DE102018215448B3 - Method of estimating the geometry of a path of movement - Google Patents

Abstract

Method for estimating the geometry of a movement path of a means of locating sensor data by means of a sensor and estimating the geometry of the path of movement based on the sensor data, determining static targets along the path of movement based on the sensor data, location points for the respective position of the targets and be determined by the sensor, a polygon is determined by the location points and the polygon is used to estimate the geometry of the movement path.

Description

The present invention relates to a method for estimating the geometry of a movement path of a means of locomotion, in particular of a vehicle, as well as a system for carrying out the method.

Technological background

The recognition and estimation of road edges or road borders is an elementary function in modern driver assistance systems and z. B. in an automatic distance control (ADR) or Adaptive Cruise Control (ACC) or an emergency brake assist (EBA) used to estimate the trajectory or the path of movement of the respective means of transport or vehicle.

The road boundary is usually expressed in the form of a clothoid. As a clothoid in modern road construction z. B. the transition arc between a straight line and a curvature. The respective course of the clothoid can z. B. estimated using static targets or objects that are located in the area of the road boundary and were detected by radar measurements. For example, such an estimate of the road boundary is based on a Kalman filter, the z. For example, to reduce errors in real metrics and to provide estimates of non-measurable system sizes.

The filter records the coefficients of the clothoids for the left and right side of the road for each cycle. Subsequently, search ranges are formed by the prediction and introduction of the covariance of the Kalman filter to define the region that is interesting for the upcoming measurement update of a cycle. Here, the search range is understood as the region in which the subsequent cycle searches for detections in which a detection occurs during the subsequent cycle or radar scan, so that it is accepted in order to calculate the clothoid parameters. Only static road targets that fall between these search areas are considered road boundary targets for the upcoming measurement update of the Kalman filtering process. This method works very well in scenarios with relatively smooth boundary edges and transitions, such as those found in B. on highways are available. However, this method is very inaccurate for highly dynamic scenarios, such. As a motorway entrance or highway exit, a roadway widening or road constriction and urban scenarios. Especially because it can lead to an overlap of the search areas, for example, when the vehicle enters a curve. Here, the uncertainty of the search areas can become so large that they overlap each other. With overlapping search areas, it is usually difficult to distinguish whether the destinations or objects must be assigned to the left or right lane. Therefore, the roadside is estimated to be the overlapped region. This leads to a very short length of the road boundary estimation, which is usually not desired.

Furthermore, there may be greater latency in adapting to changing environments, taking several cycles for the Kalman filter to calculate a roadside clothoid. This can lead to a wrong alignment of the search areas and thus a misalignment of the entire clothoid. As a result, road bends can be misjudged or non-existent road bends can be determined as road bends. In addition, too narrow search areas can be formed, so that real targets on the street side are wrongly classified as outliers. This leads to an inaccurate estimation of the clothoid or to an inaccurate road width estimation. In road construction work using lane separators, the density of target reflections from lane separators may be much lower than the density of a guardrail. The Kalman filter can give the wrong impression that such targets are to be classified as outliers and therefore only need to be weighted slightly to determine the road boundary. It may even happen that the algorithm identifies the guardrail as a road boundary rather than a narrowed lane width due to roadworks. The method of road boundary estimation using the Kalman filter thus has a certain potential for error, especially in dynamic scenarios.

Printed prior art

From the US Pat. No. 6,751,547 B2 For example, a method for accurately estimating the geometry of the forward path of a vehicle based on a two-clothoid road model is known. Road data collected by a camera or a radar system is collected here. Based on the road data, a measurement transfer function of the two-clothoid road model is then calculated. The clothoid coefficients for the near range and the far range are simultaneously estimated and the forward travel of the vehicle is estimated from the data provided by the two-clothoid model.

The US Pat. No. 6,718,259 B1 discloses a method for estimating the geometry of a forward path of a vehicle using an adaptive Kalman filter bank and a two-clothoid road model. Here are several Kalman filters where weighting factors are provided for each filter indicating the likelihood of whether the upcoming road geometry coincides with the road model assumed in the filter. After each weighted value has been assigned to the filters, the weighted value road models are fused by means of a fusion element so that a weighted road model can be output.

From the DE 10 2015 013 084 A1 and from the DE 10 2013 201 796 A1 Each method according to the preamble of claim 1 are known.

Object of the present invention

The object of the present invention is to provide, starting from the prior art, a method in which the estimation of the geometry of the movement path is improved in a simple and cost-effective manner.

Solution of the task

The above object is achieved by the entire teaching of claim 1 and the independent claim. Advantageous embodiments of the invention are claimed in the subclaims.

In the method according to the invention for estimating the geometry of the movement path of a means of locomotion, in particular of a vehicle, sensor data are collected by means of a sensor, wherein the estimation of the geometry of the movement path is based on the sensor data. This takes place in particular by determining static targets or objects along the path of movement on the basis of the sensor data, and locating points for the respective position of the targets and the sensor. Furthermore, a polygon is determined based on the location points, z. For example, by using a definable number of location points as vertices of the polygon. In particular, the polygon may include location points of each street side. Subsequently, the polygon is used to estimate the geometry of the path of movement, i. H. the inner road boundary points can be z. B. be determined for both road boundaries. This provision may, for. In the form of an algorithm, i. E. H. By means of the invention, the inner road boundary of the road can be determined or calculated on the basis of the sensor measurements. This results in the advantage that the algorithm has low computational complexity and is easy to implement or retrofit. Further, the method has no dependency on page classification or calculation of previous radar cycles. Thus, the inventive method can also be used for the classification or estimation of the trajectory or movement path geometry in dynamic scenarios, such. As in road boundary development or narrowing, roadworks, missing crash barriers or the like.

According to a preferred embodiment, a radar sensor is provided as the sensor. However, other known from the prior art sensors may be provided, such. B. camera or Lidarsensoren.

According to the invention, a plurality of polygons, in particular continuously, are determined so that an entire polygon mesh is determined. This mesh can then be used to estimate the geometry of the path of motion. The estimation of the geometry can be made even safer, so that the method can be used reliably even in highly dynamic driving scenarios. It has proved to be particularly advantageous if the polygon is a triangle, since this can be easily calculated with little computational effort. In the same advantageous manner, an entire polygon mesh can be created on the basis of triangular formations. In a simple way this can be done by means of a Delaunay triangulation or Delaunay triangulation.

Preferably, location points are set only for static targets, since usually only static targets mark a road boundary. Movable or mobile targets or objects, on the other hand, are not suitable for determining the road boundary. As a result, no location points are set for these destinations, e.g. B. to spare the computing capacity. The error rate is thereby further reduced.

Expediently, a segment is arranged in each case between two loci, which is thus part of the polygon or of the (Delaunay) triangle.

It is particularly advantageous if the location points and / or segments are classified as belonging to the road or to the road surface or to the road boundary. The classification can then be used in a practical way to estimate the geometry of the path of movement.

Preferably, a test line is provided which runs through the middle or at least the middle region of a segment associated with the road surface and whose inclination is determined on the basis of the inclination of a segment associated with the road boundary. The test line can be (new) calculated and inserted for the respective segment.

According to a particular embodiment of the invention, at least two test lines can be provided, wherein each of the test lines is assigned to one of the road boundaries, ie, is arranged in the vicinity or adjacent to this road side. For example, the test line can intersect the segment from the location point in the first third, preferably in the first quarter, and more preferably in the first fifth of the segment.

Furthermore, the respective innermost road boundary points of the two road boundaries can be determined by the two test lines. The road boundary points are then compared with each other, which can be concluded by comparing the road boundary points on divergent road courses by a set value for the deviation is exceeded or fallen below. By deviating the road boundary points can thus be concluded in a simple manner on divergent road courses. As a result, intersections or road bifurcation, as z. B. occur at motorway exits or construction sites, are detected. Furthermore, the estimation of the geometry of the movement path can be further improved. In addition, this embodiment implements an additional function for detecting tram paths.

Further, the present invention includes a system for estimating the geometry of the travel path of a vehicle. For this purpose, the system has a sensor for generating sensor data, wherein the estimate of the geometry of the movement path takes place on the basis of the sensor data and static targets along the movement path can be determined. The system is further adapted to perform the estimation of the geometry of the path of travel using the method of the invention, locating points for the respective position of the targets and the sensor, determining a polygon from the locus points, and using the polygon to estimate the geometry of the path of travel becomes.

In summary, the uncertainty associated with overlap of search ranges can be reduced by the present invention in addition to the allocation. Furthermore, filtering of useful detections for road boundary estimation can be enhanced by narrow search ranges. In addition, latency can be improved by adapting to changing environments and by determining road boundaries when working in traffic constrictions.

list of figures

• 1 eine vereinfachte schematische Darstellung eines Verfahrens zur Schätzung der Geometrie eines Vorwärtsweges eines Fortbewegungsmittels gemäß dem Stand der Technik,
• 2 eine vereinfachte schematische Darstellung eines Ausführungsbeispiels des erfindungsgemäßen Verfahrens;
• 3 eine vereinfachte schematische Darstellung eines weiteren Ausführungsbeispiels des erfindungsgemäßen Verfahrens;
• 4 eine vereinfachte schematische Darstellung einer Ausgestaltung eines Ablaufplanes des erfindungsgemäßen Verfahrens;
• 5 eine vereinfachte schematische Darstellung eines Verfahrensschritts des Ausführungsbeispiels aus 4;
• 6 eine vereinfachte schematische Darstellung eines weiteren Verfahrensschritts des Ausführungsbeispiels aus 4;
• 7 eine vereinfachte schematische Darstellung eines weiteren Verfahrensschritts des Ausführungsbeispiels aus 4, sowie
• 8 eine vereinfachte schematische Darstellung eines weiteren Verfahrensschritts des Ausführungsbeispiels aus 4.

In the following the invention will be explained in more detail with reference to practical embodiments. Show it:

• 1 a simplified schematic representation of a method for estimating the geometry of a forward path of a means of transportation according to the prior art,
• 2 a simplified schematic representation of an embodiment of the method according to the invention;
• 3 a simplified schematic representation of another embodiment of the method according to the invention;
• 4 a simplified schematic representation of an embodiment of a flowchart of the method according to the invention;
• 5 a simplified schematic representation of a method step of the embodiment of 4 ;
• 6 a simplified schematic representation of a further method step of the embodiment of 4 ;
• 7 a simplified schematic representation of a further method step of the embodiment of 4 , such as
• 8th a simplified schematic representation of a further method step of the embodiment of 4 ,

In 1 For example, an estimate of the geometry of a vehicle's travel path using an adaptive Kalman filter bank and a clothoid road model according to the prior art is shown. The respective course of the clothoid K is based on goals or objects O estimated. The estimation of the road boundary is based on a Kalman filter. The Kalman filter records the coefficients of the clothoids for each cycle K for the left side of the street kl and right side Kr. However, this method of road boundary estimation using a Kalman filter has, as described at the outset, a large potential for error, especially in dynamic scenarios.

In the method according to the invention, the geometry of the movement path is estimated by determining or calculating the inner road boundary on the basis of sensor measurements. In the following, exemplary embodiments are described in which a radar sensor is used as the sensor. Explicitly, however, include other known from the prior art sensors, such. B. Lidar- or camera sensors. As in 2 shown, the radar sensor delivers in each scan cycle a list of objects or goals with the corresponding relative position of the respective targets and their speed relative to the ego vehicle, which comprises the radar sensor and the location point 0 (x = 0, y = 0) corresponds. Based on the relative speed of the ego vehicle, the objects are then classified as static or mobile targets. For the estimation of the road boundary, however, only the static targets are used in the present case, to each of which a location point can be assigned.

The estimation is performed by calculating a Delaunay triangulation (Delaunay triangulation or Delaunay triangulation) based on the position of the static targets and their location points and that of the ego vehicle, respectively. Then the first left and right points in the triangulation diagram are searched or searched.

As in 2 shown, the search is carried out by first the first left point (location point 2 ) and the first right point (location point 54 ) and thus a segment ( 2 . 54 ) form. Thereafter, the first left and the first right point are directly connected to the location point 0 and the position of the ego vehicle, the first left point (location point 2 ) a negative y-coordinate and the first right-hand point (location point 54 ) a positive one y Coordinate has. Consequently, the first triangle is through the location points 0 . 2 and 54 limited and set for further search in the Delaunay triangulation.

Appropriately, then continue with the next triangle, which is recognized at the first left and right points. In the embodiment according to 2 So it is the triangle between the locations 2 . 4 and 54 , Then the location points 2 . 4 and 54 of the new triangle or segments ( 2 . 4 ) and ( 4 . 54 ), whether they belong to the roadside or the road surface or to the road surface. For this purpose, a test line T inserted, which deals with the segments of the new triangle ( 2 . 4 . 54 ) can cross. If the test line T crosses with segment, z. B. segment ( 4 . 54 ), then the segment is classified as belonging to the road surface. In the same way, a segment that is the test line T not crossed classified as belonging to the road boundary, such. B. the other segment ( 2 . 4 ).

The algorithm repeats for each newly determined segment until one is identified as belonging to the road surface. Consequently, as long as the next adjacent triangle on the segment ( 4 . 54 ) until it is decided whether the segments ( 4 . 7 ) or ( 7 . 54 ) belong to the road border or the road surface.

The test line T goes through the middle of the segment of the road surface, which was determined in the previous algorithm step. The slope of the test line T is based on the inclination of the road boundary segments, ie the test line T has the same slope as the road boundary segments. Further, the slope can be smoothed by a low-pass filter (based on exponential smoothing) whose input value is the slope of the detected road boundary. The algorithm is terminated as soon as no further triangle or segment has been found, which has a common corner point on the road surface.

Conveniently, the algorithm can also, according to 3 , for detecting intersections or road bifurcations, such. B. at motorway exits or construction sites, are used. For example, by having two test lines T be used. Each of the test lines can do this T each run near one of the road boundaries, in which case the same inner road boundary points are obtained. If different or at least more divergent road boundary points are obtained, this is used as an indicator that the road boundaries or road courses deviate from each other, so that one can draw conclusions about a road fork or intersection.

The following is based on an embodiment according to the flowchart in 4 as well as the diagrams in the 5-8 an algorithm for an actual radar measurement on the highway is shown.

First, according to 5 , as an initialization step, the slope of the test line T adjusted to the vertical and calculated the Delaunay graph. In the following first step, according to 5 , the objects become 2 and 54 classified as left and right place point, with the segment between the place points 2 and 54 is defined as opening gate to the street. As a result, the first Delaunay triangle (location points 0 . 2 and 54 ) set. Following this, it is examined whether the triangle with the location points 2 . 4 and 54 next adjacent triangle to the segment ( 2 . 54 ) is to create. This segment can then be stored or saved in an external list.

In the second step, according to 6 , the test line T is set, which determines the slope of the segment ( 2 , 4) or the same angle and through the middle of the segment ( 4 . 54 ) runs. If now the segment (eg the segment between the points 2 and 4 ) not with the test line T crosses, this segment ( 2 . 4 ) classified as belonging to the road boundary and the location point 4 becomes point of view due to the immediate vicinity 2 classified as left-hand location. When the segment (for example, the segment between the points 4 and 54 ) but with the test line T it is classified as belonging to the road surface. Subsequently, it is examined whether the triangle with the location points 4 . 7 and 54 next adjacent triangle to the segment ( 4 . 54 ) is to create. Following this, the segment ( 4 . 54 ) are removed from the diagram.

In the third step, the points of the second step are repeated, with the result that the segment ( 4 . 7 ) is classified as the left side of the road and the segment ( 7 . 54 ) is then removed from the graph. Then the triangle ( 7 . 10 . 54 ). The process is then terminated, such as in 7 shown when a segment ( 53 . 69 ) no longer has any other supporting triangles. This stops the search and the corresponding segment ( 53 . 69 ) is removed from the chart. As a result, the algorithm provides the inner road boundary points and the boundaries of the road for the left and right road sides, as in FIG 8th with the Delaunay graphic separated into two sub-graphs. Further, each of the sub-graphs includes all the location points of a roadside.

Claims (11)

Method for estimating the geometry of a path of travel of a means of locating sensor data by means of a sensor and estimating the geometry of the path of movement based on the sensor data, determining targets along the path of movement based on the sensor data, location points for the respective position of the targets and Sensors are determined, a polygon is determined based on the location points and the polygon is used to estimate the geometry of the path of movement, characterized in that a plurality of polygons are determined, so that a polygon mesh is determined, which is used to estimate the geometry of the path of movement. Method according to Claim 1 , characterized in that a camera, radar or Lidarsensor is provided as the sensor. Method according to Claim 1 or 2 , characterized in that the polygons are determined continuously. Method according to at least one of the preceding claims, characterized in that the polygon is a triangle and / or the polygon mesh is determined on the basis of a Delaunay triangulation. Method according to at least one of the preceding claims, characterized in that location points are set only for static targets. Method according to at least one of the preceding claims, characterized in that a respective segment is arranged between two location points. Method according to at least one of the preceding claims, characterized in that the location points and / or segments are classified as belonging to the road surface or road boundary and the classification is used to estimate the geometry of the movement path. Method according to at least one of the preceding claims, characterized in that a test line (L) is provided which runs through the middle of the segment of the road surface and whose inclination is determined by the inclination of the segment of the road boundary. Method according to at least one of the preceding claims, characterized in that at least two test lines (T1, T2) are provided, wherein each of the test lines (T1, T2) is assigned to one of the road boundaries. Method according to Claim 9 , characterized in that based on the assignment of the test lines (T1, T2) for each road boundary road boundary points are obtained, the road boundary points are compared and closed by comparing the road boundary points on divergent road courses by a definable value for the deviation on - or falls below. System for estimating the geometry of a movement path of a means of locomotion, with a sensor for generating sensor data, wherein Based on the sensor data of the sensor, the estimation of the geometry of the movement path takes place by the system is adapted to perform the estimation of the geometry of the path of travel by a method according to at least one of the preceding claims.

Images