Method and assistance system for estimating path of travel of motor vehicle based on information filter

Abstract

The invention relates to a method and an assistance system (7) for repeatedly estimating a current travel path (14) of a motor vehicle (5), wherein the method can be carried out periodically and repeatedly during the operation of the motor vehicle (5) and to a correspondingly configured motor vehicle (5). In the method, a plurality of measurement data (11) having a relevant uncertainty are detected, said measurement data characterizing traffic conditions of the motor vehicle (5). An update information matrix and an update information vector in the form of an information filter are thus estimated in each case in an update step. An updated estimate of the travel path (14) is then determined from the update information matrix and the update information vector.

Description

Technical Field

[0001] The invention relates to a method and an assistance system for estimating a travel path of a motor vehicle during operation of the motor vehicle. The invention also relates to a correspondingly configured motor vehicle. Background Art

[0002] Nowadays, there is a need for increasingly automated vehicle guidance and further improvement of assistance functions and systems for motor vehicles. As a basis for this, an estimate, i.e. a prediction, of the travel path of each motor vehicle can be used that is as accurate as possible. In the current sense, such a travel path, which can also be referred to as a travel path, specifies an area ahead of the motor vehicle in the direction of travel, through which the motor vehicle is at least expected to pass or along which it moves. Such a travel path can be delimited, for example, by the edge of the road and / or by defined edges.

[0003] For example, DE 10 2013 2005 950 B4 describes a method for detecting the edge of a road on a driving road to detect a free path. Here, candidate lines are detected based on an input image, and an initial vanishing point in the image is identified as a function of the candidate line. For each candidate line, an area of ​​interest in the image is identified, wherein each area of ​​interest includes the candidate line and a surrounding area. Features related to the candidate line and extracted from the area of ​​interest are input into a classification device to identify whether the candidate line is a potential road edge. The classification device assigns a confidence value to the candidate line, which identifies the probability of whether the candidate line is a road edge. Based on the confidence value being greater than a predetermined value, the potential candidate line is identified as a reliable road edge.

[0004] As described, for example, in DE 10 2006 040 334 A1, the current speed of the respective vehicle and its yaw rate can be taken into account for estimating the driving path. However, the corresponding known methods are not always sufficient for more complex driver assistance systems. Therefore, a method for lane recognition is proposed there. In this method, lane markings in the area in front of the vehicle are detected using a sensor system of the vehicle. Reference points with coordinates of a first coordinate system are assigned to the lane markings, which are then converted into a second coordinate system. The course of the lane markings and / or the lane is reconstructed from the position of the reference points in the second coordinate system.

[0005] DE 10 2016 003 935 A1 discloses a method for obtaining edge building information describing edge buildings along a road section traveled by a motor vehicle. Trajectory data describing the past trajectory of the motor vehicle and / or the pre-calculated trajectory of the motor vehicle are used as input data. In this method, the acquisition area covering the described trajectory is divided into interval areas perpendicular to the longitudinal direction of the vehicle. For each interval area, the position of obstacles located in the interval area is sorted according to the position of the trajectory belonging to the left edge building or the right edge building. The edge building information is then obtained from the obstacle position closest to the trajectory for each interval area. Therefore, it should be possible to obtain edge building information from sensor data provided by a universal data interface independently of the model and the specific sensor type used.

[0006] A problem with existing approaches for estimating the driving path is that they are usually relatively computationally intensive, whereas in motor vehicles only limited hardware and computing resources are usually available, such as controllers designed as embedded systems, etc. Therefore, there is a need not only for improved environment recognition and driving path estimation, but also for reducing the computing power required for this purpose. Summary of the invention

[0007] The object of the present invention is to enable a particularly efficient estimation of the travel path of a motor vehicle during operation of the motor vehicle.

[0008] According to the invention, this object is achieved by the subject matter of the independent patent claims. Possible embodiments of the invention are disclosed in the dependent claims, the description and the drawings.

[0009] The method according to the present invention can be applied to repeatedly estimate the current driving path of a motor vehicle, which is correspondingly located in front of the motor vehicle along the driving direction. The method can be applied to the operation of each motor vehicle. With the help of the method according to the present invention, the corresponding current driving path can be estimated continuously or regularly repeatedly during driving or the estimated driving path can be continuously or regularly repeatedly updated. In the method steps of the method according to the present invention, measurement data are detected together with the relevant uncertainty, and the measurement data are given or characterize the corresponding traffic conditions of the motor vehicle. The measurement data can be given or arranged, for example, as a vector. Uncertainty can be given or arranged, for example, as an uncertainty matrix, which can only contain the uncertainty of the measurement data. Uncertainty can, for example, be given to the measurement error, confidence interval or corresponding limited accuracy of the measurement data. Uncertainty can be provided and / or calculated by the sensor together with each measurement data or measurement value, for example, by means of a pre-given heuristic method, etc.

[0010] The measurement data and their associated uncertainties can hereby together form or expand a measurement space or be given in a measurement space. The measurement space is an abstract mathematical space. If a plurality of measurement data, for example of different types and / or at different locations, etc., are used for estimating or updating the driving path, the measurement space can be correspondingly high-dimensional. In practice, the uncertainty matrix can be, for example, of the order of 100 rows and columns or more.

[0011] The measurement data can describe or reflect the corresponding surroundings of the motor vehicle, especially in front of the motor vehicle in the direction of travel. Similarly, the measurement data can be or include, for example, the operation or status data of the motor vehicle, the motion data of other traffic participants, etc., which can, for example, give the trajectory of a third-party vehicle driving in the corresponding surroundings at a previous time point. In particular, a single measurement data can be of multiple different types, from multiple different sensors, from different measurement points, i.e., measurement locations in the surroundings, etc. This can achieve a particularly robust and accurate estimation or update of the driving path. However, in the current solutions, it may also be accompanied by a corresponding large amount of data processing or calculation effort. However, this problem is circumvented or avoided by the method according to the present invention.

[0012] For this purpose, in a further method step of the method according to the invention, an update information matrix and an update information vector are estimated in an update step in the form of an information filter or mechanism, respectively, i.e., as soon as new measurement data are available or have been detected. In other words, an information filter is used here in order to take into account the uncertainty contained in each new or updated measurement data in the estimation or updating of the driving path. An information filter is a Cayman filter, also known in technical terms as an inverse covariance filter.

[0013] In a further method step, an updated estimate of the driving path is determined from the update information matrix and the update information vector, respectively, i.e., for example, in each run of the method or after each update step. In other words, a state or a state vector describing the driving path according to the corresponding current estimate or prediction can be determined accordingly.

[0014] The application of the information filter form proposed here enables an updated estimate of the driving path without the need to invert the uncertainty matrix, as is usually necessary in previous solutions. Instead, individual measurements can be considered or processed independently of each other in the updating step. For this purpose, correspondingly smaller matrices used for individual measurements, such as individual sensors or individual measuring locations or measuring positions, etc., can be inverted. The information matrix to be inverted, if necessary, can be smaller, in particular very small, i.e., has a smaller dimension than the uncertainty matrix for each of the entirety of the detected or newly detected measurement data. Therefore, the present invention enables an estimate of the driving path or an update of the driving path that results in an equivalent solution to the previous solution, in which the update of the estimate of the driving path by taking into account each new measurement data requires the inversion of a relatively high-dimensional uncertainty matrix, which represents all correspondingly new measurement data or the uncertainty of the measurement. Here, compared to this method so far, the method according to the present invention can be performed with very little computational effort. This allows for an equivalent or even improved travel path estimation during operation of the motor vehicle compared to previous methods, for example by using more measurement data and / or more frequent or high-frequency updates, in particular also when using limited hardware or computing resources, such as controllers or embedded systems commonly used today.

[0015] The driving path or its estimate can be described here by parameters or values ​​which are relatively few in comparison with the set or number of measurement data and their uncertainty, so that even for the calculation steps or method steps, they can be carried out with particularly little computational effort in the corresponding state space, which is expanded by the parameters for describing the driving path or its state. Since a correspondingly accurate and reliable driving path estimate can form the basis for many different assistance systems and / or at least partially automated driving of a motor vehicle, effort and energy can be saved without drawbacks and / or improved safety and / or improved comfort can be achieved particularly easily. Thus, for example, the updating of the driving path estimate can also be carried out using the computing hardware that is common in motor vehicles today, for example in an average time of less than 1 ms. Thus, a correspondingly rapid response to new measurement data or a changing traffic situation of the motor vehicle can be achieved.

[0016] In the sense of the present invention, the traffic situation of a motor vehicle can be given by the environmental situation or the surroundings of the motor vehicle. Likewise, the traffic situation can include the corresponding current state of the motor vehicle, such as its position, speed, steering angle, etc. The surroundings can be given or described here, for example, by the presence and possible position of the corresponding local traffic infrastructure, surrounding features, objects, obstacles, other traffic participants, etc. and their sensor recognizability. Corresponding features or objects in the surroundings can be or include, for example, lanes or lane markings, curbs, road or carriageway curves, elements of road facilities, road or lane edges, guardrails, vegetation strips, etc.

[0017] The problem discussed and the conception proposed here to solve the problem can be summarized as follows, which will also be explained in more detail below: Matrix inversion for large matrices is in principle very complicated. This involves updating the measurement uncertainty matrix to be inverted in the previous method, which is much larger than the state uncertainty matrix. The estimation problem can be expressed based on information matrices and information vectors. Therefore, an information filter can be applied at least as part of the solution. Here, all individual measurements can be considered independently. Therefore, instead of inverting the relatively large entire measurement uncertainty matrix for updating, only the inversion of multiple relatively very small measurement uncertainty matrices can be performed. Finally, the inversion of the equally relatively small information matrix is ​​used to transfer the updated estimate to the state space, which is significantly less troublesome than inverting the entire measurement uncertainty matrix.

[0018] In a possible embodiment of the invention, the measurement data at least describe the road course ahead of the motor vehicle along its driving direction at different distances from the motor vehicle, i.e., the current position corresponding thereto. Thus, the measurement data can characterize, image, or scan the corresponding surroundings or roads, for example, in the range of centimeters or decimeters or with a resolution of the order of 1 m, for example, up to a distance of several tens of meters. As a result, the surroundings or the road course ahead can be particularly accurately identified, which ultimately also enables a particularly accurate, robust, and reliable estimation of the driving path. Here, by particularly efficient data processing, i.e., a particularly efficient driving path estimation, such detailed measurement data can be used as the basis for estimating the driving path without significant disadvantages compared to previous solutions, which, given the computing power or computing hardware, could only process fewer or less detailed measurement data in real time.

[0019] In another possible configuration of the present invention, the measurement data are arranged in a measurement data vector, and the associated uncertainty is arranged in an uncertainty matrix. In the updating step, the individual measurement data and their associated uncertainty are considered independently of each other, without inverting the entire uncertainty matrix. For example, the individual submatrices corresponding to the individual measurement data of the entire uncertainty matrix can be inverted individually, that is, individually, instead of inverting the entire uncertainty matrix. These submatrices can have dimensions of 3x3 or 4x4, for example, and the dimension of the entire uncertainty matrix can be in the order of magnitude of 100x100 or greater. In particular, the dimension of the submatrix can be an integer divisor of the dimension of the entire uncertainty matrix. Therefore, the advantages of the information filter form can be utilized here, that is, by adding the corresponding information matrix and the information vector, the measurement or individual measurement data can be filtered in each run or time step of the method. In the current sense, a single measurement data can be a single measurement point or a data point from one or more data sources. Here, a single sensor can provide a single measurement data or multiple single measurement data at each re-measurement. The embodiment of the invention proposed here provides an easily implementable approach for estimating the driving path without loss of accuracy and in doing so avoids the computational or data processing effort for inverting the entire uncertainty matrix.

[0020] In another possible configuration of the present invention, the state or state vector corresponding to the update information matrix and the update information vector is first determined for the driving path. This state or state vector characterizes or describes the driving path by pre-given, especially geometric parameters. Here, the number of these pre-given parameters is less than, especially much less than, the dimension of the measurement data or measurement space, in which the measurement data and its uncertainty are given. When the measurement data is arranged in the form of a vector or matrix, for example, there may be dozens or hundreds of rows or even columns. On the contrary, the number of pre-given parameters for characterizing or describing or giving the driving path or the state or state vector for the driving path may be, for example, up to 20 or up to 10, for example, 4. The number of parameters may be, for example, up to one-fifth or up to one-tenth of the number of rows of the measurement data vector or the relevant uncertainty matrix. Therefore, in this way, the driving path, i.e., its corresponding current estimate, can be given or processed with particularly little data processing effort.

[0021] In a possible extension of the present invention, the pre-set parameters for characterizing the driving path include its curvature, especially at multiple pre-set distances from the motor vehicle or its current position, that is, in the positions spaced apart from each other along the driving direction of the motor vehicle or along the longitudinal direction of the driving path or along the longitudinal extension scale space. Thus, the driving path or its trend can be described or given in a particularly simple, efficient and data-saving manner. This characterization of the driving path can be sufficient for practical purposes, because it must always be given or satisfied, that is, a certain characteristic of the driving path or the corresponding boundary conditions for the trend of the driving path can be fixedly pre-set accordingly. This can include, for example, that the driving path always extends on the local road surface, can have a pre-set width, can have a maximum curvature that is pre-set or determined, for example, by the technical characteristics of the motor vehicle, must extend continuously, that is, without interruption, and is not allowed to have any jumps, steps or lateral offsets, etc. Similarly, the pre-set parameters for characterizing the driving path can include, for example, its radius, length, width, orientation, etc. The values ​​of these parameters can be given or input in the state vector mentioned at other positions, for example. This can achieve particularly simple and efficient data processing.

[0022] In a possible extension of the invention, the last determined update information matrix is ​​inverted for determining the updated state of the driving path. The inverted information matrix is ​​then multiplied by the last determined update information vector. This constitutes a particularly simple and low-effort possibility for describing the driving path or its state or estimate by real, clearly understandable, for example, geometrical dimensions, i.e., specifying the driving path or its state or estimate in a correspondingly clear and realistic state space. This enables the corresponding current driving path estimate to be used particularly simply and effectively by other assistance systems and assistance functions. For example, the latter can be performed independently of the form used for the driving path estimate.

[0023] In a possible extension of the present invention, the trend of the corresponding updated estimate of the driving path in the corresponding surrounding environment is determined by the updated state of the driving path respectively by means of a predetermined model. Here, the predetermined model models or takes into account at least one predetermined condition of the actual driving path allowed. The model can, for example, model the characteristics or boundary conditions mentioned at other locations of the driving path. By the combination of a state that can be given, for example, by a basic parameter value or a point parameter value and at least one predetermined condition, the complete driving path can be determined by the model, for example, as a two-dimensional or three-dimensional longitudinally extending geometric object, which can be positioned relative to each current position of the motor vehicle and / or in a predetermined, for example, fixed coordinate system in the world. Therefore, the model can be given or modeled how to derive the real trend of the driving path from each measurement data or the state or state vector of the driving path estimated thereby. The model can, for example, be stored and implemented in an estimation device that is set to estimate the state of the driving path. The model can be based on the corresponding current state, for example, from a circular arc segment or a curve, etc. to construct a real driving path trend. Here, for example, it can be predefined as a condition that the transition between successive arc segments or curves should be jump-free, that is, these arc segments or curves must be connected or transitioned into each other without jumps, offsets or distances. This enables particularly effective modeling of the actual driving path trend.

[0024] In another possible configuration of the present invention, environmental data are correspondingly detected as part of the measurement data. These environmental data describe the surroundings of the motor vehicle, in particular the surroundings ahead in the direction of travel of the motor vehicle. As part of the measurement data, steering angle data of the motor vehicle are also correspondingly detected. These steering angle data can, for example, give or include the current steering angle or the current change of the steering angle, such as its change direction and / or change speed. It is then also determined whether a lane change maneuver or a turning maneuver of the motor vehicle is expected to occur in the area from the corresponding current position of the motor vehicle to the predetermined distance from the motor vehicle in the direction of travel of the motor vehicle. This can be achieved, for example, based on navigation data, the driving history of the motor vehicle and / or the corresponding driver, a corresponding predetermined motion model, etc.

[0025] If such a maneuver is expected, an updated estimate of the travel path up to a predetermined distance is determined based on the steering angle data and, starting from the predetermined distance based on the surrounding data, an updated estimate is determined without taking into account the steering angle data. The travel path up to the predetermined distance can be estimated, for example, based solely on the steering angle data or based on the steering angle data and the surrounding data. This allows a particularly accurate and reliable estimate of the travel path up to the predetermined distance, since the steering angle in the corresponding short range primarily determines and limits the direction of movement of the motor vehicle.

[0026] At the same time, by ignoring the steering angle data from a predetermined distance, a further driving path course connected thereto can also be estimated particularly reliably, since unforeseen steering angle changes or steering maneuvers may occur between the current position of the motor vehicle and the predetermined distance, so that the current steering angle data at the current position of the motor vehicle do not constitute a reliable basis for estimating the driving path course from the predetermined distance.

[0027] The embodiment of the invention proposed here therefore makes it possible to achieve a particularly accurate and reliable estimation of the travel path both in a short range up to a predetermined distance and in a long range exceeding the predetermined distance with particularly little computational or data processing effort.

[0028] The predetermined distance may be, for example, related to the respective motor vehicle or its technical characteristics and / or the corresponding surroundings and / or the current speed of the motor vehicle. Therefore, the predetermined distance may be dynamically predetermined, i.e., automatically and dynamically adjusted accordingly during the operation of the motor vehicle. The predetermined distance may also be predetermined as a fixed value. The predetermined distance may be, for example, between 10 m and 30 m, for example 15 m. Other values ​​may also be possible depending on the application.

[0029] For example, a larger distance can be predefined or set at a smaller maximum possible steering angle of the motor vehicle and / or a smaller maximum possible steering angle change speed. Likewise, a larger distance can be predefined or set at a higher speed of the motor vehicle. For example, a longer distance can also be predefined or set on a highway or a highway-like extension than in a city. This makes it possible to achieve a driving path estimate that is particularly adapted to the situation, that is, optimized in each case in different situations or surroundings.

[0030] The invention also relates to an assistance system for a motor vehicle. The assistance system according to the invention has an interface implemented in hardware and / or software for detecting measurement data characterizing the traffic situation of the motor vehicle. In addition, the assistance system according to the invention has a processing device, i.e., for example, a microchip, a microprocessor or a microcontroller, etc., and a computer-readable data memory coupled thereto. The assistance system according to the invention is hereby configured to perform the method according to the invention, in particular automatically. Therefore, the assistance system according to the invention can be configured to estimate or predict the driving path of the motor vehicle or update the estimated or predicted driving path of the motor vehicle regularly or continuously or almost continuously, for example, in relation to the measurement or recording frequency of the measurement data. For this purpose, a corresponding operation or computer program can be stored in the data memory, which encodes or executes the method steps, measures or processes or corresponding control instructions mentioned in conjunction with the method according to the invention. Then, the operation or computer program can be executed by means of a process device in order to execute the corresponding method or to promote the execution of the method.

[0031] The invention also relates to a motor vehicle having an environment sensor device for recording environment data and an assistance system according to the invention. The environment data may be data describing or characterizing the corresponding surroundings of the motor vehicle, in particular in front of it in the direction of travel of the motor vehicle. The environment sensor device may be part of the assistance system or be connected to the assistance system, for example via an onboard network of the motor vehicle. The motor vehicle according to the invention may in particular be or correspond to the motor vehicle mentioned in conjunction with the method according to the invention and / or in conjunction with the assistance system according to the invention.

[0032] Further features of the invention are apparent from the claims, the drawings and the description of the drawings. The features and feature combinations mentioned above in the description of the drawings and the features and feature combinations mentioned below in the description of the drawings and / or shown individually in the drawings can be used not only in the respectively given combination but also in other combinations or individually without departing from the scope of the invention.

[0033] The accompanying drawings show an exemplary schematic overview of a traffic situation in a single figure for intuitively explaining an information filter-based travel path estimation for a vehicle. In the figure, for the sake of clarity, only a partially representative selection of elements that occur multiple times are explicitly marked.

[0034] The identification of objects relevant to the longitudinal and lateral control of the vehicle may require the identification of a possible driving path of the vehicle, and then all identified objects can be associated with this driving path. In order to ensure a high availability of the driving path or the corresponding current estimate of the driving path in various traffic situations and road conditions, a predetermined fusion algorithm can be applied to correspondingly fuse different measurements or input data from different data sources, such as intelligent, different sensors, etc. Therefore, different aspects of each environment, that is, each surrounding environment, such as data about lane markings, edge buildings, grass, curbs, etc., can be detected and summarized into a coherent overall image of the driving path. The objects that are identified in the corresponding surrounding environment, especially movable, can then be assigned to the driving path. The movable objects in the vehicle surroundings, especially assigned to the driving path, can be used or taken into account, for example, for longitudinal control functions such as ACC, intersection assistance, adaptive energy recovery, etc., and for lateral control functions such as steering and lane guidance assistance, etc.

[0035] Approaches to date have generally been based on the assumption that the vehicle's driving path corresponds to the course of the current road or lane on which the vehicle is currently traveling, as long as the driving path is not known anyway by the planning component of the system for automatic vehicle guidance, and must be determined with the aid of sensors that detect the corresponding environment. This course can be determined, for example, in a recursive estimation process that compares the current assumption or estimate of the course of the road or the corresponding lane with lane markings detected in sensor data, for example in a camera image.

[0036] Therefore, the problem to be solved is to determine or estimate the corresponding driving path ahead or its trend. So far, the corresponding estimation process usually works in the corresponding state space of the solution, where the uncertain state is updated with each uncertain measurement, that is, each recording or detection of uncertain measurement data, or for example, each time the lane marking or road or lane trend is identified with uncertainty. This state can be given, for example, by parameters or parameter values ​​of the parameterized representation of the geometry of the road or the lane being traveled. With such an update, the estimated state can be better consistent with each measurement or matched to the measurement, while reducing the uncertainty of the state estimate. In the current scheme, the uncertainty of the measurement or measurement data is usually represented by a high-dimensional matrix of the covariance of the measured parameters.

[0037] The disadvantages of the previous solutions become more apparent if more and more different measurements or measurement data are to be taken into account in the estimation process, for example measurement data from different sensors or finely resolved measurement data of the road course at different distances from the vehicle. In the previous solutions, updating the estimate by taking into account each new measurement or measurement data may require the inversion of a corresponding high-dimensional matrix, which represents the uncertainty of the measurement. Such a matrix may have a size of 500x500, for example. Inverting such a matrix may require significant computational effort, but at the same time, embedded systems, such as those usually used as the basis for today's driver assistance systems, can only provide very limited computing time or computing power. DETAILED DESCRIPTION

[0038] In this context, Figure 1 An exemplary schematic and partial overview of a traffic situation is shown. Specifically, a section of a road 1 is shown, which is bounded by a road edge 2. Here, for example, there are two lanes on the road 1, which are separated from each other by a lane marking 3. In addition, a guardrail 4 extends here along one of the road edges 2.

[0039] A motor vehicle 5 is currently moving on a road 1. The motor vehicle 5 may be equipped with an assistance function that uses the corresponding current driving path of the motor vehicle 5 as a data basis. In order to meet this requirement, the motor vehicle 5 is configured to continuously or regularly estimate the driving path or to update the estimate of the driving path accordingly. For this purpose, the motor vehicle 5 has an environment sensor device 6 and an assistance system 7. The assistance system 7 can detect the environment data recorded by means of the environment sensor device 6 and other data, such as steering angle data of the motor vehicle 5, via an interface 8. In order to process these data to estimate the driving path, the assistance system 7 can have a processor 9 and a computer-readable data memory 10, as shown schematically here by way of example.

[0040] The detected data can be of different types, originate from different sources and characterize or describe the surroundings of the motor vehicle 5 at different locations. For example, a plurality of corresponding data or measurement points 11 are represented here, of which only a selection is explicitly indicated for the sake of clarity. The data or measurement points 11 can, for example, originate from a third-party vehicle 12 also traveling on the road 1, describing its trajectory 13, including one or more points derived from the current steering angle or current position of the steering wheel 16 of the motor vehicle 5, i.e., on the driving route 17 of the motor vehicle 5 projected into the future based on the steering angle, representing the measurement or detection of the lane edge 2, lane markings 3, guardrails 4, etc.

[0041] On this basis, the assistance system 7 continuously and repeatedly determines the corresponding current driving path estimate 14 during the operation of the motor vehicle 5, which is also schematically shown here. Compared with the number or the dimension of the data detected as the basis for the driving path estimate 14, the corresponding driving path estimate 14 can be described by relatively few parameters. In the concrete case, the state to be estimated for the driving path can consist of four parameters, for example, which respectively specify the curvature of the driving path at different distances, which are represented here as reference points 15, for example. The assistance system 7 can then describe, i.e. determine the complete course of the driving path, i.e. the complete driving path estimate 14 for the motor vehicle 5, with the help of a predetermined model, for example stored in the data memory 10, which uses the corresponding estimated state as input. The corresponding current driving path estimate 14 can then be stored, for example, in the data memory 10 and / or provided or transmitted via the interface 8.

[0042] Starting from the observation that the dimension of the corresponding state space expanded by the parameters for describing the state or trend of the driving path is smaller, in particular much smaller than the dimension of the measurement space given or representing the detected data or their uncertainty, a simplified, less computationally complex and therefore more efficient determination or update of the driving path estimate 14 can be performed compared to previous solutions. For this purpose, instead of directly estimating the state of the driving path and the associated uncertainty, the information vector and the associated information matrix are estimated by the auxiliary system 7 in the form of an information filter based on the corresponding detected data. Therefore, the use of an information filter is provided here, wherein the corresponding update of the driving path estimate 14 can be realized, that is, can be performed without the complex inversion of the entire uncertainty matrix of the detected data. Instead, the individual detected data or measurements or measurement data of, for example, the individual measuring points 11 are considered independently of each other in the corresponding update step. In this case, the inversion of the entire uncertainty matrix can be effectively replaced by the inversion of the corresponding smaller matrix, which is simpler to calculate. Then, the estimate of the corresponding state can obtain a lower-dimensional information matrix from the information vector after the inversion.

[0043] For the first starting point, i.e. the first driving path estimate 14 when or after the motor vehicle 5 is put into operation, a simplified driving path estimate 14 can be determined, for example, based on the current steering angle of the motor vehicle 5, and then this driving path estimate can be continuously updated as described above, in particular taking into account surrounding data or based on this estimate of the course of the road 1 in the driving direction ahead of the motor vehicle 5. Thus, starting from the driving path initially based on the current steering angle of the motor vehicle 5, the estimate of the parameters of a given driving path state, such as the above-mentioned curvature at a plurality of reference points 15, can be continuously adapted to the correspondingly latest detected data or the correspondingly latest measurement or estimate of the actual road course of the road 1.

[0044] For this purpose, a plurality of different types of measurements from different sensors can be provided at any time, all of which can be converted into a uniform form for data fusion. Such measurements can be, for example, or include or are given by building structures or devices along the road or roadway edge 2, such as guardrails 4, which reflect back to the radar sensor of the environment sensor device 6, markings, such as lane markings 3 and / or vegetation or grass, etc., recognized in the camera image of the camera of the environment sensor device 6, the trajectory 13 of the third-party vehicle 12 in front, the collective movement of other traffic participants in the corresponding surroundings, the steering angle or steering angle change of the motor vehicle 5, the position, orientation, direction of movement and speed of the motor vehicle, etc.

[0045] Thus, for example, for each updating step, several hundred uncertain measurements, for example in the form of certain angles of the road course of the road 1, can be provided for fusion at a certain number of different distances. By transferring the problem into information space, i.e., expressing it via information vectors and information matrices and applying information filters, the corresponding updates can be performed with particularly little computational effort and thus particularly quickly or with correspondingly few computing resources, in particular in real time. This makes it possible, for example, to implement such an updating step on conventional embedded systems or controllers, in particular even under unfavorable conditions, with an average of up to 1 ms.

[0046] In this case, an estimate of the driving path, a description of the driving path state, is proposed by combining fewer parameters, in particular geometric parameters, for example up to 20 parameters, and information filters compared to the correspondingly detected data. As a result, a large number of different input data can be processed particularly easily, with little effort and quickly into a corresponding current driving path estimate 14.

[0047] Overall, the described example shows how operation-optimized data fusion can be implemented for determining a driving path, in particular also on embedded systems.

[0048] Reference numerals list

[0049] 1 Road

[0050] 2 Road edge

[0051] 3 Lane markings

[0052] 4 Guardrail

[0053] 5 Motor Vehicles

[0054] 6 Environmental sensor device

[0055] 7 Auxiliary systems

[0056] 8 Interfaces

[0057] 9 Processor

[0058] 10 Data storage

[0059] 11 Measurement points

[0060] 12 Third-party vehicles

[0061] 13 Trajectory

[0062] 14 Driving Path Estimation

[0063] 15 Benchmarks

[0064] 16 Steering wheel

[0065] 17 Driving route

Claims

1. A method for repeatedly estimating a current driving path (14) of a motor vehicle (5), wherein during operation of the motor vehicle (5) the following is periodically repeated: - detecting a plurality of measurement data (11) together with associated uncertainties, said measurement data characterizing a corresponding traffic situation of said motor vehicle (5), - thus estimating the updated information matrix and the updated information vector in the form of the information filter in the update step, respectively, and - determining an updated estimate of the driving path (14) from the update information matrix and the update information vector, respectively.

2. The method according to claim 1, It is characterized in that The measurement data (11) at least describe the course of the road ahead of the motor vehicle (5) at different distances from the motor vehicle (5) in each case in the direction of travel.

3. The method according to any one of the preceding claims, It is characterized in that The measurement data (11) are arranged in a measurement data vector and the associated uncertainties are arranged in an uncertainty matrix, and in the updating step, the individual measurement data (11) and their associated uncertainties are considered independently of one another without inverting the entire uncertainty matrix.

4. The method according to any one of the preceding claims, It is characterized in that Firstly, an update state of the driving path (14) is determined from the update information matrix and the update information vector, wherein the update state characterizes the driving path (14) by predetermined parameters, in particular geometric parameters, wherein the number of the parameters is smaller than the dimension of the measurement data (11).

5. The method according to claim 4, It is characterized in that The predefined parameters for characterizing the driving path (14) include a curvature of the driving path, in particular a curvature at a plurality of predefined distances (15) from the motor vehicle (5).

6. The method according to claim 4 or 5, It is characterized in that In order to determine the updated state of the driving path (14), the last determined update information matrix is ​​inverted, and the inverted information matrix is ​​multiplied by the last determined update information vector.

7. The method according to any one of claims 4 to 6, It is characterized in that A corresponding updated course of the driving path (14) in the corresponding surroundings is determined from the updated state using a predetermined model, wherein the model models at least one predetermined condition for a permissible actual driving path course (14).

8. The method according to any one of the preceding claims, It is characterized in that Environmental data describing the surroundings of the motor vehicle (5) and steering angle data of the motor vehicle (5) are respectively detected as part of the measurement data (11), and it is determined whether a lane change maneuver or a turning maneuver of the motor vehicle (5) is expected in a region from the respective current position of the motor vehicle (5) to a predetermined distance from the motor vehicle (5) in the direction of travel of the motor vehicle, and if this is the case, an updated estimate of the driving path (14) up to the predetermined distance is determined based on the steering angle data, and from the predetermined distance onwards, an updated estimate of the driving path (14) is determined based on the environmental data without taking into account the steering angle data.

9. An assistance system (5) for a motor vehicle (5), comprising an interface for detecting measurement data (11) characterizing a traffic situation of the motor vehicle (5), a processor device (9) and a computer-readable data memory (10) coupled to the processor device, wherein the assistance system (5) is configured to execute a method according to any one of the preceding claims.

10. A motor vehicle (5) having a surroundings sensor device (6) for recording surroundings data and having an assistance system (5) according to claim 9.

Images