordance with a human-like vehicle control model, with respect to the first vehicle control signal and contextual information associated with the vehicle, wherein

the second vehicle control signal adjusts the first vehicle control signal to achieve human-like vehicle control behavior, and

further comprising a human-like vehicle control model generator implemented by the processor and configured for generating the human-like vehicle control model which comprises a plurality of human-like vehicle control sub-models, each sub-model corresponding to a group of multiple drivers having similar driving characteristics, and wherein the human-like vehicle control model is established by:

recording human driving data of multiple drivers, each record comprising training data including at least corresponding vehicle state, vehicle control data, and environment data that characterizes a surrounding condition under which the vehicle control data yielded the corresponding vehicle state;

generating, for each record of human driving data, a vehicle kinematic model based vehicle control signal based on the corresponding vehicle state and the vehicle control data of the record in accordance with the vehicle kinematic model;

updating, via machine learning, one of the plurality of human-like vehicle control sub-models based on the training data of each record and the vehicle kinematic model based vehicle control signal, such that at least the one of the plurality of human-like vehicle control sub-models is used to provide the achieved human-like vehicle control behavior.

14. The system of claim 13 , wherein the recorded human driving data is based on manual driving data.

15. The system of claim 14 , wherein the human driving data comprises lane following and lane changing behaviors.

16. The system of claim 15 , wherein the environment data provides context information related to a driving scenario.

17. The system of claim 16 , wherein the context information specifies at least one of road condition and weather condition associated with the autonomous vehicle.