CN-116736243-B - Stable radar tracking speed initialization using multiple hypotheses

Abstract

This document describes an object tracker that performs stable velocity initialization for radar tracking, including when only a sparse radar point cloud is available, using a variety of assumptions. In the case where only a single point is scanned at a time, the tracker creates multiple hypotheses for the direction and speed of the object. A least squares function may be applied to each hypothesis to derive each respective initial velocity, which is tracked using a kalman filter during a hypothesis tracking period. As hypotheses are initialized and tracked over each hypothesis tracking period, their tracking error scores are calculated. Based on the hypothesized tracking error scores, hypotheses with low evidence are discarded during the hypothesized tracking period. When the hypothesis tracking period ends, the hypothesis with high evidence initializes the speed of tracking. Parallel hypothesis evaluation enables the tracker to quickly and accurately initialize the speed by selecting only the best hypothesis, which may enable safer driving.

Inventors

• LIU ZIXIN

Assignees

• 安波福技术有限公司

Dates

Publication Date

20260512

Application Date

20230310

Priority Date

20220311

Claims (20)

1. 1. A system, the system comprising a processor configured to: obtaining point cloud sensor data indicative of a return of a signal reflected from an object in an environment, and Using the point cloud sensor data to establish tracking of an object in the environment, including initializing a velocity measurement of the object by: Creating a plurality of hypotheses for a predicted movement of the object; determining a tracking error score for each of the plurality of hypotheses, the tracking error score calculated as accumulated location and distance rate-of-change errors associated with one or more points of the point cloud sensor data associated with the tracking; For each hypothesis of the plurality of hypotheses, determining an initial speed and a first associated level of evidence of the initial speed supporting the hypothesis, the first associated level of evidence corresponding to a tracking error score for the hypothesis; Generating a fused hypothesis by combining the multiple hypotheses, the initial speed of the fused hypothesis and the first associated evidence level based on a set of initial speeds of the multiple hypotheses and the first associated evidence level; including the fused hypothesis in the plurality of hypotheses, and In response to including the fused hypothesis in the plurality of hypotheses, selecting a first best hypothesis based on a first associated evidence level of the plurality of hypotheses, the first best hypothesis having a lowest associated tracking error score of the plurality of hypotheses; time updating an initial velocity for each hypothesis of the plurality of hypotheses; For each hypothesis of the plurality of hypotheses, determining a second associated evidence level supporting an updated initial velocity for the hypothesis, the second associated evidence level corresponding to an updated tracking error score for the hypothesis; removing from the plurality of hypotheses any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy an evidence threshold; In response to determining that more than two hypotheses remain in the plurality of hypotheses after canceling any hypothesis in the plurality of hypotheses having values of the second associated evidence level that do not satisfy the evidence threshold: For each of the remaining plurality of hypotheses, performing a measurement update of the predicted movement of the object, and Selecting a second best hypothesis to replace the previously selected first best hypothesis based on a second associated evidence level of the remaining plurality of hypotheses, the second best hypothesis having a lowest associated tracking error score of the remaining plurality of hypotheses after the measurement update, and Responsive to determining that only two hypotheses of the plurality of hypotheses remain after canceling any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy the evidence threshold: Selecting a third best hypothesis to replace the previously selected first or second best hypothesis based on a second associated evidence level of the remaining two hypotheses, the third best hypothesis having the lowest associated tracking error score of the remaining two hypotheses, and Outputting the tracking of the object for a vehicle system, the tracking comprising a speed parameter initialized to an initial speed of the third best assumption.
2. 2. The system of claim 1, wherein the processor is further configured for, prior to only two hypotheses remaining: time updating an initial velocity for each hypothesis of the plurality of hypotheses; For each hypothesis in the plurality of hypotheses, determining a second associated level of evidence supporting an updated initial velocity for the hypothesis, and Any hypothesis of the plurality of hypotheses having a value of the second associated evidence level that does not satisfy the evidence threshold is eliminated from the plurality of hypotheses.
3. 3. The system of claim 1, wherein the second best hypothesis for replacing the first best hypothesis previously selected is selected further in response to determining that a hypothesis tracking period ends.
4. 4. The system of claim 3, wherein the hypothesis tracking period comprises a plurality of frames of a radar system from which the point cloud sensor data is obtained.
5. 5. The system of claim 4, wherein the hypothesis tracking period comprises approximately fifteen frames of the radar system.
6. 6. The system of claim 1, wherein the processor is configured for determining, for each hypothesis of the plurality of hypotheses, a first associated level of evidence supporting an initial velocity for the hypothesis by determining accumulated location and range rate errors for one or more points of the point cloud sensor data.
7. 7. The system of claim 1, wherein the processor is configured for using a constant motion model to make temporal or measurement updates to the plurality of hypotheses.
8. 8. The system of claim 1, wherein the evidence threshold comprises a calculated evidence ratio for each of the plurality of hypotheses.
9. 9. The system of claim 8, wherein the evidence ratios calculated for each of the plurality of hypotheses comprise unique evidence ratios among all of the plurality of hypotheses.
10. 10. The system of claim 1, wherein the point cloud sensor data comprises point cloud radar data.
11. 11. A method, the method comprising: Obtaining, by an object tracker, point cloud sensor data from a radar system, the point cloud sensor data being indicative of radar echoes reflected from objects in an environment, and Using the point cloud sensor data to establish tracking of an object in the environment, including initializing a velocity measurement of the object by: Creating a plurality of hypotheses for a predicted movement of the object; determining a tracking error score for each of the plurality of hypotheses, the tracking error score calculated as accumulated location and distance rate-of-change errors associated with one or more points of the point cloud sensor data associated with the tracking; For each hypothesis of the plurality of hypotheses, determining an initial speed and a first associated level of evidence of the initial speed supporting the hypothesis, the first associated level of evidence corresponding to a tracking error score for the hypothesis; Generating a fused hypothesis by combining the multiple hypotheses, the initial speed of the fused hypothesis and the first associated evidence level based on a set of initial speeds of the multiple hypotheses and the first associated evidence level; including the fused hypothesis in the plurality of hypotheses, and In response to including the fused hypothesis in the plurality of hypotheses, a first best hypothesis is selected for initializing the speed measurement of the object based on a first associated evidence level of the plurality of hypotheses, the first best hypothesis having a lowest associated tracking error score of the plurality of hypotheses.
12. 12. The method of claim 11, further comprising: time updating an initial velocity for each hypothesis of the plurality of hypotheses; For each hypothesis of the plurality of hypotheses, determining a second associated evidence level supporting an updated initial velocity for the hypothesis, the second associated evidence level corresponding to an updated tracking error score for the hypothesis; removing from the plurality of hypotheses any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy an evidence threshold; In response to determining that more than two hypotheses remain in the plurality of hypotheses after canceling any hypothesis in the plurality of hypotheses having values of the second associated evidence level that do not satisfy the evidence threshold: For each of the remaining plurality of hypotheses, performing a measurement update of the predicted movement of the object, and A second best hypothesis is selected to replace the previously selected first best hypothesis based on a second associated evidence level of the remaining plurality of hypotheses.
13. 13. The method of claim 11, further comprising: time updating an initial velocity for each hypothesis of the plurality of hypotheses; For each hypothesis of the plurality of hypotheses, determining a second associated evidence level supporting an updated initial velocity for the hypothesis, the second associated evidence level corresponding to an updated tracking error score for the hypothesis; removing from the plurality of hypotheses any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy an evidence threshold; Responsive to determining that only two hypotheses of the plurality of hypotheses remain after canceling any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy the evidence threshold: Selecting a third best hypothesis to replace the previously selected first or second best hypothesis based on a second associated evidence level of the remaining two hypotheses, the third best hypothesis having the lowest associated tracking error score of the remaining two hypotheses, and Outputting the tracking of the object for a vehicle system, the tracking comprising a speed parameter initialized to an initial speed of the third best assumption.
14. 14. The method of claim 11, further comprising: time updating an initial velocity for each hypothesis of the plurality of hypotheses; For each hypothesis of the plurality of hypotheses, determining a second associated evidence level supporting an updated initial velocity for the hypothesis, the second associated evidence level corresponding to an updated tracking error score for the hypothesis; removing from the plurality of hypotheses any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy an evidence threshold; In response to determining that more than two hypotheses remain in the plurality of hypotheses after canceling any hypothesis in the plurality of hypotheses having values of the second associated evidence level that do not satisfy the evidence threshold: For each of the remaining plurality of hypotheses, performing a measurement update of the predicted movement of the object, and Selecting a second best hypothesis to replace the previously selected first best hypothesis based on a second associated evidence level of the remaining plurality of hypotheses, the second best hypothesis having a lowest associated tracking error score of the remaining plurality of hypotheses after the measurement update, and Responsive to determining that only two hypotheses of the plurality of hypotheses remain after canceling any hypothesis of the plurality of hypotheses having a value of a second associated evidence level that does not satisfy the evidence threshold: Selecting a third best hypothesis to replace the previously selected first or second best hypothesis based on a second associated evidence level of the remaining two hypotheses, the third best hypothesis having the lowest associated tracking error score of the remaining two hypotheses, and Outputting the tracking of the object for a vehicle system, the tracking comprising a speed parameter initialized to an initial speed of the third best assumption.
15. 15. The method of claim 14, wherein the second best hypothesis for replacing the first best hypothesis previously selected is selected further in response to determining that a hypothesis tracking period expires.
16. 16. The method of claim 15, wherein the hypothesis tracking period comprises a plurality of frames of the radar system.
17. 17. The method of claim 14, wherein the evidence threshold comprises a unique evidence ratio calculated for each of the plurality of hypotheses.
18. 18. The method of claim 11, wherein the object tracker is configured to determine, for each hypothesis of the plurality of hypotheses, a first associated evidence level supporting an initial velocity for the hypothesis by determining a cumulative location and distance rate of change error for an associated portion of the point cloud sensor data.
19. 19. The method of claim 11, wherein the object tracker is configured to use a constant motion model to make temporal or measurement updates to the plurality of hypotheses.
20. 20. A computer readable storage medium comprising instructions that, when executed, cause a processor of a radar system to execute an object tracker configured to perform the method of any of claims 11-19.

Description

Stable radar tracking speed initialization using multiple hypotheses Background A perception system for a vehicle (e.g., an advanced safety or autopilot system) may rely on the output of a radar tracker. Each trace characterizes a point cloud of radar echoes detected over multiple frames, which may be grouped to represent individual objects. The initial velocity of tracking may be derived based on point cloud detection using the positional variance of points measured over multiple frames. Note that the tracking velocity measurement is initialized to be as accurate as possible, otherwise tracking splits may occur. For example, if the velocity of the object is incorrectly initialized, the reported object position may deviate from its true position. For more distant objects, a sparse point cloud (e.g., a single point) may be available to track all information generated, which makes speed initialization difficult. To address this problem, filters or advanced algorithms may be used, including linear least squares based algorithms such as iterative least squares or normalized estimated error squares. However, the results of using these techniques may be too unstable to be used with vehicle controls that desire tracking accuracy to ensure safety. Furthermore, even if an object can be detected at a great distance from a single point back, erroneous decisions made due to the use of erroneous assumptions about the object's position, direction or speed may propagate downstream to the tracked user, which may lead to unsafe or uncomfortable driving. Existing procedures at least fail to provide consistent or stable results, resulting in these problems, which reduce security because objects cannot be accurately tracked. Disclosure of Invention This document describes techniques and systems for stable radar tracking speed initialization using multiple hypotheses. In one example, a method includes obtaining, by an object tracker, point cloud sensor data from a radar system, the point cloud sensor data indicative of radar echoes reflected from objects in an environment, and establishing tracking of the objects in the environment using the point cloud sensor data. The method includes initializing a speed measurement of an object by creating multiple hypotheses for a predicted movement of the object, determining, for each of the multiple hypotheses, an initial speed and a first associated evidence level supporting the initial speed of the hypothesis, generating, by combining the multiple hypotheses, a fused hypothesis, the initial speed and the first associated evidence level of which are based on a set of the initial speed and the first associated evidence level of the multiple hypotheses, including the fused hypothesis in the multiple hypotheses, and selecting, in response to including the fused hypothesis in the multiple hypotheses, a first best hypothesis for initializing the speed measurement of the object based on the first associated evidence level of the multiple hypotheses. In some examples, the method further includes temporally updating an initial velocity for each of the plurality of hypotheses, determining, for each of the plurality of hypotheses, a second associated evidence level supporting the updated initial velocity for the hypothesis, and eliminating from the plurality of hypotheses any hypothesis having a value for the second associated evidence level that does not satisfy the evidence threshold. Further, in response to determining that more than two hypotheses of the plurality of hypotheses remain after eliminating any hypothesis of the plurality of hypotheses having values of the second associated evidence level that do not satisfy the evidence threshold, the method further includes, for each hypothesis of the remaining plurality of hypotheses, performing a measurement update on the predicted movement of the object, and selecting a second best hypothesis to replace the previously selected first best hypothesis based on the second associated evidence level of the remaining plurality of hypotheses. In response to determining that only two hypotheses remain in the plurality of hypotheses after eliminating any hypothesis in the plurality of hypotheses having values of the second associated evidence level that do not satisfy the evidence threshold, the method further includes selecting a third best hypothesis to replace the previously selected first best hypothesis or second best hypothesis based on the remaining second associated evidence levels of the two hypotheses, and tracking a speed parameter including an initial speed initialized to the third best hypothesis for the vehicle system output tracking of the object. By implementing these and other examples contemplated by the present disclosure, stable radar tracking speed initialization using multiple hypotheses may be implemented to initialize speed measurements more accurately than using other radar tracking techniques, even where only a single point ret