Method for estimating a road boundary for a radar sensor

Abstract

The present invention relates to a method for estimating a road boundary for a radar sensor (9c, 9d) which is arranged on an ego vehicle (1) such that its field of view (10) is in the reverse direction and at least partially covers a side area of ​​the ego vehicle (1), wherein the radar sensor (9c, 9d) emits radar signals in a 90° angular range in vehicle coordinates, and in this angular range a detection (11) is searched for which no or a very low relative speed is then assigned to the road boundary.

Description

[0001] The present invention relates to methods for estimating a road boundary for a radar sensor, as well as a corresponding radar sensor and a vehicle having a radar sensor according to the invention. Technological background

[0002] Modern vehicles, such as cars and motorcycles, are increasingly equipped with driver assistance systems. These systems use sensors to perceive the surroundings, recognize traffic situations, and support the driver, for example, by applying the brakes or steering, or by issuing visual or audible warnings. Radar sensors, lidar sensors, cameras, and similar devices are regularly used as sensor systems for environmental perception. The sensor data collected by these sensors allows conclusions to be drawn about the environment. Environmental perception using radar sensors is based on the emission of focused electromagnetic waves and their reflection, for example, by other road users, obstacles on the road, or roadside structures. The radar sensor generates measurement points or detections, which are then used, for example, to determine the vehicle's position.An object can be assigned to it, which can then be tracked using a radar object tracker. Based on this, instructions can then be issued for driver warnings or information, or for (partially) autonomous steering, braking, and acceleration. Through the processing of sensor and environmental data, assistance functions can, for example, prevent accidents with other road users or facilitate complex driving maneuvers by supporting or even completely taking over the driving task or vehicle control (partially or fully automated). For example, the vehicle can perform autonomous emergency braking (Automatic Emergency Brake) using an emergency brake assist system (EBA) or speed and following distance control using an adaptive cruise control system (ACC).Furthermore, assistance functions are also considered, in which rear-facing sensors detect traffic behind the vehicle and then issue a warning to the driver and / or steer the vehicle accordingly, depending on the traffic volume or objects detected. Examples include blind spot detection (BSD), occupant safe exit (OSE), and lane change assist (LCA).

[0003] For systems of the type described above, radar sensors can also be used in combination with sensors of other technologies, such as cameras or lidar sensors. Radar sensors have the advantage, among others, of operating reliably even in adverse weather conditions and, in addition to measuring the distance to objects, can also directly measure their radial relative velocity via the Doppler effect. Transmission frequencies of 24 GHz, 77 GHz, and 79 GHz are typically used. Due to the increasing functional scope of such systems, the requirements are constantly rising, particularly regarding the maximum detection range. Besides the environmental sensing of vehicles for systems of the type described above, the interior monitoring of vehicles is also becoming increasingly important, for example, to detect which seats are occupied; frequencies in the 60 GHz range are used for this purpose.

[0004] In this context, the present invention addresses the problem of how to determine the lateral distance and curvature of an existing road boundary. A necessary prerequisite is a radar sensor directed to the rear (left or right) of the ego vehicle, as is typical for sensor systems designed to support functions such as BSD, OSE, and LCA.

[0005] A reliable road boundary detection algorithm is essential for ensuring safe and efficient operation, as it significantly improves object detection accuracy. For example, knowing the road boundary allows for mirror target classification. Furthermore, so-called "outliers" or detections located beyond the road boundary can be excluded from object tracking, or such objects can be classified as "behind the boundary" and thus excluded from the function. Known road boundary detection algorithms require information from the radar about the road boundary, such as a chain of radar detections, filled grid cells, or tracked stationary objects, to determine the road boundary's characteristics.However, as soon as this information is no longer available, these algorithms can no longer provide road boundary information. For example, radar information can be lost if the physical road boundary does not consist of a metal guardrail or poles, but rather, for instance, a low concrete wall or a change in the road boundary's structural features. In such cases, the radar no longer detects reflections, as, for example, a low concrete wall acts as a mirror surface and reflects the radar waves away instead of back to the radar sensor. Printed state of the art

[0006] From JP 2015 001 773 A, a lane estimation device is known which includes a sensor that detects a stationary object and a moving object located around a vehicle. The lane estimation device further comprises a first lane estimation unit that estimates a first lane based on the position of the stationary object detected by the sensor, a second lane estimation unit that estimates a second lane based on the trajectory of the moving object detected by the sensor, and a third lane estimation unit that estimates a third lane to be driven on by the vehicle based on the first and second lane estimates.

[0007] From DE 10 2022 207 736 B3, a method for determining the rearward road path of a vehicle is known, in which at least one rearward-facing sensor for environmental detection, in particular a radar sensor and / or lidar sensor, is provided, wherein the environmental detection takes place within successive measurement cycles of the environmental sensor and objects are detected based on the sensor's detections, a filter is provided for the detections by which static detections are selected, a grid with a vertical and a horizontal axis is created in which the static detections are entered, and a histogram is created during the sensor's environmental detection based on the filtered detections, wherein the number of filtered detections for a definable distance to the vehicle is determined along the horizontal axis.and a road boundary is determined based on the number of filtered detections and the distance to the vehicle on the horizontal axis.

[0008] Furthermore, DE 10 2015 108 759 A1 describes a method for classifying an object in the lateral vicinity of a motor vehicle, in which a signal is emitted into the lateral vicinity by a vehicle-side sensor device for monitoring the lateral vicinity and the signal reflected by the object is received, wherein a monitoring area of ​​the sensor device for monitoring the lateral vicinity is segmented into a first detection area and a second detection area, wherein a first value of a distance of the object to the motor vehicle is determined on the basis of the signal received from the first detection area and it is determined whether the object is a stationary object located to the side of the motor vehicle.Based on the signal received from the second detection area, a second value for the distance of the object to the motor vehicle is determined, and based on the first and second values ​​of the distance, it is determined whether the object identified as stationary is an elongated, stationary object extending in a longitudinal direction of the vehicle. The invention also relates to a driver assistance system and a motor vehicle.

[0009] Furthermore, DE 10 2015 105 206 A1 discloses a method for selectively reducing or filtering data, wherein the data are supplied by one or more sensors mounted on the vehicle before this data is used to detect, track and / or estimate a stationary object arranged along the side of a road, such as a guardrail or barrier. Object of the present invention

[0010] Starting from the prior art, the object of the present invention is now to provide a method for estimating a road boundary for a radar sensor and a corresponding radar sensor with which the road boundary can be determined in a simple and cost-effective manner and the accuracy of the determination can be improved. Solution to the task

[0011] The foregoing problem is solved by the entire teaching of claim 1 and the dependent claims. Advantageous embodiments of the invention are claimed in the dependent claims.

[0012] In the inventive method for estimating a road boundary for a radar sensor, the sensor is arranged on an ego-vehicle such that its field of view is directed backwards and at least partially covers a side area of ​​the ego-vehicle. Essentially, radar sensors of a vehicle are used that are arranged in the rear and / or side area of ​​the rear fenders. The radar sensor emits radar signals in a 90° angular range in vehicle coordinates, covering at least an area of ​​approximately 80°–100°, in particular 85°–95°. Within this angular range, a detection (i.e., a radar signal reflected back from an object and detected by the radar sensor) is sought that exhibits no or only a very low relative velocity, which can then be assigned to the road boundary.For the purposes of the invention, vehicle coordinates are understood to be a coordinate system in which the respective radar sensor serves as the origin and is located at 0° or 360° in the direction of travel, extending to the area where a road boundary is to be detected, e.g., 10 m or 15 m or the like. Naturally, a 90° angle range is also understood here to be an area that lies (depending on which side of the vehicle the respective radar sensor is located) within the 270° angle range (or 260° to 280° or 265° to 275°) or within the negative (-)90° angle range (or -80° to -100° or -85° to -95°). Detection at 90° (or also at -90° or 270°) is also available in the case of a concrete wall or continuous smooth metal surface, since no mirror-away effect occurs, but is reflected back onto the ego vehicle and thus onto the respective radar sensor.

[0013] The procedure for estimating a road boundary for a radar sensor comprises the following procedural steps: Step I: Transmitting and receiving radar signals in a 90° angular range, searching within that range for a detection that could correspond to the road boundary and has a very low relative speed; Step IIa: If no such detection is found that could correspond to a road boundary, a reliability value for an existing road boundary is lowered; Step IIb: If such a detection has been found, at least one motion detection is sought at the same lateral distance above and / or below the detected detection that shows a significant difference from the theoretical stationary velocity, which can be calculated; Step IIIa: If such motion detection was found in step IIb, the reliability value for an existing road boundary is reduced; Step IIIb: If no such motion detection was found in step IIb, a check is performed to see if a tracked moving object is present in the 90° angle range; Step IVa: If a valid moving object was found in step IIIb, the reliability value for an existing road boundary is reduced; Step IVb: If no valid moving object was found in step IIIb, the reliability value for an existing road boundary is increased; Step V: Assigning the detection to a road boundary, provided the reliability value reaches a definable threshold.

[0014] Furthermore, the procedure can also additionally include step VI, in which an ego-path curvature is determined, preferably by using points on the (past) ego-path, wherein a polynomial through the points is used with the help of a set of linear equations to determine the polynomial parameters.

[0015] Preferably, a detection of a road boundary is only assigned if it has a relative speed of < 0.2 m / s, i.e., has almost no relative speed or is within the error tolerance range.

[0016] Furthermore, the lateral velocity of the ego vehicle can also be taken into account, for example, by integrating the lateral acceleration, which is already present in most vehicles. This allows the expected Doppler value of a steady-state detection at 90° to be determined very accurately, even during lane changes, which will naturally differ from 0 m / s. Typical values ​​for the lateral velocity during lane changes on highways are, for example, between 1 and 3 m / s. These values ​​could then be considered when evaluating the 90° reflection.

[0017] It is advantageous to determine the curvature or curve of the road boundary. To determine the curvature of the road boundary, the ego-vehicle's path (or the driven trajectory) is used, with a series of points (e.g., 100 points) distributed along a straight line behind the ego-vehicle at fixed intervals, e.g., every 2 m. These points are then shifted based on the ego-vehicle's speed and yaw rate. The actual radar signal is not required for this. A polynomial function is then fitted through these points to obtain a mathematical representation of the curvature. By combining lateral distance measurement, 90° angle range detection, and the curvature of the driven ego-path, a high degree of accuracy in the road boundary representation can be achieved.The length of the road boundary is independent of any radar reflection behind the ego vehicle and is only limited by the necessary subsequent logic; a typical length could be, for example, 100 m.

[0018] Furthermore, the road boundary information (such as offset and curvature) can be sent to a sensor on the opposite side of the vehicle (e.g., from the rear right radar sensor to the rear left radar sensor and vice versa), so that each sensor can utilize a complete representation of the road boundaries. This allows the improvement measures described above to be applied to both the road boundary facing the radar sensor and the road boundary opposite it.

[0019] Advantageously, additional data from another environmental sensor, in particular a camera facing forward, or data from a navigation system, or data transmitted to the ego vehicle via V2X communication, can also be used to increase the confidence score, with the additional data including the curvature of the lane. Using this additional data allows for verification of the determined lane curvature, road edge characteristics, or other information, thus further increasing the accuracy of the determination.

[0020] Furthermore, the present invention also claims a radar sensor configured to perform a method according to the invention.

[0021] Furthermore, the present invention also claims a vehicle that has a radar sensor according to the invention. Description of the invention using exemplary embodiments

[0022] The invention will now be described in more detail using practical embodiments. The figures show: Fig. 1 a simplified representation of an embodiment of a vehicle according to the invention with radar sensors according to the invention; Fig. 2 a simplified representation of an embodiment of a process sequence according to the invention; Fig. 3A a simplified schematic representation of step I; Fig. 3B a simplified schematic representation of step IIb; Fig. 3C a simplified schematic representation of step IIIb; Fig. 4A a simplified schematic representation of step VI, wherein a series of points are distributed at fixed distances from each other behind the ego vehicle along the ego path; Fig. 4B a simplified schematic representation of step VI, where a polynomial is drawn through the points, as well as Fig. 4C is a simplified schematic representation of step VI, where the points or the line drawn through the points are moved to the road boundary based on the speed and yaw of the ego vehicle.

[0023] Reference number 1 in Fig. 1 designates a vehicle according to the invention, which comprises a control unit 2 (ECU, Electronic Control Unit or ADCU, Assisted and Automated Driving Control Unit) by which sensor control, sensor data fusion, environment and / or object recognition, trajectory planning and / or vehicle control can be carried out, in particular (partially) autonomously. For vehicle control, the control unit 2 can access various actuators (steering 3, motor 4, brake 5). Furthermore, the vehicle 1 has a front radar sensor 6, a lidar sensor 7, a camera 8 and radar sensors 9a-9d for environment detection. Advantageously, the sensor data can be used for environment and object recognition, so that various assistance functions, such as emergency brake assist (EBA), adaptive cruise control (ACC), lane keeping assist, etc., can be performed.A lane keeping assist system (LKA) or the like can be implemented, for which a road boundary estimation can be used, which is carried out using a method according to the invention. Furthermore, the execution of the assistance functions can also be carried out via the control unit 2. The radar sensors 9c and 9d in the rear area of ​​the vehicle 1 are radar sensors according to the invention, on the basis of which a road boundary estimation according to the invention can be carried out.

[0024] In the method according to the invention, road boundary estimation is performed solely by observing the detections at the level of the laterally directed radar sensor(s) at approximately 90° in vehicle coordinates (the so-called "Doppler zero" or "90° detection"; i.e., a zero measurement in the case of radar Doppler measurement), since this detection does not indicate relative speed. If a road boundary is present, this detection provides a direct measure of the lateral distance of the road boundary. Additional information from the radar that extends beyond the road boundary or concerns objects behind the road boundary is therefore no longer required to control the corresponding detection and functional algorithms. Fig. The vehicle coordinate system is shown in the upper right corner, with the origin always located at the sensor.

[0025] Since radar measurements of actual overtaking, overtaken, or parallel-driving objects or vehicles do not show relative speed at this 90° position, a number of plausibility checks can be applied to avoid the (false) detection of a road boundary in such cases. For example, after a clearly moving detection in the vicinity of the 90° area, a search can be conducted at the same lateral distance that indicates a moving object at this position. Furthermore, in the absence of a moving detection, a search can be made for a moving object that has entered this position from above or below the 90° area and is currently only being measured at 90°. If no indication of such a moving target is found in the 90° area, the previously detected 90° reading is considered a road boundary detection or roadside measurement.If this is confirmed in several measurement cycles, the confidence threshold can be reached, allowing a highly precise lateral distance to be specified. In subsequent cycles, the lateral distance is then filtered at 90° upon renewed detection, meaning that the lateral distance can change over time. This can occur, for example, due to the ego vehicle changing lanes.

[0026] The method according to the invention is primarily designed for laterally facing rearward radar sensors, as the curvature of the ego path can be utilized in this case. The lateral offset to the road edge alone is also useful, for example, to suppress ego-mirror objects behind the road edge at approximately 90°. This information could also be fused with external road information, such as that provided by the forward-facing camera, which provides information about the lane curvature. An additional approach could involve comparing the measured values ​​or offsets to the road edge or road boundary from sensors on the same side (e.g., a front and a rear radar sensor) to confirm or verify the results and thus increase confidence in the road boundary detection.Such fusion could be performed on each sensor individually by sending the lateral offset information to the other sensor via a bus interface, or the information could be processed on a central control unit or ADCU for surround systems (such as control unit 2).

[0027] One embodiment of the process according to the invention is described in Fig. 2 shown, whereby this process could be carried out using the following procedural steps: Step I: Transmitting and receiving radar signals within a 90° angular range, which, for example, encompasses an angular window of 85° to 95° in vehicle coordinates, with the respective radar sensor serving as the origin and extending to the area where a road boundary is to be located, e.g., 10 m or 15 m or the like. Within this angular range, a detection is then sought that could correspond to the road boundary and exhibits a very low relative speed, e.g., < 0.2 m / s or the like. Fig. Figure 3A schematically depicts the Ego vehicle 1, which includes the radar sensor 9c, which is located in the rear of the Ego vehicle 1 and has a field of view 10, which covers the 90° angle range, and in which a corresponding detection 11 with Doppler - 0 m / s or < 0.2 m / s was recorded. Step IIa: If no such detection is found that could correspond to a road boundary, the confidence or reliability value for an existing road boundary decreases, e.g. realized by a low-pass filter. Step IIb: However, if such a detection is found, at least one further detection is sought at the same lateral distance above and below the detected detection, e.g., + / - 5 m in the x-direction. This detection, in turn, should be moving slightly, as it should exhibit a significant difference from its theoretical stationary relative velocity, which is based on the kinematics of the ego vehicle. This detection can therefore also be referred to as motion detection. The significant difference in relative velocity could, for example, be > 2 m / s. Fig. Figure 3B illustrates this, where motion detection 12a was recorded at -5 m in the x-direction and motion detection 12b at 5 m in the x-direction. Such a motion detection, together with the 90° detection, represents a physical moving object that currently covers at least part of the 90° angle range and therefore does not allow a view of the road edge. Step IIIa: If such a motion detection was found in Step IIb, the confidence level or reliability value for an existing road edge decreases. Step IIIb: If no such motion detection was found in Step IIb, the system checks whether a tracked moving object is present within the 90° angle range. It is possible that a tracked object is only measured near 90° in certain measurement cycles. This object should have previously been detected by the sensors outside the 90° angle range, e.g., by entering the field of view from above or approaching the ego vehicle from behind, i.e., from the rear or back, as described in Fig. 3C based on the tracked or followed moving object 13, which is approaching from behind or from the rear. Step IVa: If a valid moving object 13 was found in step IIIb, the confidence or reliability value for an existing road boundary decreases. Step IVb: If no valid moving object was found in step IIIb, the confidence or reliability value for an existing road boundary increases. Step V: If the confidence or reliability value reaches a definable threshold or limit, a road boundary is assumed or recorded; if this threshold or limit is not reached, it is not. Step VI: Combining the lateral distance (lateral offset) to the road boundary with the curvature of the driven ego path 15. The ego path 15 is continuously determined independently of radar detections, so that an ego path curvature is always available and can be determined, for example, by forming a polynomial through these points using a set of linear equations to determine the polynomial parameters. Fig. 4A and Fig. Figure 4B illustrates how a series of points (marked as black stars) are recorded along the ego path (or the trajectory driven by the ego vehicle 1) (e.g., 100 points behind the ego vehicle 1 along a straight line 14 at fixed intervals of, for example, 2 m). These points are then moved from the ego path 15 to the road boundary 16 based on the speed and yaw of the ego vehicle 1, as shown in Fig. 4C is shown. The actual radar signal is not required for this. A polynomial function is then fitted through these points to obtain a mathematical representation of the curvature.

[0028] By combining lateral distance measurement (i.e., the y-distance or y-offset), 90° angle range detection, and the curvature of the driven ego path 15, a high degree of accuracy in the representation of the road boundary can be achieved. The length of the road boundary is independent of any radar reflection behind the ego vehicle 1 and is limited only by the necessary subsequent logic; a typical length could, for example, be 100 m.

[0029] For example, the calculation of the points and the mathematical representation of the curvature could be done by using the following parameters: Cycle time = Δt Yaw rate=γ˙ Ego speed = s Yaw angle = γ = Δt ⋅ γ˙

[0030] For all points Pi', an update of the translation can then be performed by Xi_new=s⋅Δt⋅Xi_old and an update of the rotation by (Xi_newYi_new)=(cos γ−sin γsin γ cos γ)⋅(Xi_oldYi_old) is carried out. Furthermore, the following can be used as a polynomial that passes through the points Pi': y=x⋅g+C0⋅x22⋅C1⋅x36 where the parameters we are looking for are g, C0, C1 and the linear regression is A · b = y, with A=[x1x122x136⋮⋮⋮xnxn22xn36]; b=[gC0C1]; y=[y1 ⋮yn]; b=(A'⋅A)−1⋅A'⋅y; where A' is the transpose of A. Reference symbol list 1 Ego vehicle 2 Control unit 3 Steering 4 Motor control 2 5 Brake 6 (Front) radar sensor 7 Lidar sensor 8 Camera 9a-9d radar sensor 10 field of vision 11 Detection 12a, 12b Motion detection 13 objects 14 Even 15 Ego Path 16 Street boundary

Claims

[1] Method for estimating a road boundary for a radar sensor (9c, 9d) arranged on an ego vehicle (1) such that its field of view (10) is in the rearward direction and covers at least part of a side area of ​​the ego vehicle (1), the method comprising the following steps: - Step I: Emitting and receiving radar signals by the radar sensor (9c, 9d) in a 90° angular range, searching within the angular range for a detection (11) which could correspond to the road boundary and has a very low relative speed; - Step IIa: If no such detection (11) is found that could correspond to a road boundary, a reliability value for an existing road boundary is lowered; - Step Ilb: If such a detection (11) has been found, at least one motion detection (12a, 12b) is searched for at the same lateral distance to the ego vehicle above and / or below the detection (11) that shows a significant difference from a theoretical stationary velocity; - Step IIIa: If such motion detection (12a, 12b) was found in step Ilb, the reliability value for an existing road boundary is reduced; - Step IIIb: If no such motion detection (12a, 12b) was found in step Ilb, check whether a tracked moving object (13) is present in the angular range; - Step IVa: If a valid moving object (13) was found in step IIIb, the reliability value for an existing road boundary is reduced; - Step IVb: If no valid moving object (13) was found in Step IIIb, the reliability value for an existing road boundary is increased; - Step V: Assign the detection (11) to a road boundary, provided the reliability value reaches a definable threshold. [2] Method according to claim 1, characterized by , that the method additionally includes step VI, wherein an ego-path curvature is determined, preferably by using detections on the (past) ego-path, wherein a polynomial is used by the detections with the help of a set of linear equations to determine the polynomial parameters. [3] Method according to any one of the preceding claims, characterized by , that a detection of a road boundary can only be assigned if it has a relative speed of < 0.2 m / s. [4] Method according to any one of the preceding claims, characterized by, that a lateral velocity of the ego vehicle (1) is taken into account in order to determine an expected Doppler value of a stationary detection (11) in the 90° angle range. [5] Method according to any one of the preceding claims, characterized by , that a curvature of the road boundary is determined by recording an ego-path (15) of the ego-vehicle (1), wherein a series of detections are distributed along a straight line at a fixed distance from each other along the ego-path (15), which are shifted to the position of the road boundary based on the speed and / or yaw of the ego-vehicle (1), in order to then fit the detections using a polynomial function to obtain the curvature of the road boundary. [6] Method according to claim 5, characterized by, that the curvature and / or offset is sent to another radar sensor (6, 9a, 9b, 9c, 9d) of the ego vehicle (1), in particular a radar sensor (9c, 9d) opposite the ego vehicle (1), or to a central control unit (2) in order to use it for object detection by the radar sensor (6, 9a, 9b, 9c, 9d) or for sensor fusion and / or data verification. [7] Method according to any one of the preceding claims, characterized by , that additional data from another environmental sensor, in particular a forward-facing camera (8), or data from a navigation system or data sent to the Ego vehicle via V2X communication, are used to increase the reliability value, the additional data including the curvature of the lane. [8] Radar sensor (9c, 9d), especially for object detection for a vehicle (1), characterized by, that the radar sensor (9c, 9d) is configured to perform a road boundary estimation according to a method according to one of the preceding claims. [9] Vehicle (1) comprising a radar sensor (9c, 9d) according to claim 8.

Images