METHOD AND SYSTEM FOR TRACKING BY A VEHICLE
A method can include: determining a set of measurements; determining a visibility representation; determining a set of observations; generating a set of hypotheses; determining a set of tracks; and/or any other suitable elements. Additionally or alternatively, the method can optionally include planning a trajectory for the vehicle and/or any other suitable elements. The method functions to track objects and the uncertainty of existence thereof in the vehicle's environment over time.
1 . A method comprising:
determining a set of measurements with a vehicle sensor suite;
determining an environmental visibility map based on the set of measurements;
using the environmental visibility map and a prior object track, estimating a probability of detection of an object in the environment, wherein the object is associated with the prior object track;
based on the set of measurements, determining an object detection;
updating the prior object track to yield a current object track;
based on the object detection and the probability of detection of the object, determining a probability of existence of the current object track; and
controlling a vehicle based on the current object track and the probability of existence.
2 . The method of claim 1 , wherein determining the probability of existence comprises using Poisson Multi-Bernoulli Mixturing (PMBM).
3 . The method of claim 1 , wherein estimating the probability of detection of the object comprises:
predicting a position of the object using a trajectory of the current object track; and
sampling the environmental visibility map using the predicted position of the object.
4 . The method of claim 1 , wherein the visibility map is a binary map, wherein the probability of detection is non-binary.
5 . The method of claim 4 , wherein determining the probability of detection comprises sampling multiple values within a boundary hull associated with the current object track.
6 . The method of claim 1 , further comprising normalizing a semantic classification of the current object track based on the probability of existence, and wherein controlling the vehicle comprises using the semantic classification.
7 . The method of claim 1 , further comprising:
determining a next set of measurements with the vehicle sensor suite;
determining a next environmental visibility map based on the next set of measurements;
determining that the object is not detected in the next set of measurements;
determining a next probability of detection of the object using the next environmental visibility map and the current object track; and
responsive to the object not being detected, updating the probability of existence of the current object track based on the next probability of detection of the object.
8 . The method of claim 1 , wherein updating the probability of existence comprises applying a dynamically updating a prior probability of existence.
9 . The method of claim 1 , wherein determining the environmental visibility map comprises ray tracing from Lidar scans of the set of measurements.
10 . The method of claim 1 , wherein the environmental visibility map represents visibility from a plurality of sensors of the vehicle sensor suite and is determined based on predetermined relative positions of the plurality of sensors.
11 . The method of claim 1 , wherein measurements used to determine the environmental visibility map comprise lidar measurements and measurements used to detect the object comprise camera measurements.
12 . A method comprising:
determining a set of measurements with a vehicle sensor suite;
determining an environmental visibility representation based on the set of measurements;
using the environmental visibility representation, estimating a probability of detection of an object associated with an object track;
determining that the object is undetected within the set of measurements;
responsive to the determination of the object being undetected within the set of measurements, determining a probability of existence for the object track based on the probability of detection of the object; and
controlling the autonomous vehicle based on the object track and the probability of existence.
13 . The method of claim 12 , wherein determining the probability of existence comprises applying a dynamic update to a prior probability of existence.
14 . The method of claim 12 , wherein the object is associated with multiple distinct object tracks.
15 . The method of claim 12 , wherein measurements used to determine the environmental visibility representation are in a different modality from measurements used to detect the object.
16 . The method of claim 12 , wherein estimating the probability of detection of the object comprises sampling the visibility representation at multiple indexes.
17 . The method of claim 16 , wherein the multiple indexes are based on a classification of the object.
18 . The method of claim 12 , wherein the environmental visibility representation comprises binary values, and the probability of detection is non-binary.
19 . The method of claim 12 , further comprising normalizing a prior classification of the object track based on the probability of existence.
20 . The method of claim 12 , wherein determining an environmental visibility representation based on the set of measurements comprises: detecting objects using a classically programmed object detector.