Granted Patent B2

US 12,516,501 · App. 18/302,429 · Granted Jan 6, 2026

Inventors: Ruitao Song (Sunnyvale, CA); Liangjun Zhang (Sunnyvale, CA)

Assignee: Baidu USA LLC

E02F9/2045E02F9/205E02F9/26G06N7/01

Abstract

Claims (37)

1 . A computer-implemented method for controlling a trajectory of a continuous track vehicle based upon vehicle trajectory input, comprising:

determining, via a controller, (i) a first command signal for a first locomotion component of a vehicle and (ii) a second command signal for a second locomotion component of the vehicle;

based upon a terrain classification, selecting (i) a first pre-trained model for the first locomotion component of the vehicle and (ii) a second pre-trained model for the second locomotion component of the vehicle;

determining (i) a first signal for the first locomotion component of the vehicle by utilizing the first command signal and the second command signal as input to the first pre-trained model and (ii) a second signal for the second locomotion component of the vehicle by utilizing the first command signal and the second command signal as input to the second pre-trained model; and

controlling (i) the first locomotion component of the vehicle using the first signal and (ii) the second locomotion component of the vehicle using the second signal;

wherein the controller comprises a nonlinear model predictive controller that determines the first command signal and the second command signal based upon vehicle state input and based upon the vehicle trajectory input; and

wherein the first locomotion component and the second locomotion component each comprise continuous track locomotion components.

2 . The computer-implemented method of claim 1 , wherein the first signal comprises a first pulse-width modulated signal, and wherein the second signal comprises a second pulse-width modulated signal.

3 . The computer-implemented method of claim 1 , wherein the terrain classification is obtained from a terrain classification model that uses image data, inertial measurement unit data, or both as inputs.

4 . The computer-implemented method of claim 3 , wherein selecting the first pre-trained model and the second pre-trained model is further based upon the image data or the inertial measurement data.

5 . The computer-implemented method of claim 1 , wherein the first model and the second model comprise a set of Gaussian Process Models (GPMs) in which each GPM is correlated to a terrain type.

6 . A computer-implemented method for facilitating trajectory control of a continuous track vehicle based upon vehicle trajectory input, comprising:

receiving a terrain classification;

accessing a set of pre-trained models, each pre-trained model of the set of pre-trained models being configured to determine a signal in response to input command signals, the set of pre-trained models comprising a plurality of subsets of pre-trained models, wherein each subset of pre-trained models is associated with a respective known terrain;

each subset of pre-trained models comprising a respective first pre-trained model of the subset and a respective second pre-trained model of the subset trained using trajectory data associated with one or more vehicles traversing the respective known terrain, wherein each respective first pre-trained model of the subset is associated with a first locomotion component that is associated with the one or more vehicles, and wherein each respective second pre-trained model of the subset is associated with a second locomotion component that is associated with the one or more vehicles;

selecting a particular subset of pre-trained models from the plurality of subsets of pre-trained models based upon at least the terrain classification;

receiving (i) a first command signal associated with the first locomotion component and (ii) a second command signal associated with the second locomotion component; and

determining (i) a first signal based on the respective first pre-trained model of the selected subset of pre-trained models and (ii) a second signal based on the respective second pre-trained model of the selected subset of pre-trained models; and

controlling the first locomotion component based on the determined first signal and controlling the second locomotion component based on the determined second signal;

wherein the first command signal and the second command signal are obtained from a nonlinear model predictive controller; and

wherein the nonlinear model predictive controller determines the first command signal and the second command signal based upon vehicle state input and based upon the vehicle trajectory input; and

wherein the first locomotion component and the second locomotion component each comprise continuous track locomotion components.

7 . The computer-implemented method of claim 6 , wherein the first pre-trained model utilizes both the first command signal and the second command signal as input to determine the first signal, and wherein the second pre-trained model utilizes both the first command signal and the second command signal as input to determine the second signal.

8 . The computer-implemented method of claim 6 , wherein the set of pre-trained models comprises a set of Gaussian Process Models (GPMs).

9 . The computer-implemented method of claim 6 , wherein the first signal comprises a first pulse-width modulated signal, and wherein the second signal comprises a second pulse-width modulated signal.

10 . The computer-implemented method of claim 6 , wherein the terrain classification is obtained from a terrain classification model that uses image data, inertial measurement data, or both as inputs.

11 . The computer-implemented method of claim 10 , wherein selecting the particular subset of pre-trained models is further based upon the image data or the inertial measurement data.

12 . A system for controlling a trajectory of a continuous track vehicle based upon vehicle trajectory input, comprising:

one or more processors; and

a non-transitory computer-readable medium or media comprising one or more sets of instructions which, when executed by at least one of the one or more processors, causes steps to be performed comprising:

determining, via a controller, (i) a first command signal for a first locomotion component of a vehicle and (ii) a second command signal for a second locomotion component of the vehicle;

selecting (i) a first pre-trained model for the first locomotion component of the vehicle and (ii) a second pre-trained model for the second locomotion component of the vehicle;

determining (i) a first signal for the first locomotion component of the vehicle by utilizing the first command signal and the second command signal as input to the first pre-trained model and (ii) a second signal for the second locomotion component of the vehicle by utilizing the first command signal and the second command signal as input to the second pre-trained model; and

controlling (i) the first locomotion component of the vehicle using the first signal and (ii) the second locomotion component of the vehicle using the second signal;

wherein the controller comprises a nonlinear model predictive controller that determines the first command signal and the second command signal based upon vehicle state input and the vehicle trajectory input; and

wherein the first locomotion component and the second locomotion component each comprise continuous track locomotion components.

13 . The system of claim 12 , wherein the first pre-trained model and the second pre-trained model are selected based upon a terrain classification, and wherein the terrain classification is obtained from a terrain classification model that uses image data, inertial measurement unit data, or both as inputs.

Assignments (1)

Continuity (1)

References Cited (101)