Classification of objects based on motion patterns for autonomous vehicle applications
Aspects and implementations of the present disclosure address shortcomings of the existing technology by enabling motion pattern-assisted object classification of objects in an environment of an autonomous vehicle (AV) by obtaining, from a sensing system of the AV, a plurality of return points, each return point comprising one or more velocity values and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system, identifying an association of the plurality of return points with an object in an environment of the AV, identifying, in view of the one or more velocity values of at least some of the plurality of return points, a type of the object or a type of a motion of the object, and causing a driving path of the AV to be determined in view of the identified type of the object.
1. A method comprising:
obtaining, by a sensing system of an autonomous vehicle (AV), a plurality of return points, each return point comprising one or more velocity values and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system;
identifying an association of the plurality of return points with an object in an environment of the AV;
segmenting, based on a first value of a translational velocity and a second value of a rotational velocity, the plurality of return points into at least a first cluster of points and a second cluster of points, wherein the first cluster of points is segmented by fitting the first cluster of points using a translational rigid-body motion associated with the first value of the translational velocity, and wherein the second cluster of points is segmented by fitting the second cluster of points using a combination of:
the translational rigid-body motion associated with the first value of the translational velocity, and
a rotational rigid-body motion associated with the second value of the rotational velocity;
identifying, using the first cluster of points and the second cluster of points, a type of the object; and
outputting a control instruction that causes one or more AV control systems to change a driving path of the AV in view of the identified type of the object.
2. The method of claim 1 , further comprising:
identifying, using at least the first value of the translational velocity, a trajectory of the object.
3. The method of claim 1 , wherein the object comprises a pedestrian.
4. The method of claim 3 , wherein the translational rigid-body motion is associated with a torso of the pedestrian and the rotational rigid-body motion is associated with at least one of an arm of the pedestrian or a leg of the pedestrian.
5. The method of claim 1 , wherein the object comprises a vehicle, wherein the first cluster of points is associated with at least a portion of a body of the vehicle, and wherein the second cluster of points is associated with a wheel of the vehicle.
6. The method of claim 5 , further comprising:
determining, based on a mismatch of the first value of the translational velocity with the second value of the rotational velocity, that the wheel of the vehicle is slipping.
7. The method of claim 1 , wherein segmenting the plurality of return points comprises;
computing, with an individual point of the plurality of return points, a distance in a coordinate-velocity space from the individual point to a centroid of the first cluster of points; and
determining association of the individual point with the first cluster of points using the computed distance in the coordinate-velocity space.
8. The method of claim 1 , wherein the object comprises a vehicle, the first cluster of points is associated with at least a portion of a body of the vehicle, and wherein the second cluster of points is associated with a door of the vehicle.
9. The method of claim 8 , wherein the rotational rigid-body motion is associated with a rotational motion of the door of the vehicle around a vertical axis.
10. 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.
11. The method of claim 10 , wherein the coherent lidar is a frequency-modulated continuous-wave lidar.
12. The method of claim 1 , wherein the one or more velocity values comprise a component of a velocity of the reflecting region along a direction from a transmitter of the sensing system to the reflecting region.
13. The method of claim 1 , wherein the one or more velocity values comprise a first velocity value obtained using a first sensor of the sensing system and a second velocity value obtained using a second sensor of the sensing system, wherein the second sensor is located differently from the first sensor.
14. A system comprising:
a sensing system of an autonomous vehicle (AV), the sensing system to:
obtain a plurality of return points, each return point comprising one or more velocity values 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:
identify an association of the plurality of return points with an object in an environment of the AV;
segment, based on a first value of a translational velocity and a second value of a rotational velocity, the plurality of return points into at least a first cluster of points and a second cluster of points, wherein the first cluster of points is segmented by fitting the first cluster of points using a translational rigid-body motion associated with the first value of the translational velocity, and wherein the second cluster of points is segmented by fitting the second cluster of points using a combination of:
the translational rigid-body motion associated with the first value of the translational velocity, and
a rotational rigid-body motion associated with the second value of the rotational velocity;
identify, using the first cluster of points and the second cluster of points, a type of the object; and
output a control instruction that causes one or more AV control systems to change a driving path of the AV in view of the identified type of the object.
15. The system of claim 14 , wherein the perception system is further to:
identify, using at least the first value of the translational velocity, a trajectory of the object.
16. The system of claim 14 , wherein the object comprises a pedestrian.
17. The system of claim 14 , wherein the translational rigid-body motion is associated with at least one of:
a torso of a pedestrian, or
a body of a vehicle; and
wherein and the rotational rigid-body motion is associated with at least one of:
an arm of the pedestrian,
a leg of the pedestrian, or
a wheel of the vehicle.
18. The system of claim 14 , wherein plurality of return points are obtained using a coherent light detection and ranging device (lidar) of the sensing system and wherein the signal emitted by the sensing system comprises a phase-modulated or a frequency-modulated electromagnetic wave.
19. A non-transitory computer-readable medium having instructions stored thereon that, when executed by a computing device, cause the computing device to:
obtain, from a sensing system of an autonomous vehicle (AV), a plurality of return points, each return point comprising one or more velocity values and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system;
identify an association of the plurality of return points with an object in an environment of the AV;
segment, based on a first value of a translational velocity and a second value of a rotational velocity, the plurality of return points into at least a first cluster of points and a second cluster of points, wherein the first cluster of points is segmented by fitting the first cluster of points using a translational rigid-body motion associated with the first value of the translational velocity, and wherein the second cluster of points is segmented by fitting the second cluster of points using a combination of:
the translational rigid-body motion associated with the first value of the translational velocity, and
a rotational rigid-body motion associated with the second value of the rotational velocity:
identify, using the first cluster of points and the second cluster of points, a type of the object; and
output a control instruction that causes one or more AV control systems to change a driving path of the AV in view of the identified type of the object.