Determining object motion and acceleration vector in a vehicle radar system

DESCRIPTION OF THE DISCLOSURE

The present disclosure relates to a vehicle radar system. For each one of a plurality of radar cycles, the vehicle radar system is arranged to provide a measured azimuth angle and a measured radial velocity for a first plurality of detections.

Many vehicle radar systems comprise radar transceivers that are arranged for generating radar signals that are transmitted, reflected and received by means of appropriate antennas comprised in the radar system. The radar signals may for example be in the form of FMCW (Frequency Modulated Continuous Wave) signals .

Radar provides important information about the environment and its participants, including moving objects and stationary environment to enable implementing automated vehicle systems such as speed control and collision prevention, as well as other such as auto alignment, online bumper fascia calibration, ego motion estimation, scene understanding, lane keeping assist (LKA) and automated driving (AD) .

At each radar cycle, the received radar signal provides a multitude of radar detections, and for each detection radial (Doppler) velocity V _d , radial distance r and azimuth angle Q are provided. The detections can be grouped as tracked objects, providing a common motion state for an extended object.

The motion state of an object describes, e.g., the distance from the ego vehicle to the object, a complete, relative or absolute, velocity vector of the object in relation to the ego vehicle, and a complete, relative or absolute, acceleration vector of the object in relation to the ego vehicle. The motion state of tracked objects is updated continuously to reflect changes in object motion. For instance, acceleration and velocity vectors comprised in the motion state will increase in magnitude if a tracked object brakes or otherwise changes its velocity.

It can be assumed that those detections that belong to a certain extended object have the same velocity vectors. Therefore, in order to be able to determine which ones of the acquired detections that have the same velocity vectors and thus can be determined to belong to a certain extended object, it is desirable to obtain complete vectors for velocity and acceleration for all the detections in as few radar cycles as possible in an uncomplicated and reliable manner.

The object of the present disclosure is thus to provide a vehicle radar system that is adapted to determine complete vectors for velocity and acceleration in as few radar cycles as possible in an uncomplicated and reliable manner.

This object is achieved by means of a vehicle radar system which, for each one of a plurality of radar cycles, is arranged to provide a measured azimuth angle and a measured radial velocity for a first plurality of detections. For each one of said plurality of radar cycles, the radar system is arranged to select one detection from said first plurality of detections for each one of two velocity components comprised in a set of components to be determined, where the two velocity components define a full velocity vector for a relative velocity, to select one detection from a second plurality of detections for each one of at least one corresponding acceleration component, and to calculate said components for the selected detections. For each one of said plurality of radar cycles, the radar system is furthermore arranged to determine a calculated radial velocity for each one of at least a part of the other detections in said first plurality of detections by means of said calculated components, to determine an error between each calculated radial velocity and the measured radial velocity, and to determine the number of inliers, where an inlier corresponds to an error that falls below an inlier threshold.

This object is also achieved by means of a method for a vehicle radar system, where, for each one of a plurality of radar cycles, the method comprises acquiring a measured azimuth angle and a measured radial velocity for a plurality of detections associated with an object, selecting one detection from said first plurality of detections for each one of two velocity components comprised in a set of components to be determined, where the two velocity components define a full velocity vector for a relative velocity, and selecting one detection from a second plurality of detections for each one of at least one corresponding acceleration component and calculating said components for the selected detections.

For each one of said plurality of radar cycles, the method further comprises determining a calculated radial velocity for each one of at least a part of the other detections in said first plurality of detections by means of said calculated components, determining an error between each calculated radial velocity and the measured radial velocity, and determining the number of inliers, where an inlier corresponds to an error that falls below an inlier threshold.

The method comprises repeating the above for all radar cycles in said plurality of radar cycles; the method then comprises choosing the set of components that results in the largest number of inliers. The radar system is furthermore arranged to choose the set of components that results in the largest number of inliers.

According to some aspects, the radar system is adapted to group those detections that correspond to inliers resulting from the chosen set of components, such that these detections form an extended object.

According to some aspects, the second plurality of detections is the same as the first plurality of detections.

According to some aspects, the second plurality of detections belongs to another radar cycle. The radar system is arranged to first calculate the velocity components for the selected detections, and then to calculate said at least one acceleration component by means of the calculated velocity components.

According to some aspects, the radar system is arranged to calculate said components by solving for these from the express on for each selected detection, where 0 _m constitutes a measured angle between a radar system reference line and a direction towards the detection in question.

According to some aspects, the radar system is adapted to substitute a _Y = a— and a _v = a—, where IlfI is the magnitude of

\V _r \ ^y \V _r \’ ' ^r ' ^¾

the relative velocity v _r , such that said components are calculated by solving for these from the expression Other examples are disclosed in the dependent claims.

A number of advantages are obtained by means of the present disclosure. Mainly, object tracking performance is improved, as well as ego dynamics estimation.

Furthermore, detections from multiple cycles can be used at the same time, which makes the algorithm robust even in case of low number of detections and in the presence of noise and outliers.

BRIEF DESCRIPTION OF THE DRAWINGS

The present disclosure will now be described more in detail with reference to the appended drawings, where:

Figure 1 shows a schematic top view of an ego vehicle and a target vehicle;

Figure 2 shows a schematic illustration of an accelerating target object;

Figure 3 shows a graphic illustration of radial Doppler velocity as a function of azimuth angle for an accelerating target object;

Figure 4 shows a flowchart for methods according to the present disclosure .

DETAILED DESCRIPTION

Figure 1 schematically shows a top view of an ego vehicle 1 arranged to run in a direction D with a velocity v _ego , here co- inciding with a system x-axis 21. The vehicle 1 comprises a vehicle radar system 3 which in turn comprises a transceiver arrangement 4 and a control unit 5. The vehicle radar system 3 is arranged to distinguish and/or resolve single targets from the surroundings by transmitting signals 2a and receiving reflected signals 2b and using a Doppler effect in a previously well-known manner, the transceiver arrangement 4 having a certain field of view 11. The vehicle radar system 3 is thus arranged to provide data regarding azimuth angle and radial velocity for a possible target object 6 by simultaneously sampling and analyzing phase and amplitude of the received signals 2b by means of Doppler signal processing in a previously known manner. The radar signals may for example be in the form of FMCW (Frequency Modulated Continuous Wave) Doppler signals operating at 77 GHz.

By means of the vehicle radar system 3, a measured radial velocity v _dm , the Doppler velocity, is determined, for example from the Doppler frequency and the Doppler phase shift. Furthermore, a measured azimuth angle 0 _m between a radial direction of observation and a radar system reference line such as the x-axis 21, as well as a measured radial distance r _m between the transceiver arrangement 4 and a detection 9, are acquired for each one of a plurality of detections 9, 20 (only a few indicated in Figure 1 for reasons of clarity) , where there is a number N of detections 9 that belong to the target object 6 among a total number M ³ N of detections 9, 20. Some of the detections 9 can be grouped as a tracked object, providing a common motion state for an extended object 10.

The motion state of an object describes, e.g., the distance from the ego vehicle 1 to a target object 6, a complete, relative or absolute, velocity vector of the object in relation to the ego vehicle, and a complete, relative or absolute, acceleration vector of the object in relation to the ego vehicle. The motion state of tracked objects is updated continuously to reflect changes in object motion.

In order to be able to identify and group those detections 9 that belong to a certain extended object 10, it is desired to determine which detections that have the same velocity vector, which now will be described for a first example.

The target object 6 and its corresponding detections 9 have a relative velocity v _t that has a velocity x-component v _x and a velocity y-component v _y . The x-component v _x runs parallel to the x-axis 21 and the y-component v _y runs parallel to a y-axis 22, which in turn runs perpendicular to the x-axis 21. For the velocity x-component v _x there is a corresponding acceleration x- component a _x , and for the velocity y-component v _y , there is a corresponding acceleration y-component a _y .

Assuming that all the detections 9 which are received from an extended object have the same direction of velocity, then the measured radial velocity V _dm of each one of these detections 9 is given by projecting the relative velocity vector v _t onto the radial direction. This is given by: as illustrated more in detail in the lower left part of Figure 1. Here, w denotes noise. The present disclosure is directed towards a principle, and although practically present, the noise w is omitted in the rest of the text.

For an extended object, multiple detections 9 originate from the same target object 6, and each detection has its individual measured radial velocity V _dm . For all N detections 9 that originate from the same target object 6, a set of equations will be :

It is desired to solve equation (2) for all M detections 9, 20 and to find the N detections 9 that originate from the same target object 6 and that correctly satisfy equation (2).

As there are two unknowns, at least two equations are needed so solve for the full velocity vector v _x, v _y . In case more detections are available, and hence more equations are available, the system becomes overdetermined and a least squares inversion can give the solution for the full velocity vector v _x, v _y .

The Random Sample Consensus (RANSAC) approach can be used to find the optimal solution, and it is robust to noise and outliers. This algorithm successively picks two random detections, and solves the above equation for the velocity components v _x, v _y using only these two detections.

Each one of the resulting random hypotheses is evaluated and the best fit which best describes the remaining detections is selected. To increase robustness, detections from multiple radar cycles from the past can be added to increase the number of detections, and hence the number of equations. This, however, is only applicable if the velocity vector does not change.

In this context, a radar cycle is one observation phase during which the vehicle radar system 3 is arranged to acquire data, process said data on several signal processing levels and to send out available results. This can be a fixed time interval, or it can be a dynamic time interval depending on environment conditions and processing load.

When the target object 6 is moving with an acceleration, the Doppler velocity of the detections 9 changes rapidly in each radar cycle, and therefore a standard RANSAC fit on data from multiple radar cycles does not give an accurate solution. With reference to Figure 2, the relative velocity v' _t v'' _t v''' _t v'''' _t for an extended target object 10' that is moving away from a radar transceiver 4' with a certain constant acceleration is shown for four consecutive radar cycles in the radar transceiver's field of view 11'.

The detections 9' , 9' ' , 9' ' ' , 9' ' ' ' for each radar cycle are shown with corresponding circles, crosses, squares and triangles .

Correspondingly, Figure 3 shows a graphic illustration of radial Doppler velocity as a function of azimuth angle for the accelerating target object 6'.

Therefore, in the cases where the target object 6 has a constant acceleration a _x, a _y , equation (1) becomes:

For all N detections 9 that originate from the same target object 6, a set of equations will be:

_N - As for equation (2), it is desired to solve equation (4) for all M detections 9, 20 and to find the N detections 9 that originate from the same target object 6 and that correctly satisfy equation (4) .

Thus, the four unknowns for the target velocity's and acceleration' s x-and y-components v _x , v _y , a _x , a _y need to be solved for. This is possible with at least four equations. Similar to the example described above for velocity only, the RANSAC method can be used to robustly estimate the four unknowns.

According to the present disclosure, for a certain radar cycle, the control unit 5 is adapted to consecutively select four detections 9 at random by means of which the four unknowns are solved. Out of the random hypotheses which are created this way, the control unit 5 is adapted to select the one which best describes the remaining detections.

This is accomplished by the control unit 5 that is adapted to:

- Determine a calculated radial velocity v _dc for each one of the selected detections by means of the calculated components v _x , v _y , a _x , a _y using equation (3) for v _dc instead

Of Vdm ·

- Determine an error between the calculated radial velocity v _dc and the measured radial velocity v _dm .

- Determine the number of inliers, where an inlier corresponds to an error that falls below an inlier threshold .

This is performed for a plurality of radar cycles such that a corresponding plurality of sets of components v _x , v _y , a _x , a _y is acquired. The control unit 5 is then adapted to choose the set of components v _x , v _y , a _x , a _y that results in the largest number of inliers . An important advantage for this approach is its robustness to noise. This is achieved because all of the four unknowns are optimized simultaneously by solving a single globally optimal equation .

Inverting the above equation is feasible in the general case, but it is computationally expensive. Therefore, according to some aspects of the present disclosure, to simplify the calculations only detections which are measured at the same cycle are used in a first processing stage for solving the velocity components v _x, v _y . Then, additional detections from another radar cycle, for example a previous radar cycle, are used to solve the acceleration components a _x, a _y .

According to some aspects, the complexity can be reduced by using only three unknowns according to a following second example.

In practical applications, the most relevant case of acceleration is in a longitudinal direction, in a driving direction. Assuming that acceleration always is in the driving direction and there is no lateral movement, the acceleration components a _x, a _y are

a _y = where \V _r \ is the magnitude of the relative velocity v _r .

Then rewriting equation (3) results in:

Simplifying equation (5) results in:

and

Substituting V dm w results in:

V _x cos (9 _m )+V _y sin(d _m )

(l + a— t) = w o:

V \ Vr\ ) and finally:

In order to make a fit, at least three detections are needed, either all from different radar cycles, or two detections from one radar cycle, i.e. the current radar cycle, and one detection from another radar cycle, i.e. the previous radar cycle. In that case, as in the previous first example, only detections which are measured at the same cycle are used in a first processing stage for solving the velocity components v _x , v _y . Then, additional detections from another radar cycle, for example a previous radar cycle, are used to solve the acceleration component a.

Optimization is achieved for three unknown variables, the target velocity's x-and y-components v _x , v _y and the simplified acceleration component a simultaneously.

An important advantage for the approach according to the present disclosure is the fact that detections from multiple cycles can be used at the same time, which makes the algorithm robust even in case of low number of detections and in the presence of noise and outliers.

With reference to Figure 4, the present disclosure also relates to a method for a vehicle radar system 3, where, for each one of a plurality of radar cycles, the method comprises:

12: Acquiring a measured azimuth angle 0 _m and a measured radial velocity v _dm for a plurality of detections 9 associated with an object 6.

13: Selecting one detection from said first plurality of detections 9, 20 for each one of two velocity components v _x , v _y comprised in a set of components v _x , v _y , a _x , a _y ; a to be determined, where the two velocity components v _x , v _y define a full velocity vector for a relative velocity v _r .

14: Selecting one detection from a second plurality of detections 9, 20 for each one of at least one corresponding acceleration component a _x , a _y ; a.

15: Calculating said components v _x , v _y , a _x , a _y ; a for the selected detections .

16: Determining a calculated radial velocity v _dc for each one of at least a part of the other detections in said first plurality of detections 9, 20 by means of said calculated components v _x , v _y , a _x , a _y ; a.

17: Determining an error between each calculated radial velocity v _dc and the measured radial velocity v _dm .

18: Determining the number of inliers, where an inlier corresponds to an error that falls below an inlier threshold. 19: Repeating the above for all radar cycles in said plurality of radar cycles.

When all radar cycles in said plurality of radar cycles have been processed according to the above, when there are no radar cycles left in the plurality of radar cycles, the method then comprises : 23: Choosing the set of components v _x , v _y , a _x , a _y; a that results in the largest number of inliers.

According to some aspects, the method further comprises:

24: Grouping those detections 9 that correspond to inliers resulting from the chosen set of components v _x , v _y , a _x , a _y; a, such that these detections 9 form an extended object 10.

The present disclosure is not limited to the examples above, but may vary freely within the scope of the appended claims. For example, the radar system may be implemented in any type of vehicles such as cars, trucks and buses as well as boats and aircraft .

The radar transceiver 4 is adapted for any suitable type of Doppler radar in a Doppler radar system. There may be any number of radar transceivers 4 in the vehicle radar system 3, and they may be arranged for transmission and signals in any suitable direction. The plurality of sensing sectors or sensing bins may thus be directed in other desirable directions, such as rearwards or at the sides of the vehicle 1.

Radar signal processing is performed in any kind of suitable control unit 5 comprised in the vehicle radar system 3, such as a DSP (Digital Signal Processor) or an RCU (Radar Control Unit) which according to some aspects comprises a DSP, or combination with separate or combined control unit arrangements.

Generally, the present disclosure relates to a vehicle radar system 3 where, for each one of a plurality of radar cycles, the radar system 3 is arranged to provide a measured azimuth angle 0 _m and a measured radial velocity v _dm for a first plurality of detections 9, 20. For each one of said plurality of radar cycles, the radar system 3 is arranged to: - select one detection from said first plurality of detections 9, 20 for each one of two velocity components v _x , v _y comprised in a set of components v _x , v _y , a _x , a _y; a to be determined, where the two velocity components v _x , v _y define a full velocity vector for a relative velocity v _r;

- select one detection from a second plurality of detections

9, 20 for each one of at least one corresponding acceleration component a _x , a _y; a;

- calculate said components v _x , v _y , a _x , a _y; a for the selected detections ;

- determine a calculated radial velocity v _dc for each one of at least a part of the other detections in said first plurality of detections 9, 20 by means of said calculated components v _x , v _y , a _x , a _y; a;

- determine an error between each calculated radial velocity v _dc and the measured radial velocity v _dm;

- determine the number of inliers, where an inlier corresponds to an error that falls below an inlier threshold;

where the radar system 3 furthermore is arranged to choose the set of components v _x , v _y , a _x , a _y; a that results in the largest number of inliers.

According to some aspects, the radar system 3 is adapted to group those detections 9 that correspond to inliers resulting from the chosen set of components v _x , v _y , a _x , a _y; a, such that these detections 9 form an extended object 10.

According to some aspects, the second plurality of detections 9, 20 is the same as the first plurality of detections.

According to some aspects, the second plurality of detections belongs to another radar cycle, where the radar system 3 is arranged to: - first calculate the velocity components v _x , v _y for the selected detections;

- then calculate said at least one acceleration component a _x , a _y ; a by means of the calculated velocity components v _x , v _y .

According to some aspects, the second plurality of detections belongs to a previous radar cycle.

According to some aspects, the radar system 3 is arranged to calculate said components v _x , v _y , a _x , a _y ; a by solving for these from the expression for each selected detection, where 0 _m constitutes a measured angle between a radar system reference line 21 and a direction towards the detection 9 in question.

According to some aspects, the radar system 3 is adapted to substitute a = a— and a _v = a—, where Ilf I is the magnitude of

\V _r \ ^y \V _r \’ ' ^r ' ^¾

the relative velocity v _r , such that said components v _x , v _y , a _x , a _y ; a are calculated by solving for these from the expression

Generally, the present disclosure relates to method for a vehicle radar system 3, where, for each one of a plurality of radar cycles, the method comprises:

12: acquiring a measured azimuth angle 0 _m and a measured radial velocity v _dm for a plurality of detections 9 associated with an object 6. For each one of said plurality of radar cycles, the method further comprises:

13: selecting one detection from said first plurality of detections 9, 20 for each one of two velocity components v _x , v _y comprised in a set of components v _x , v _y , a _x , a _y; a to be determined, where the two velocity components v _x , v _y define a full velocity vector for a relative velocity v _r;

14: selecting one detection from a second plurality of detections 9, 20 for each one of at least one corresponding acceleration component a _x , a _y; a;

15: calculating said components v _x , v _y , a _x , a _y; a for the selected detections;

16: determining a calculated radial velocity v _dc for each one of at least a part of the other detections in said first plurality of detections 9, 20 by means of said calculated components v _x , v _y , a _x , a _y; a;

17: determining an error between each calculated radial velocity v _dc and the measured radial velocity v _dm; and

18: determining the number of inliers, where an inlier corresponds to an error that falls below an inlier threshold; and

19: repeating the above for all radar cycles in said plurality of radar cycles;

where the method then comprises:

23: choosing the set of components v _x , v _y , a _x , a _y; a that results in the largest number of inliers.

According to some aspects, the method further comprises:

24: grouping those detections 9 that correspond to inliers resulting from the chosen set of components v _x , v _y , a _x , a _y; a, such that these detections 9 form an extended object 10.

According to some aspects, the second plurality of detections 9, 20 is the same as the first plurality of detections. According to some aspects, the second plurality of detections belongs to another radar cycle, where the radar system 3 is arranged to:

- first calculate the velocity components v _x , v _y for the selected detections;

- then calculate said at least one acceleration component a _x , a _y ; a by means of the calculated velocity components v _x , v _y .

According to some aspects, the second plurality of detections belongs to a previous radar cycle.

According to some aspects, the method comprises calculating said components v _x , v _y , a _x , a _y ; a by solving for these from the express on for each selected detection, where 0 _m constitutes a measured angle between a radar system reference line 21 and i direction towards the detection 9 in question.

According to some aspects, the radar system (3) is adapted to substitute a _x a— and a., = a—, where \V _r \ is the magnitude of the relative velocity v _r , such that said components v _x , v _y , a _x , a _y ; a are calculated by solving for these from the expression