Velocity estimation and object tracking for autonomous vehicle applications

1. A method comprising:

obtaining, by a sensing system of an autonomous vehicle (AV), for each of a plurality of times, a plurality of return points, each return point comprising a velocity value and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system;

tracking a motion state vector characterizing a motion of a physical object, wherein the motion state vector comprises one or more components of an angular acceleration of the physical object, and wherein tracking the motion state vector comprises, for each of the plurality of times,

predicting a motion state vector for a next of the plurality of times;

identifying a difference between a corresponding plurality of return points obtained for the next of the plurality of times and the predicted motion state vector;

updating the predicted motion state vector in view of the identified difference;

causing a driving path of the AV to be determined in view of the tracked motion state vector characterizing the motion of the physical object; and

controlling the AV to perform a driving maneuver using the determined driving path, wherein the driving maneuver comprises changing at least one of a direction of the AV or a speed of the AV.

2. The method of claim 1 , wherein the sensing system of the AV comprises a coherent light detection and ranging device (lidar) and wherein the signal emitted by the sensing system comprises a phase-modulated or a frequency-modulated electromagnetic wave.

3. The method of claim 2 , wherein the coherent lidar is a frequency-modulated continuous-wave lidar.

4. The method of claim 1 , wherein the velocity value is a component of a velocity of the reflecting region along a direction from a transmitter of the sensing system to the reflecting region.

5. The method of claim 1 , wherein each of the plurality of times corresponds to a different cycle of a transmitter of the sensing system.

6. The method of claim 1 , further comprising determining, using the motion state vector, a representation of motion of the physical object, wherein the representation of motion comprises:

one or more components of a translational velocity of the physical object; and

one or more components of an angular velocity of the physical object.

7. The method of claim 6 , wherein determining the representation of motion comprises:

determining, based on a first plurality of return points obtained for a first set of one or more times of the plurality of times, an aggregated value representing a combination of two or more latent velocity components, and

disambiguating, using a second plurality of return points obtained for a second set of one or more times of the plurality of times, individual values of the two or more latent velocity components from the aggregated value.

8. The method of claim 1 , wherein a first sub-plurality of the plurality of return points are obtained by a first sensor of the sensing system, and wherein a second sub-plurality of the plurality of return points are obtained using a second sensor of the sensing system, the second sensor located differently than the first sensor.

9. The method of claim 1 , wherein the motion state vector further comprises one or more coordinates of the physical object and one or more components of velocity of the physical object.

10. The method of claim 1 , wherein updating the predicted motion state vector further comprises:

using a covariance matrix for the state vector.

11. A method comprising:

obtaining, by a sensing system of an autonomous vehicle (AV), a first plurality of return points, each return point of the first plurality of return points comprising a measured velocity value and one or more measured coordinates of a reflecting region that reflects a respective signal emitted by the sensing system;

estimating a state vector of an object corresponding to the first plurality of return points, the state vector comprising a translational velocity of the object, a rotational velocity of the object, and one or more components of an angular acceleration of the object;

obtaining, using the estimated state vector, an estimated plurality of return points;

obtaining an improved state vector of the object, based, at least in part, on a difference between the first plurality of return points and the estimated plurality of return points;

causing a driving path of the AV to be determined in view of the improved state vector of the object; and

controlling the AV to perform a driving maneuver using the determined driving path, wherein the driving maneuver comprises changing at least one of a direction of the AV or a speed of the AV.

12. The method of claim 11 , wherein obtaining the improved state vector comprises using an estimated covariance matrix for components of the state vector of the object.

13. The method of claim 12 , wherein estimating the state vector is based, at least in part, on a second plurality of return points, the second plurality of return points obtained at an earlier time compared with the first plurality of return points.

14. A system comprising:

a sensing system of an autonomous vehicle (AV), the sensing system to:

obtain, for each of a plurality of times, a plurality of return points, each return point comprising a velocity value and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system; and

a perception system of the AV, the perception system to:

track a motion state vector characterizing a motion of a physical object, wherein the motion state vector comprises one or more components of an angular acceleration of the physical object, and wherein to track the motion state vector, the perception system is to, for each of the plurality of times:

predict a motion state vector for a next of the plurality of times;

identify a difference between a corresponding plurality of return points obtained for the next of the plurality of times and the predicted motion state vector;

update the predicted motion state vector in view of the identified difference;

cause a driving path of the AV to be determined in view of the tracked motion state vector characterizing the motion of the physical object; and

control the AV to perform a driving maneuver using the determined driving path, wherein the driving maneuver comprises changing at least one of a direction of the AV or a speed of the AV.

15. The system of claim 14 , wherein the sensing system comprises a coherent light detection and ranging device (lidar) and wherein the signal emitted by the sensing system comprises a phase-modulated or a frequency-modulated electromagnetic wave.

16. The system of claim 14 , wherein the perception system is further to:

determine, using the motion state vector, a representation of motion of the physical object, wherein the representation of motion comprises:

one or more components of a translational velocity of the physical object; and

one or more components of an angular velocity of the physical object.

17. The system of claim 16 , wherein to determine the representation of motion the perception system is to:

determine, based on a first plurality of return points obtained for a first set of one or more times of the plurality of times, an aggregated value representing a combination of two or more latent velocity components, and

disambiguating, using a second plurality of return points obtained for a second set of one or more sensing frametimes of the plurality of sensing frametimes, individual values of the two or more latent velocity components from the aggregated value.

18. The system of claim 14 , wherein a first sub-plurality the plurality of return points are obtained by a first sensor of the sensing system, and wherein a second sub-plurality of the plurality of return points are obtained using a second sensor of the sensing system, the second sensor located differently than the first sensor.

19. A non-transitory computer-readable medium having instructions stored thereon that, when executed by a computing device, cause the computing device to:

obtain, by a sensing system of an autonomous vehicle (AV), for each of a plurality of times, a plurality of return points, each return point comprising a velocity value and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system;

track a motion state vector characterizing a motion of a physical object, wherein the motion state vector comprises one or more components of an angular acceleration of the physical object, and wherein to track the motion state vector, the computing device is to, for each of the plurality of times:

predict a motion state vector for a next of the plurality of times;

identify a difference between a corresponding plurality of return points obtained for the next of the plurality of times and the predicted motion state vector; and

update the predicted motion state vector in view of the identified difference;

cause a driving path of the AV to be determined in view of the tracked motion state vector characterizing the motion of the physical object; and

control the AV to perform a driving maneuver using the determined driving path, wherein the driving maneuver comprises changing at least one of a direction of the AV or a speed of the AV.