Multisensor locating device

State of the art

The invention relates to a locating device with a plurality of sensors for measuring the distances of objects, a triangulation module for calculating the spatial coordinates of the objects on the basis of the measured distances and a selection module for distinguishing between real objects and false objects based on plausibility criteria.

Such locating devices are intended in particular for installation in motor vehicles and serve to locate other vehicles driving in front of their own vehicle and thus to create an Adaptive Cruise Control (ACC) data base in which the speed of the own vehicle is determined using a downstream controller is automatically adjusted so that an appropriate safety distance is maintained to the vehicle in front. In this case, a distinction must be made between vehicles in their own lane and irrelevant objects, for example vehicles on secondary lanes. Therefore, it is not sufficient to measure only the distances to the objects, but the location coordinates of the objects in a two-dimensional coordinate system must be determined.

In a multi-sensor locating device, as known for example from DE 199 49 409 A1, the location coordinates are determined by triangulation. The locating device comprises at least two sensors mounted on the front of the vehicle, in the transverse direction of the vehicle in a certain, arranged as a base width distance from one another, so that the location coordinates of a given object, for example, the x and y coordinates in a Cartesian coordinate system, whose x-axis coincides with the vehicle's longitudinal direction, can be calculated on the basis of the different distances measured by the sensors and on the basis of the known base width. The sensors are, for example 24-GHz pulsed radar sensors with which the object distances and, if desired, the relative velocities of the objects can be measured on the basis of the signal propagation times of the radar waves reflected on the object.

For a larger number n of objects, each sensor provides n distance values. In the case of two sensors, a total of 2n distance values ​​are obtained. It is not clear from the outset how the distance values ​​of one sensor are to be assigned to those of the other sensor. Overall, n result² possible combinations of distance values ​​representing a corresponding number of object candidates. In addition to the n real objects, one thus obtains a large number of false objects which are based only on a faulty assignment of the distance values. These false objects must then be eliminated by dynamic object observation. In doing so, one makes use of the circumstance that implausible changes in the coordinate values ​​often occur in the dummy objects, ie, the changes in the coordinate values ​​represent movements of the object candidates, for example extreme longitudinal or transverse movements, which are not to be expected in a real traffic situation.

In the known locating device, an individual sensor list is initially created for each sensor, in which the measured distance values ​​are arranged increasing. Starting with the smallest distance values, the associated location coordinates are then calculated for all conceivable combinations of distance values ​​from the various lists, and a multisensor list is created in which all object candidates are represented by their location coordinates. The radar measurements and the evaluation steps described above are repeated periodically, for example at intervals of the order of one millisecond. In a so-called tracking procedure, the movements of the object candidates are tracked over several measuring cycles. If it turns out that an object candidate performs an implausible movement, then this candidate is identified as a dummy object and ignored in the further evaluation.

Since, on the one hand, the number of apparent objects to be tracked in the tracking procedure can be very large and, on the other hand, the object location must be performed in real time, the known method requires a high resource outlay, in particular a high computing and storage capacity. It should be noted that even a single vehicle, in particular a truck with a highly rugged rear face, in practice usually forms a plurality of focal points of reflection, which are initially interpreted as separate objects in the evaluation. This leads to a considerable increase in the number of object candidates to be tracked, with the result that the resource requirement for the management and execution of the multisensor list increases further.

Advantages of the invention

The invention provides a locating device that enables reliable object location with reduced resource requirements.

This is achieved in a locating device of the type mentioned above in that a speed module is provided to calculate at least the first time derivatives of the location coordinates directly from the time derivatives of the measured distances, and that in the selection module at least one plausibility criterion is implemented, the calculated evaluates temporal derivatives of the location coordinates.

In the locating device according to the invention, the individual sensor lists created for each individual sensor contain, in addition to the distance data, the associated relative speeds. These can in principle be calculated by deriving the distance data after the time, but are preferably measured directly with the aid of the radar sensors. In the triangulation module, the location coordinates for each object candidate are calculated in a known manner on the basis of the distance data. In addition, at least the first temporal derivatives of the location coordinates are also calculated in the velocity module, that is, for example, the relative velocities of the object in the x and y directions. For these calculations, it is not necessary to resort to the tracking procedure, since the x and y components of the velocity can be calculated directly from the radial velocities inherited from the individual sensor lists. The longitudinal and transverse velocities (or even the angular velocities in the case of polar coordinates) of the object candidates are therefore already available at a much earlier time than in the conventional method. Thus, it is possible to recognize abnormal or implausible velocity components of an object candidate already in the triangulation and to eliminate the candidate immediately. In this way the scope of the multisensor list to be processed is considerably reduced.

Advantageous developments and refinements of the invention emerge from the subclaims.

An additional simplification of the evaluation and a further reduction of the resource requirements is made possible by the fact that the speeds of the real objects calculated by the speed module can be taken over directly by the tracking module in order to track the movements of these objects.

Since at least part of the plausibility checks can be performed independently of the tracking procedure, there is also the advantageous possibility of decoupling the cycle time of the triangulation algorithm from the cycle time in single-sensor signal processing.

If a single vehicle forms multiple reflection centroids, then the object candidates corresponding to these reflection centroids will not only have the same y-velocity component but also the same x-velocity component. This makes it possible to identify these object candidates as belonging together with high reliability. According to an advantageous development of the invention, therefore, the object candidates that are not rejected as implausible, are combined into clusters that have substantially the same (vectorial) speed. Since all members of a cluster represent the same physical object, only a single representative needs to be included in the multisensor list for each cluster, preferably the one with the smallest spacing, ie the smallest x coordinate. This considerably simplifies the processing of the multisensor list.

drawing

An embodiment of the invention is illustrated in the drawings and explained in more detail in the following description.

Show it:<ul id="ul0001" list-style="none"><li>Figure 1 is a sketch for explaining the operation of the locating device;</li><li>Figure 2 is a block diagram of the locating device; and</li><li>Figure 3 is a flow chart for explaining the operation of the locating device.</li></ul>

Description of the embodiment

In Figure 1, two objects A and B are shown in a two-dimensional Cartesian coordinate system, which are located with two distance-sensitive sensors S1 and S2. The sensors S1, S2 are, for example Radar sensors, which are installed at the front of a motor vehicle, not shown, whose longitudinal center axis coincides with the x-axis of the coordinate system. The sensors are arranged symmetrically to the x-axis in a specific, referred to as base width b distance from each other.

For the objects A and B, it can, for example to act two vehicles in front or, more precisely, idealized as punctiform assumed focal points of reflection of these vehicles. The object A has the distance r11 to the sensor S1 and the distance r21 to the sensor S2. The object B has the distance r12 to the sensor S1 and the distance r22 to the sensor S2. The locations of equal distance to the sensors S1 and S2 are shown in the drawing by circular arcs. Each of the sensors S1, S2 receives radar waves which are reflected by the two objects A and B. The amplitude maxima of the received signals have signal delays corresponding to the respective object distances. However, it is initially unknown which amplitude maximum originates from which object. The received signal pattern could therefore also be caused by two other objects C and D, which are also located at points of intersection of the circular arcs with the radii r11, r12, r21, r22. In the simplified example according to FIG. 1, in which only two real objects A, B are present, the measurement thus yields a total of four object candidates, namely the two real objects A, B corresponding to the distance pairs r11, r21 and r12, r22 and two dummy objects C. , D corresponding to the distance pairs r11, r22 and r12, r21. In the general case, with n real objects, one gets n² Object candidates corresponding to all possible pairings of the distances measured by the sensors S1, S2. Which of these object candidates are the real objects must be decided in a later evaluation step.

The x and y coordinates of the object candidates can be calculated by triangulation from the measured distance data. For example, for the apparent object C with the coordinates (x, y) from the theorem of Pythagoras, we obtain the relations:<maths id="math0001" num="(1)"><math display="block"><mrow><msup><mrow><mtext>r11</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> = x</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + (y + b / 2)</mtext></mrow><mrow><mtext>2</mtext></mrow></msup></mrow></math><img file="EP1522873A1_D0001.tif" /></maths><maths id="math0002" num="(2)"><math display="block"><mrow><msup><mrow><mtext>r22</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> = x</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + (y - b / 2)</mtext></mrow><mrow><mtext>2</mtext></mrow></msup></mrow></math><img file="EP1522873A1_D0002.tif" /></maths>

Multiplying the equations (1) and (2) yields<maths id="math0003" num="(3)"><math display="block"><mrow><msup><mrow><mtext>r11</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> = x</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + y</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + b * y + b</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>/ 4</mtext></mrow></math><img file="EP1522873A1_D0003.tif" /></maths><maths id="math0004" num="(4)"><math display="block"><mrow><msup><mrow><mtext>r22</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> = x</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + y</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> - b * y + b</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>/ 4</mtext></mrow></math><img file="EP1522873A1_D0004.tif" /></maths>

By subtracting equation (4) from equation (3), one obtains<maths id="math0005" num="(5)"><math display="block"><mrow><msup><mrow><mtext>y = (r11</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> - r22</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>)/2 B</mtext></mrow></math><img file="EP1522873A1_D0005.tif" /></maths>

Addition of equations (3) and (4) yields<maths id="math0006" num=""><math display="block"><mrow><msup><mrow><mtext>2 * x</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> = r11</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + r22</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> - 2 * y</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> - b</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>/ 2</mtext></mrow></math><img file="EP1522873A1_D0006.tif" /></maths> and resolved to x:<maths id="math0007" num="(6)"><math display="block"><mrow><msup><mrow><mtext>x = [(r11</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + r22</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext>) / 2 - y</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> - b</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext>/ 4]</mtext></mrow><mrow><mtext>1.2</mtext></mrow></msup></mrow></math><img file="EP1522873A1_D0007.tif" /></maths>

Here, the value obtained from equation (5) can be used for y, so that both coordinates x and y can be calculated from the measured object distances r11 and r22 and the known base width b. In an analogous way, the coordinates of the other object candidates can also be calculated.

By differentiating equation (5) by time, one obtains:<maths id="math0008" num="(7)"><math display="block"><mrow><mtext>vy: = dy / dt = (r11 * V11 -r22 * V22) / b</mtext></mrow></math><img file="EP1522873A1_D0008.tif" /></maths>

Where V11 = d (r11) / dt and V22 = d (r22) / dt are the radial velocities that can be measured directly from the Doppler effect using sensors S1 and S2.

If equation (6) is differentiated by time, one obtains<maths id="math0009" num="(8)"><math display="block"><mrow><mtext>vx: = dx / dt = </mtext><mfrac><mrow><mtext>r11 * V11 + r22 * V22 - 2 * y * vy</mtext></mrow><mrow><msup><mrow><mtext>2 * [(r11</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> + r22</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext>) / 2 - y</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext> - b</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><msup><mrow><mtext>/ 4]</mtext></mrow><mrow><mtext>1.2</mtext></mrow></msup></mrow></mfrac></mrow></math><img file="EP1522873A1_D0009.tif" /></maths>

Here y and vy can again be used from equations (5) and (7).

As soon as the coordinates of an object candidate have been determined by triangulation, the time derivatives of these coordinates, that is to say the x and y components vx, vy of the velocity of the object candidate, can thus also be calculated directly.

If desired, higher derivatives of the coordinates can also be calculated analogously, although the derivatives of the radial velocities V11, V22 required for this purpose are to be determined in each case from measurements carried out shortly after one another.

Knowing the velocities and possibly the higher derivatives of the coordinates makes it possible, in most practical cases, to immediately recognize at least some of the object candidates as fake objects, as will now be explained with reference to FIG.

In Figure 1, it has been assumed as an example that the object A rests relative to the sensors, while the object B approaches the sensors, that is, has negative radial velocities with respect to the two sensors. In the drawing, this is symbolized by the fact that the object positions B ', C', D 'are drawn in dashed lines for the object B and for the dummy objects C and D, which appear after expiration of a short time interval Δt, ie at a point in time r12 has decreased by V12 * Δt and r22 by V22 * Δt. It can be seen that the x-coordinate of the object B has decreased by a corresponding, relatively small amount. For the dummy object C, on the other hand, the position C 'is displaced relatively laterally to the original position, ie, the y-coordinate has increased by a relatively large amount, which would correspond to a large velocity component of the object C in the y-direction. Accordingly, a relatively large velocity component in the negative y-direction is also observed in the case of the object D.

In a real traffic situation, however, the relative speed of an object is subject to certain physical limitations. In particular, the relative speeds in the direction of the y-axis are relatively small in magnitude, because both the vehicle in front and the own vehicle can only perform slow transverse movements. Transverse movements, which are simulated by a yaw movement of their own, the sensors S1, S2 bearing the vehicle are low in magnitude. For each speed component, therefore, one can specify a specific interval within which "plausible" values ​​must lie for vy, wherein the interval limits can depend on other parameters characterizing the traffic situation, in particular on the absolute speed and the yaw rate of the own vehicle. Also for the relative velocities in the direction of the x-axis there are physical limits. If one of the velocity components vx or vy calculated for an object candidate in the manner described above is outside the plausibility limits, it can be decided immediately that the subject object candidate is a dummy object.

If a single object candidate has been identified as a dummy object, as in the example shown, the object candidate C, this at the same time further limits the selection for the real objects among the remaining object candidates. There must be at least one real object which has the same distance (r11) from the sensor S1 as the detected dummy object C. In the example shown, this may be only the object A. Likewise, there must be at least one real object B that has the same distance (r22) from the sensor S2 as the dummy object C. Thus, in the simplified example situation according to FIG. 1, it is already clear that A and B are the real objects, and D can consequently be recognized as a dummy object, even if the calculated velocity components for D should still be within the plausibility limits. Because of this context, there is a high probability that fake objects are really recognized as such.

Of course, there are also physical plausibility limits for the components of the relative accelerations in the y direction and in particular also in the x direction. By including higher derivatives of the location coordinates, the probability of detecting false objects can therefore be further increased.

FIG. 2 shows the structure of a locating device, which operates according to the above-described functional principle, on the basis of a block diagram. The sensor S1 supplies the measured distances r11, r12 and the associated radial velocities V11, V12 to an evaluation device, which uses this data to apply an individual sensor list 10 sorted according to growing distances. If the sensor S1 detects all real objects, the number of entries in the individual sensor list 10 thus corresponds to the number n of real objects (in the example under consideration, n = 2). Each object is in the individual sensor list by its distance r11 or r12 represents. In addition, for each object, the associated radial velocity V11 or V12 indicated. Accordingly, a single sensor list 12 is created for the measured data of the sensor S2, which contains the measured distances r21, r22 and the associated radial speeds V21, V22.

The distance data rli (i = 1... N) and r2j (j = 1... N) from both individual sensor lists are fed to a triangulation module 14, which derives from this for all n² Object candidates computes the coordinates xij, yij. A speed module 16 calculates the velocity components vxij and vyij for each object candidate, accessing the coordinate values ​​calculated by the triangulation module 14 as well as the radial velocities Vli, V2j stored in the individual sensor lists 10, 12. The results, together with the coordinates xij, yij, are forwarded to a selection module 18 which creates a multisensor list of all object candidates arranged according to increasing x coordinates and checks in turn for each object candidate whether the velocity components vxij and vyij comply with the respectively associated plausibility limits. The candidates for which at least one plausibility limit has been violated (in the example shown the object candidate C) are eliminated here, as are the candidates, which then necessarily have to be false objects, as in the example shown the object candidate D. Only the remaining candidates, which are then likely to be real objects, are passed to the downstream processing stages.

In the example shown, the selection module 18 is followed by a cluster module 20, which examines the velocity components vxij, vyij of the remaining candidates to see if there are clusters of candidates that have both a matching x-component and a matching y-component of the velocity respectively. The candidates of such a cluster almost certainly represent different centers of reflection of one and the same physical object, for example different parts of the rear face of one and the same vehicle. They can therefore be treated as a single real object. Accordingly, the cluster module 20 selects from each cluster that object which has the smallest x-coordinate, that is closest to the vehicle equipped with the sensors S1 and S2.

In the very simple example of Figure 1, no such clusters are present, so that the cluster module 20 remains ineffective. In practice, however, such clusters occur very frequently, so that a considerable reduction in the number of remaining objects is achieved by the cluster module 20. The multisensor list with the remaining objects, which are now to be regarded as real objects, is transferred to a tracking module 22. In the example shown, the multisensor list contains the real objects A and B with the coordinates (x11, y11) and (x22, y22). Preferably, it also contains the associated velocities (vx11, vy11) and (vx22, vy22).

The indices 11 and 22 here are the indices ij from the individual sensor lists 10 and 12.

When new data from the sensors S1 and S2 arrive in successive measuring cycles, the tracking module stores the data of the real objects for several consecutive measuring cycles, and by comparing the new coordinates with the stored previous coordinates and speeds, the movements of the individual objects are tracked. Other plausibility criteria can also be checked in order to subsequently eliminate any apparent objects contained in the list. The data of the objects tracked in this way are then available for evaluation, for example in an adaptive speed controller.

In order to save computing capacity, the described embodiment can be modified so that the processing operations in the selection module 18 and in the cluster module 20 and possibly also the operations in the triangulation module 14 and in the speed module 16 are not performed in every single measurement cycle, but only at longer intervals need.

For this purpose, it is expedient to additionally store the associated entries from the individual sensor lists 10, 12 in the object list in the tracking module 22 for each object. Since this storage is done only for the real objects, the additional storage space requirement is low. The advantage is that now the tracking can also be performed directly on the individual sensor data.

A possible procedure is shown in FIG. 3 with reference to a flowchart. It should be assumed that a number of real objects has already been identified in previous measuring and tracking cycles. In step 100, new individual sensor lists 10, 12 are created for the current measurement cycle. In step 101, the new entries in the individual sensor lists with the corresponding previous entries for the already known real objects are identified by a tracking procedure. The calculations in the triangulation module 14 and in the velocity module 16 then only need to be performed for the few index pairs i, j that correspond to known real objects, and the results can be transferred directly to the tracking module 22.

In step 102, it is checked whether or not one of a plurality of control conditions, which will be explained later, is satisfied. If no control condition is met, a return is made to step 100 and steps 100 to 102 are run again for the next measurement cycle.

One of the control conditions checked in step 102 is that a new object appears, ie that at least one of the sensors S1, S2 measures a distance and a radial speed that can not be assigned to a known object. Another control condition may be that a previously tracked object is lost. Finally, a control condition may be that the loop of steps 100 through 102 has been traversed a certain number of times without any of the other control conditions having occurred.

If at least one of these control conditions is met, a branch is made to step 103, and in the selection module 18, a new multisensor list is created using the triangulation module 14 and the speed module 16, which now contains all index pairs i, j and thus represents all object candidates. In step 104, in the selection module 18, the plausibility checks for all members of the multi-sensor list and, if necessary, Clustering performed in the cluster module 20. In step 105, a new list of real objects is then passed to the tracking module 22, and the objects are tracked there. Thereafter, a return is made to step 100.

This process ensures that the computationally expensive steps 103 and 104 do not need to be executed every measuring cycle, but only when there is a real need or when a longer time interval comprising several measuring cycles has elapsed.