DE102022000032A1 - Method for estimating a vehicle's own motion - Google Patents

Abstract

The invention relates to a method for estimating the movement of a vehicle, wherein an odometry (OF, OL, OR) of the vehicle is determined using a plurality of odometry determination units based on different data and at least one of the determined odometries (OF, OL, OR ) the movement of the vehicle itself is determined. In this case, the vehicle's own movement is determined as a function of a ascertained confidence from data which were ascertained by odometry ascertainment units which exceed a predefined confidence value.

Description

The invention relates to a method for estimating the movement of a vehicle according to the preamble of claim 1.

• - einen GPS-Empfänger,
• - einen Radarsensor,
• - eine Datenspeichervorrichtung, in der Programminstruktionen gespeichert sind,
• - eine symbolische Karte, die innerhalb der Datenspeichervorrichtung gespeichert ist, und
• - eine Steuereinheit, die mit der Datenspeichervorrichtung, dem GPS-Empfänger und dem Radarsensor wirkgekoppelt ist.

Die Steuereinheit ist dafür ausgestaltet, die Programminstruktionen auszuführen, um Daten von dem GPS-Empfänger, Daten von dem Radarsensor und Daten von der gespeicherten symbolischen Karte zu analysieren. Die Steuereinheit ist des Weiteren dafür ausgestaltet, einen probabilistischen Fahrzeugstandort auf Grundlage der Analyse zu bestimmen. Die symbolische Karte ist eine Karte mit fahrspurspezifischen Details und der probabilistische Fahrzeugstandort ist ein Standort innerhalb einer Fehler-Enveloppe eines berechneten GPS-Standorts, der höchstwahrscheinlich der tatsächliche Standort des Fahrzeugs ist. Das Analysieren der Daten von dem GPS-Empfänger, der Daten von dem Radarsensor und der Daten von der gespeicherten symbolischen Karte umfasst ein Bestimmen einer Geschwindigkeit für jedes der mehreren durch das Radar detektierten Objekte und ein Klassifizieren jedes der mehreren durch das Radar detektierten Objekte auf der Grundlage der bestimmten Geschwindigkeit. Das Klassifizieren eines jeden der mehreren Objekte umfasst ein Klassifizieren eines jeden der mehreren durch das Radar detektierten Objekte auf der Grundlage der bestimmten Geschwindigkeit als statisch oder dynamisch. Das Bestimmen eines probabilistischen Fahrzeugstandortes umfasst ein Generieren einer Liste möglicher Fahrzeugstandorte innerhalb der Fehler-Enveloppe des berechneten GPS-Standortes und ein Bestimmen einer einzelnen dynamischen Objektausrichtungswertung für jedes der mehreren Objekte, die als dynamisch klassifiziert wurden, für jeden der möglichen Fahrzeugstandorte, und ein Bestimmen einer einzelnen statischen Objektausrichtungswertung für jedes der mehreren Objekte, die als statisch klassifiziert wurden, für jeden der möglichen Fahrzeugstandorte. Eine jeweilige einzelne dynamische Objektausrichtungswertung basiert auf einem prognostizierten Standort eines jeweiligen der mehreren Objekte, die als dynamisch klassifiziert wurden, mit Bezug auf eine erste angenommene Mitte einer Fahrspur. Die jeweilige einzelne statische Objektausrichtungswertung basiert auf einem prognostizierten Standort eines jeweiligen der mehreren Objekte, die als statisch klassifiziert wurden, mit Bezug auf eine zweite angenommene Mitte einer Fahrspur. Die prognostizierten Standorte werden unter Verwendung der relativen Position eines jeden der detektierten Objekte mit Bezug auf die ausgewählte Fahrzeugposition unter Verwendung der Radardaten definiert. Die Steuereinheit ist des Weiteren dafür ausgelegt, innerhalb der Fehler-Enveloppe eine Anzahl von Fahrzeugpositionen auszuwählen, wobei ausgewählte Fahrzeugpositionen, die zu prognostizierten Positionen für dynamische detektierte Objekten führen, die näher bei einer Mitte einer Fahrspur liegen, höher gewertet werden als ausgewählte Fahrzeugpositionen, die zu prognostizierten Positionen für dynamische detektierte Objekte führen, die weiter von einer Mitte einer Fahrspur entfernt liegen. Ausgewählte Fahrzeugpositionen, die zu prognostizierten Positionen für statische detektierte Objekte führen, die weiter von einer Mitte einer Fahrspur entfernt liegen, werden höher gewertet werden als ausgewählte Fahrzeugpositionen, die zu prognostizierten Positionen für statische detektierte Objekte führen, die näher bei einer Mitte einer Fahrspur liegen. Ferner ist die Steuereinheit dafür ausgelegt, den Fahrzeugstandort mit der höchsten Wertung als den probabilistischen Fahrzeugstandort auszuwählen. Weiterhin wird ein Verfahren zum Lokalisieren eines Fahrzeugs beschrieben.From the EP 2 715 281 B1 a localization system within a vehicle is known. The localization system includes

• - a GPS receiver,
• - a radar sensor,
• - a data storage device in which program instructions are stored,
• - a symbolic map stored within the data storage device, and
• - a control unit operatively coupled to the data storage device, the GPS receiver and the radar sensor.

The control unit is configured to execute the program instructions to analyze data from the GPS receiver, data from the radar sensor and data from the stored symbolic map. The controller is further configured to determine a probabilistic vehicle location based on the analysis. The symbolic map is a map with lane specific detail and the probabilistic vehicle location is a location within an error envelope of a computed GPS location that is most likely the vehicle's actual location. Analyzing the data from the GPS receiver, the data from the radar sensor, and the data from the stored symbolic map includes determining a velocity for each of the plurality of radar-detected objects and classifying each of the plurality of radar-detected objects on the basis of the determined speed. Classifying each of the plurality of objects includes classifying each of the plurality of objects detected by the radar as static or dynamic based on the determined velocity. Determining a probabilistic vehicle location includes generating a list of possible vehicle locations within the error envelope of the computed GPS location and determining a single dynamic object orientation score for each of the multiple objects classified as dynamic for each of the possible vehicle locations, and determining a single static object orientation score for each of the multiple objects classified as static for each of the possible vehicle locations. Each individual dynamic object orientation score is based on a predicted location of each of the plurality of objects classified as dynamic with respect to a first assumed center of a lane. The respective individual static object alignment score is based on a predicted location of each of the plurality of objects classified as static with respect to a second assumed center of a lane. The predicted locations are defined using the relative position of each of the detected objects with respect to the selected vehicle position using the radar data. The control unit is further configured to select a number of vehicle positions within the error envelope, where selected vehicle positions that result in predicted positions for dynamic detected objects that are closer to a center of a lane are weighted higher than selected vehicle positions that lead to predicted positions for dynamic detected objects that are farther from a center of a lane. Selected vehicle positions that result in predicted positions for static detected objects that are farther from a center of a lane will be ranked higher than selected vehicle positions that result in predicted positions for static detected objects that are closer to a center of a lane. Furthermore, the control unit is configured to select the vehicle location with the highest score as the probabilistic vehicle location. A method for locating a vehicle is also described.

The invention is based on the object of specifying a novel method for estimating the movement of a vehicle.

The object is achieved according to the invention by a method which has the features specified in claim 1.

In a method for estimating a vehicle's own movement, an odometry of the vehicle is determined using a plurality of odometry determination units based on different data in each case, and the vehicle's own movement is determined from at least one of the determined odometries.

• - wird in einem ersten Verfahrensschritt eine absolute Ego-Position des Fahrzeugs relativ zu einer festen Referenzposition ermittelt,
• - wird in einem zweiten Verfahrensschritt in einem Merkmalsraum, welcher sich in einem definierten Umkreis um die absolute Ego-Position erstreckt, ein Grad einer Urbanisierung anhand von in einem Synthetischen-Apertur-Radarbild enthaltenen Daten und/oder anhand einer digitalen Straßenkarte ermittelt,
• - wird in einem dritten Verfahrensschritt in einem weiteren Merkmalsraum, welcher sich in einem definierten Umkreis um die absolute Ego-Position erstreckt, ein Grad einer Bodenfeuchte und eines Fahrbahnzustands anhand von in einem Synthetischen-Apertur-Radarbild enthaltenen Daten ermittelt,
• - wird in einem vierten Verfahrensschritt in einem weiteren Merkmalsraum, welcher sich in einem definierten Umkreis um die absolute Ego-Position erstreckt, eine Topografie mit einem Grad von Oberflächenunebenheiten anhand eines digitalen Höhenmodells ermittelt,
• - wird in einem fünften Verfahrensschritt in einem weiteren Merkmalsraum, welcher sich in einem definierten Umkreis um die absolute Ego-Position erstreckt, ein Grad einer Dynamik von Umgebungsobjekten anhand von in einem Synthetischen-Apertur-Radarbild enthaltenen Daten ermittelt,
• - werden die vorherigen Verfahrensschritte mehrfach zu definierten Zeitpunkten wiederholt,
• - werden in einem sechsten Verfahrensschritt mittels einer lokalen Entscheidungslogik für die in den wiederholt ausgeführten vorherigen Verfahrensschritten ermittelten Merkmalsräume lokale Priorisierungen ermittelt,
• - werden in einem siebten Verfahrensschritt die lokalen Priorisierungen jedes Merkmalsraums für jeden Merkmalsraum zu einer globalen Priorisierung auf Basis einer jeweiligen Gewichtung kombiniert,
• - wird in einem achten Verfahrensschritt eine Konfidenz der Odometrie-Ermittlungseinheiten auf Basis der globalen Priorisierung skaliert und
• - wird in einem neunten Verfahrensschritt die Eigenbewegung des Fahrzeugs in Abhängigkeit der Konfidenz aus den Daten bestimmt, welche von Odometrie-Ermittlungseinheiten, die einen vorgegebenen Konfidenzwert überschreiten, ermittelt wurden.

According to the invention

• - in a first method step, an absolute ego position of the vehicle is determined relative to a fixed reference position,
• - In a second process step, a degree of urbanization is determined in a feature space, which extends in a defined area around the absolute ego position, using a synthetic aperture data contained in the radar image and/or determined using a digital road map,
• - in a third method step, a degree of soil moisture and a road condition is determined in a further feature space, which extends in a defined radius around the absolute ego position, using data contained in a synthetic aperture radar image,
• - in a fourth method step, a topography with a degree of surface unevenness is determined in a further feature space, which extends in a defined area around the absolute ego position, using a digital elevation model,
• - in a fifth method step, a degree of dynamics of surrounding objects is determined in a further feature space, which extends in a defined area around the absolute ego position, using data contained in a synthetic aperture radar image,
• - the previous procedural steps are repeated several times at defined times,
• - in a sixth method step, local prioritizations are determined by means of a local decision logic for the feature spaces determined in the repeatedly executed previous method steps,
• - in a seventh method step, the local prioritizations of each feature space are combined for each feature space to form a global prioritization based on a respective weighting,
• - In an eighth method step, a confidence of the odometry determination units is scaled on the basis of the global prioritization and
• - In a ninth method step, the vehicle's own movement is determined as a function of the confidence from the data which were determined by odometry determination units which exceed a predetermined confidence value.

A large number of driver assistance functions are provided in vehicles for automated driving operation of the vehicles or selected vehicle functions, such as automated parking. Reliable trajectory planning for the respective vehicle and reliable perception of a vehicle environment are required for these driver assistance functions. This in turn requires precise information about a self-motion or a reliable self-motion estimate of the ego vehicle, also referred to as motion estimation.

In the present case, various odometry determination units are used for this self-motion estimation. These can, for example, determine a chassis-supported odometry, with a self-motion estimation being based on data from chassis sensors and/or wheel speed sensors and a wheel speed, a steering angle, an acceleration and other values being able to be derived via cycles. However, compared to their environment, these sensors exhibit a high degree of variance in their error quality. For example, if the wheels lose contact with the roadway and this results in slippage or because a wheel has been changed or a wheel diameter has changed, incorrect movement estimates can occur.

Furthermore, a radar-based odometry can be determined for estimating the own movement using the odometry determination units, with features of radar sensor units in the vehicle being taken into account for this purpose and the features being tracked over time and the own movement being derived. This means that in the radar-based estimation of the own movement, the features are recorded via the radar sensor units, partially classified and then tracked in order to derive the own movement. However, in the case of movements performed on non-planar surfaces, a sudden loss of features due to highly dynamic scenarios and a feature space with few features, for example in the case of so-called rural drives, there can be a high susceptibility to errors.

Furthermore, for the purpose of estimating one's own motion, the odometry determination units can be used to determine an odometry based on data from inertial measurement units and position data of the vehicle determined using a satellite-based position determination system. For example, the vehicle's own movement can be determined by several acceleration and speed sensors. However, only relative movement can be derived, which is why the combination with the satellite-based position determination system is carried out. However, this results in disadvantages in absolute position determination, primarily in urban environments, due to the lack of coverage of the satellite-based position determination system.

The present method enables a combination of different methods for estimating one's own motion, with corresponding disadvantages of the methods being evaluated situationally and thus odometry determination units and their determined data can be prioritized. This is done using a satellite-based approach, in wel chem synthetic aperture radar images and multidimensional feature spaces can be used. This means that the method enables prioritization for self-motion estimation based on so-called multi-source odometry units.

In this case, the prioritization can take place in a particularly advantageous manner independently of the vehicle's own sensor system. Furthermore, an optimized utilization of the odometry determination units is achieved due to their prioritization and the method can be used independently of the prevailing weather and light situations, independent of day and night times, independent of own, lateral and longitudinal movements, independent of a parameterization of the own vehicle (e.g. a wheel size etc.) and can be carried out independently of space and time. Any extensions of the respective feature space are possible and no further odometry information, for example an intrinsic speed, etc., is required to carry out the method.

Exemplary embodiments of the invention are explained in more detail below with reference to drawings.

• 1 schematisch einen Ablauf eines Verfahrens zur Schätzung einer Eigenbewegung eines Fahrzeugs,
• 2 schematisch eine Visualisierung eines ersten Verfahrensschritts des Verfahrens gemäß 1,
• 3 schematisch eine Visualisierung eines zweiten Verfahrensschritts des Verfahrens gemäß 1,
• 4 schematisch eine Visualisierung eines dritten Verfahrensschritts des Verfahrens gemäß 1,
• 5 schematisch eine Visualisierung eines vierten Verfahrensschritts des Verfahrens gemäß 1,
• 6 schematisch eine Visualisierung eines fünften Verfahrensschritts des Verfahrens gemäß 1,
• 7 schematisch eine Visualisierung eines sechsten Verfahrensschritts des Verfahrens gemäß 1,
• 8 schematisch eine Visualisierung eines siebten Verfahrensschritts des Verfahrens gemäß 1,
• 9 schematisch eine Visualisierung eines achten Verfahrensschritts des Verfahrens gemäß 1 und
• 10 schematisch eine Visualisierung eines neunten Verfahrensschritts des Verfahrens gemäß 1.

show:

• 1 schematically a sequence of a method for estimating a movement of a vehicle,
• 2 schematically a visualization of a first step of the method according to 1 ,
• 3 schematically a visualization of a second step of the method according to 1 ,
• 4 schematically a visualization of a third step of the method according to 1 ,
• 5 schematically a visualization of a fourth step of the method according to 1 ,
• 6 schematically a visualization of a fifth step of the method according to 1 ,
• 7 schematically a visualization of a sixth step of the method according to 1 ,
• 8th schematically a visualization of a seventh step of the method according to 1 ,
• 9 schematically a visualization of an eighth step of the method according to 1 and
• 10 schematically a visualization of a ninth step of the method according to 1 .

Corresponding parts are provided with the same reference symbols in all figures.

In 1 a sequence of a possible exemplary embodiment of a method for estimating a movement of a vehicle with the method steps VS1 to VS10 is shown. the 2 until 10 each show a visualization of method steps VS1 to VS9 of the method.

In the exemplary embodiment shown, the method comprises nine method steps VS1 to VS9, with several odometry determination units using different data in each case to determine an odometry OF, OL, OR of the vehicle and from at least one of the determined odometries OF, OL, OR the movement of the vehicle Vehicle is determined.

The odometries OF, OL, OR are determined using the following methods.

In a first method, an in 7 chassis-supported odometry OF described in more detail, with a self-motion estimation being based on data from chassis sensors and/or wheel speed sensors and a wheel speed, a steering angle, an acceleration and other values being able to be derived via revolutions.

In a second method, an in 7 radar-based odometry OR, shown in more detail, is determined, for which purpose features of radar sensor units in the vehicle are taken into account and the features are tracked over time and the own movement is derived.

In a third method, an in 7 odometry Ol shown in more detail based on data from inertial measurement units and using a satellite-based position determination system determined position data of the vehicle. For example, the vehicle's own movement is determined by several acceleration and speed sensors and combined with data from the satellite-supported positioning system.

Using the present method, a prioritization for a self-motion estimation of a vehicle based on multi-source odometry determination units is made possible. For this purpose, satellite-based input data in the form of in the 2 until 6 Syntheti shown in more detail cal-aperture radar images SARB1 to SARB4 are used to construct a multi-dimensional feature space M1 to M4.

First of all, in a first method step VS1, an absolute ego position POS1 of the vehicle is determined relative to a fixed reference position.

In a second method step VS2, a degree Sat_U of an urbanization is determined in a feature space M1, which extends in a defined radius, for example 50 m, around the absolute ego position POS1, using data contained in a synthetic aperture radar image SARB1 and/or or determined or extracted using a digital road map, for example a so-called open street map. The urbanization represents a degree of infrastructure in an immediate environment of the vehicle that is to be defined. These include, for example, buildings, traffic lights, sidewalks, etc. The degree of urbanization Sat_U can be specified, for example, in a range from [0, . . . , 1]. will.

In a third method step VS3, in a further feature space M2, which extends in a defined radius, for example 50 m, around the absolute ego position POS1, a degree Sat_B of soil moisture and a roadway condition is determined using in a synthetic aperture radar image SARB2 contained data determined. Based on a mode of operation of synthetic aperture radars, the state of the road and water, ice or a covering of snow on a road can be derived on the basis of a frequency band, also referred to as X,C,L band, a scatter distribution and a polarization. The feature space M2 can correspond to the feature space M1. The degree Sat_B of the soil moisture and the condition of the road can be specified in a range of [0, . . . , 1], for example.

In a fourth method step VS4, in a further feature space M3, which extends in a defined radius, for example 50 m, around the absolute ego position POS1, a topography with a degree Sat_O of surface irregularities is created using a digital elevation model in a synthetic aperture Radar image SARB3 determined. For example, what is known as a digital elevation model, which models a topography of the individual movement, is used here as a further feature space M3. Such models are available, for example, as open source solutions. The feature space M3 can correspond to the feature space M1 and/or the feature space M2. The degree Sat_O of the asperities can be specified in a range of [0,...,1], for example.

In a fifth method step VS5, in a further feature space M4, which extends in a defined radius, for example 50 m, around the absolute ego position POS1, a degree Sat_D of a dynamic of surrounding objects is determined based on contained in a synthetic aperture radar image SARB4 data determined. In this case, a movement classification on synthetic aperture radar images SARB4 is used as a further feature space M4. That movement classification is implemented, for example, by means of machine learning methods, for example by means of what is known as moving target detection or other methods. The feature space M4 can correspond to the feature space M1 and/or the feature space M2 and/or the feature space M3. The degree Sat_D of the dynamics of surrounding objects can be specified in a range of [0,...,1], for example.

The feature spaces M1 to M4 are now determined in every nth time stamp in the defined environment of the vehicle, for example in a radius of 50 m or 100 m. This means that the previous method steps VS1 to VS5 are repeated several times at defined points in time.

In a sixth method step VS6, local prioritizations are determined by means of a local decision logic for the feature spaces determined in the previously executed method steps VS1 to VS5. This means that on the basis of the multi-dimensional feature spaces M1 to M4, the decision logic with a subsequent combinatorics is used in order to establish the prioritization between the odometry determination units. The decision logic is, for example, a so-called fuzzy logic. A membership function of the fuzzy logic for each odometry determination unit and each feature space M1 to M4 is determined by an initial input data set, ie the different grades Sat_U, Sat_B, Sat_O and Sat_D. how 7 shows, the odometries OF, OL, OR are determined for the degree Sat_U of urbanization, for the degree Sat B of soil moisture and the condition of the road, for the degree Sat_O of surface unevenness and for the degree Sat_D of the dynamics of surrounding objects in the various feature spaces M1 to M4.

As a result, as in 8th exemplified for the Sat_U degree of urbanization and the Sat_O degree of asperity, for each satellite-based feature a scalar, which are combined by a prior weighting. For this purpose, in a seventh step VS7 combines the local prioritization of each feature space M1 to M4 for each feature space M1 to M4 to form a global prioritization PR, PI based on the respective weighting. For the visualizations shown, this results in a prioritization PR for the radar-based odometry OR and a prioritization PI for the odometry OL determined based on data from inertial measurement units and position data determined by means of a satellite-based position determination system.

As in 9 shown, based on the resulting prioritization PF, PI, PR, a confidence of the odometry determination units is scaled in an eighth method step VS8 on the basis of the global prioritization PF, PI, PR. The inputs of all odometry determination units are still taken into account, but their confidence is controlled by the prioritizations PF, PI, PR. For example, when driving on rural routes, the confidence of the radar-based odometry OR is degraded, whereas a higher confidence is assigned to the odometry OL determined based on data from inertial measurement units and using a satellite-based position determination system. Advantageously, there is a higher availability of satellite-based data, which is at least essentially spatially and temporally, in particular as a function of a
Position POS1 to POSn of the vehicle and a service provider are independent. In addition, an optimal utilization of the odometry determination units can be achieved and the feature spaces M1 to M4 can be expanded as desired.

As in 10 shown then, in a ninth method step VS9, the vehicle's own movement is determined as a function of the confidence from the data determined by odometry determination units that exceed a predetermined confidence value.

In a tenth method step VS10, a cyclic repetition of all previous method steps VS1 to VS9 is then initialized, for example after a predetermined number of time stamps, for example 600 s.

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• EP 2715281 B1 [0002]

Claims (1)

A method for estimating a vehicle's own movement, wherein - using a plurality of odometry determination units based on different data in each case, an odometry (OF, OL, OR) of the vehicle is determined and - from at least one of the determined odometries (OF, OL, OR) the Own movement of the vehicle is determined, characterized in that - in a first method step (VS1) an absolute ego position (POS1 to POSn) of the vehicle is determined relative to a fixed reference position, - in a second method step (VS2) in a feature space ( M1), which extends within a defined radius around the absolute ego position (POS1 to POSn), a degree (Sat_U) of urbanization based on data contained in a synthetic aperture radar image (SARB1) and/or based on a digital road map is determined, - in a third method step (VS3) in a further feature space (M2), which is in a defined area around the abs olute ego position (POS1 to POSn), a degree (Sat_B) of soil moisture and of a roadway condition is determined using data contained in a synthetic aperture radar image (SARB2), - in a fourth method step (VS4) in a further feature space (M3), which extends in a defined radius around the absolute ego position (POS1 to POSn), a topography with a degree (Sat_O) of surface irregularities is determined using a digital elevation model in a synthetic aperture radar image SARB3, - in a fifth method step (VS5) in a further feature space (M4), which extends in a defined radius around the absolute ego position (POS1 to POSn), a degree (Sat_D) of a dynamic of surrounding objects based on a synthetic aperture Radar image (SARB4) contained data is determined, - the previous method steps (VS1 to VS5) are repeated several times at defined times, - in a sixth Ver process step (VS6) local prioritizations are determined by means of a local decision logic for the feature spaces (M1 to M4) determined in the repeatedly executed previous process steps (VS1 to VS5), - in a seventh process step (VS7) the local prioritizations of each feature space (M1 to M4 ) are combined for each feature space (M1 to M4) to form a global prioritization (PF, PI, PR) based on a respective weighting, - in an eighth method step (VS8), a confidence of the odometry determination units based on the global prioritization (PF, PI, PR) is scaled and - in a ninth method step (VS9), the vehicle's own movement is determined as a function of the confidence from the data which were determined by odometry determination units which exceed a predetermined confidence value.

Images