DE102017216442A1 - Automatic parking system - Google Patents

Abstract

The present invention relates to an automatic parking system and a corresponding automatic method for parking a motor vehicle, in which possible candidates for free parking spaces are automatically determined and selected. This is done on the basis of a population state map in which sensor data of one or more vehicle sensors are fused. The data are prepared accordingly and visualized for a driver so that he can follow the process at any time and if necessary intervene. For a selected parking space, a route is planned and, optionally after confirmation by the driver, the automatic parking process is carried out.

Description

BACKGROUND OF THE INVENTION

Field of the invention

The present invention relates to an automatic parking system. In particular, the present invention relates to an automatic parking system, which is able to independently detect vacant parking spaces in a parking lot.

State of the art

The present invention relates to the field of automotive engineering, in particular driver assistance systems (FAS, also known by the name "Advanced Driver Assistance System", ADAS for short). Recently, automatic parking systems (also referred to as automatic parking assistance systems) have received much attention because they provide valuable assistance in particularly challenging driving situations and thus significantly increase ride comfort. In particular, in the field of sensor technology, such as finding obstacles such as already parked vehicles that exclude a certain area of a motor vehicle in a parking lot as (free) parking space, great progress has been made in recent years. As vehicle sensors come in the field of automatic parking systems today in addition to ultrasound in particular radar (radio detection and ranging), lidar (light detection and ranging) and camera-based sensors used.

Automatic parking systems customary today are based primarily on ultrasonic sensors for detecting obstacles and free spaces. Due to the low resolution and limited detection range of ultrasonic sensors, ultrasound-based parking assistance systems have some principal weaknesses. First, the vehicle must be moved sufficiently close (for example, to about 2 meters (m)) to a vacant parking space to detect them. Furthermore, the ultrasonic sensor (sonar) can detect only the front but not the sides of the parked vehicles. This is in 2 shown. Therefore, the orientation and the depth of the free parking space can not be determined precisely in a single drive past a motor vehicle. Finally, there is a high level of uncertainty and a high detection error rate, which requires additional movement to confirm a detected parking space. In practical terms, this means that the motor vehicle has to perform several maneuvers (back and forth) before the rear ultrasonic sensors can detect the position of a target parking space with sufficient accuracy.

It is therefore increasingly being used to also use other types of environmental sensors on the vehicle, in particular the above-listed types of sensors. In addition to ultrasonic sensors, fish-eye cameras with a wide viewing angle are becoming increasingly popular for the environmental detection of modern commercial vehicles in "all-round view". However, this entails new challenges in terms of processing the plurality of data acquired by one or more sensors in order to prepare them so that they are maximally usable for an automatic parking assistance system. Methods for preparing sensor data for map representations of the environment of a sensor-equipped object and sensor fusion are, for example, in the articles "Dynamic Map Building for an Autonomous Mobile Robot" by John J. Leonard, Hugh F. Durrant-Whyte and Ingemar J. Cox, The International Journal of Robotics Research, Vol. 4, August 1992, pages 286-298 and "Sensor Fusion for Flexible Human-Portable Building-Scale Mapping" by Maurice F Fallon, Hordur Johansson, Jonathan Brookshire, Seth Teller and John J. Leonard, Proceedings often he 2012 IEEE / RSJ International Conference on Intelligent Robotics and Systems (IROS), p. 4405-4412 described.

What is still missing in particular is an efficient method of displaying (visualizing) the vehicle environment and enabling the detection of parking spaces while at the same time avoiding unclear situations that can lead to wrong decisions and resulting traffic accidents.

SUMMARY OF THE INVENTION

The present invention aims to efficiently use an improved automatic parking system capable of efficiently utilizing a plurality of data acquired from one or more vehicle sensors of a vehicle for automatic parking, and thus to increase overall user convenience, as well as to provide a corresponding method ,

This is achieved with the features of the main claim 1 and the independent claim 10.

According to a first aspect of the present invention, an automatic parking system for a motor vehicle is provided. The automatic parking system comprises a memory device for storing a population state map of a vehicle environment of the motor vehicle, the population state map comprising a two-dimensional grid of cells, each of which is associated with a parameter value representing one of the possible occupation states "free", "occupied" and " unknown ". The automatic parking system further comprises a calculation device configured to determine vacant areas of the vehicle environment based on the occupation state map as candidate for parking spaces and a visualization device configured to visualize the vehicle environment including the identified candidates for parking spaces and to visualize a parking space selected from the determined candidates. Finally, the automatic parking system comprises a planning device configured to determine a planned travel path from a current vehicle position into the selected parking space, taking into account given vehicle parameters, and a control device configured to automatically steer the motor vehicle into the selected parking space along the planned travel path.

According to a second aspect of the present invention, there is provided a method for automatically parking a motor vehicle with an automatic parking system. The automatic parking system includes storage means for storing a population status map of a vehicle environment of the motor vehicle, the occupancy state map comprising a two-dimensional grid of cells, each associated with a parameter value that assumes one of the possible occupation states "free", "occupied" and "unknown" , The method includes the steps of automatically determining vacant areas of the vehicle environment based on the occupancy map as candidates for parking spaces, visualizing the vehicle environment including the identified candidates for parking spaces, selecting a parking space from the identified candidates and visualizing the selected parking space, scheduling a travel path from a current vehicle position in the selected parking space, taking into account given vehicle parameters and the automatic control of the motor vehicle in the selected parking space along the planned route.

It is the particular approach of the present invention to provide an automatic parking assist system which makes it possible to visually represent the occupied / vacant states of parking spaces in a vehicle environment based on a occupancy state map as a mathematical model of the vehicle environment (an exterior of the vehicle). In particular, the population status card may be the result of sensor fusion of data from one or more vehicle sensors. The visualized information is then used for automatic parking.

Preferably, the occupancy status map is the result of a sensor fusion of data determined by at least one vehicle sensor of the motor vehicle. An occupied state card thus determined is therefore also called a sensor fusion card. Sensor fusion in only one vehicle sensor is to be understood, for example, as meaning that the vehicle sensor during a driving operation of the vehicle takes time in succession data, which are then merged in a subsequent sensor fusion. In other words, the vehicle drives over the parking lot and data relating to different areas of the parking lot are successively recorded and merged. Sensor fusion in multiple vehicle sensors involves both a collection of data taken in succession and taken by different sensors.

More preferably, the occupancy state map comprises occupancy information at least with respect to cells in the region of a front line of objects located during the data acquisition by the sensors along one side of the motor vehicle. The calculation device is preferably set up to determine a free area as a candidate for a parking space by determining a front line of an object as a straight line running through a plurality of occupied cells along the side of the motor vehicle, an orientation of the object after one predetermined predetermined algorithm based on the front line, defining a central axis of a free area as a candidate for a parking space based on the thus determined orientation of an object or two adjacent objects and a front line of the free area perpendicular to the fixed central axis so determined that this substantially with the front line of the one or two adjacent objects is aligned.

Such an approach is particularly suitable when using only ultrasonic or radar sensors, as they provide as already stated no depth information. As "objects" are expected in the assumed situation of driving a parking lot especially other, already parked on parking spaces of the parking lot motor vehicles. The above-described exemplary method for determining a candidate for a free parking space is therefore particularly applicable if only the positions of the front fronts of the vehicles already parked, for example by ultrasound or radar sensors, are known. For the occupancy status card, this means that cells in the front fronts area of already parked vehicles are marked as occupied. Other cells in the immediate vicinity of the vehicle equipped with the sensors, in the area of which there are no other objects, are marked as free. These are in particular cells in the area of a free parking space, the one equipped with the sensors Motor vehicle was directly adjacent to the data recording. All other cells, in particular cells behind the front line of a recognized object or cells of a vacant parking space removed further from the motor vehicle equipped with the sensors, are marked as "unknown".

Even more preferably, in the predetermined algorithm, the orientation of the object is determined to be perpendicular to the front line. As a result, an orientation of the object can be determined without requiring additional data about cells of the population status card other than those in the area of the front line.

According to another preferred embodiment, in addition to the occupancy information of cells in the area of the front line, the occupation state map also includes occupation information regarding cells in the area of side edges of objects located during the data acquisition by the sensors along one side of the motor vehicle. The predetermined algorithm determines a sideline of an object as a cell characterized by a plurality of busy as in a direction substantially perpendicular to the front line and further defines an orientation of the object as the orientation of the sideline. By using the additional information in this case, a more precise determination of the position of the object, as well as an estimate of the size of the object is possible.

Such a procedure is particularly suitable if (in addition to ultrasound or radar sensors or exclusively) lidar devices and / or cameras are available, since they can additionally provide depth information, such as information about pages of objects.

In addition, the calculation device additionally uses information recorded by a vehicle sensor about lane markings of parking spaces for determining the central axis of a free area as a candidate for a parking space, provided that these lane markings are not wholly or partially located during a data acquisition by the sensors along one side of the motor vehicle Objects are obscured. Sensors that are able to absorb depth information, ie information in a direction perpendicular to the vehicle side or direction of travel of the motor vehicle not only in the immediate vicinity, detect such parking space markings. However, when defining a free area as a possible candidate for a parking space, object boundaries, in particular boundaries of vehicles already parked, are to be given priority over lane markings.

Preferably, the data is recorded by the vehicle sensors while driving, in particular while driving on a parking space.

Preferably, the visualization device additionally visualizes the planned infrastructure.

The selection of the parking space from the determined candidates preferably takes place automatically. Alternatively, preferably, the parking space is selected by the driver by a user input based on the visualized vehicle environment.

Preferably, the visualization device is further configured to visualize a respective continuous section of an automatic parking process. This can be realized, for example, by fading in a corresponding designation on a display. Possible designations for individual sections of the parking process are, in particular, "search", "parking space found or detected", "parking operation in progress" and "parking process completed".

Preferably, the at least one vehicle sensor comprises an ultrasonic sensor, a radar device, a camera or a lidar sensor.

Preferably, a message is additionally issued to the driver of the motor vehicle with the request to accept a selected parking space. The driver must then confirm the selected parking space by a user input, such as by pressing a button on a touch screen, before the route is planned and the automatic parking process is initiated. This possibility is particularly important in connection with an automatic selection of the parking space. If the driver does not acknowledge / reject the selected parking space (such as by other user input, or by not performing the user input intended for acceptance within a predetermined time), then the automatic parking system may perform a service in accordance with the available vehicle environment data suggest other free parking spaces if several candidates for parking spaces have been identified. Alternatively, in this case, the automatic parking operation can be aborted.

Also preferably, the driver is informed of the successful completion of the automatic parking operation by issuing a corresponding message after completion of the automatic parking process. Such a message can be made, for example, acoustically, visually (graphically) or by a predetermined vibration process.

Further features and advantages of the present invention are the subject of dependent claims.

list of figures

• 1 einen allgemeinen Überblick über die Funktionsweise eines automatischen Parksystems gemäß der vorliegenden Erfindung gibt;
• 2 eine Übersichtszeichnung ist, die die Ermittlung eines Kandidaten für eine freie Parklücke gemäß einer Ausführungsform der vorliegenden Erfindung für einen Fall zeigt, in welchem die Sensordaten lediglich Informationen bezüglich einer dem vorbeifahrenden Kraftfahrzeug zugewandten Seite von Objekten (geparkten Fahrzeugen) liefern;
• 3 eine detailliertere Darstellung der Situation aus 2 für verschiedene Fälle ist;
• 4 eine weitere Detaildarstellung für eine ähnliche Situation wie in 2 ist;
• 5 das Auffinden eines Kandidaten für eine freie Parklücke für einen Fall darstellt, wenn sich keine weiteren Objekte in dem relevanten Umgebungsbereich des Kraftfahrzeugs befinden;
• 6 eine Übersichtszeichnung ist, die die Ermittlung eines Kandidaten für eine freie Parklücke gemäß einer Ausführungsform der vorliegenden Erfindung für einen Fall zeigt, in welchem die Sensordaten Informationen sowohl bezüglich einer dem vorbeifahrenden Kraftfahrzeug zugewandten Seite als auch Tiefeninformation bezüglich einer zu letzterer im wesentlichen senkrecht verlaufenden Seite eines Objekts (geparkten Fahrzeugs) liefern;
• 7 eine detailliertere Darstellung der Situation aus 6 für verschiedene Fälle ist;
• 8 eine weitere Detaildarstellung für eine ähnliche Situation wie in 6 ist;
• 9 das Auffinden eines Kandidaten für eine freie Parklücke für einen Fall veranschaulicht, wenn zusätzlich Fahrbahnmarkierungen für Parklücken vorhanden sind und von Fahrzeugsensoren erfasst werden;
• 10 einen weiteren Fall veranschaulicht, bei dem Fahrbahnmarkierungen für Parklücken vorhanden sind, aber teilweise von Objekten (geparkten Fahrzeugen) verdeckt sind;
• 11 eine Veranschaulichung einer Umgebung eines Kraftfahrzeugs auf der Grundlage einer Besetzungszustandskarte ist;
• 12 die Auswahl eines von mehreren Kandidaten für eine Parklücke gemäß einer Ausführungsform der vorliegenden Erfindung sowie die Planung eines Fahrwegs veranschaulicht;
• 13 ein Beispiel für eine Visualisierung der Fahrzeugumgebung einschließlich einer ausgewählten Parklücke und eines geplanten Fahrwegs zeigt; und
• 14 ein Ablaufdiagramm für einen automatischen Parkvorgang gemäß einer Ausführungsform der vorliegenden Erfindung ist.

Additional features and advantages of the present invention will be set forth below in the detailed description and illustrated in the accompanying drawings, in which:

• 1 gives a general overview of the operation of an automatic parking system according to the present invention;
• 2 FIG. 11 is an overview drawing showing the determination of a vacancy candidate according to an embodiment of the present invention for a case where the sensor data merely provides information regarding a passing vehicle side of objects (parked vehicles); FIG.
• 3 a more detailed account of the situation 2 for different cases;
• 4 a further detail for a similar situation as in 2 is;
• 5 finding a candidate for a free parking space for a case when there are no other objects in the relevant surrounding area of the motor vehicle;
• 6 FIG. 11 is an outline drawing showing the determination of a candidate for a free parking space according to an embodiment of the present invention for a case where the sensor data includes information regarding both a passing vehicle side and depth information with respect to a substantially perpendicular side of the vehicle Object (parked vehicle);
• 7 a more detailed account of the situation 6 for different cases;
• 8th a further detail for a similar situation as in 6 is;
• 9 finding a candidate for a free parking space for a case when additionally there are lane markings for parking spaces and are detected by vehicle sensors;
• 10 illustrates another case where lane markers for parking spaces exist but are partially obscured by objects (parked vehicles);
• 11 an illustration of an environment of a motor vehicle based on a staff condition map;
• 12 illustrates the selection of one of a plurality of parking slot candidates according to an embodiment of the present invention and the planning of a travel path;
• 13 shows an example of a visualization of the vehicle environment including a selected parking space and a planned travel path; and
• 14 Fig. 10 is a flow chart for an automatic parking operation according to an embodiment of the present invention.

In the various figures, the same reference numerals designate the same or corresponding features, so that a renewed description is optionally omitted.

DETAILED DESCRIPTION

The present invention relates to an automatic parking assist system based on a occupancy status map of a vehicle environment of a motor vehicle to be parked in a parking lot. For this purpose, the vehicle is equipped with vehicle sensors which record data during a parking search in the area of the designated parking space, the evaluation of which lead to the determination of possible suitable free parking spaces (candidates for parking spaces) on the basis of a population status card. The term "vehicle environment" is to be understood as an exterior area of the vehicle, for example a parking space or certain areas of a parking space. It is not necessarily assumed that the vehicle environment contemplated here completely surrounds the vehicle (includes an angle of 360 ° about the vehicle), but according to embodiments of the invention, also sections of the outer area covering a smaller angle are covered by the concept of the vehicle environment , Essentially only in the direction of travel left or only lying in the direction of travel to the right of the vehicle area. The exact size of the considered vehicle environment is not essential to the invention, it may e.g. Depending on the local conditions and / or the type of sensors that the vehicle is equipped.

A population state map is generated using a data fusion of sensor data. Vehicle sensors used herein include, for example, particularly ultrasonic sensors, cameras, radar and lidar devices. A sensor fusion is possible even if the vehicle is equipped with only a single sensor or a single type of sensors. There then takes place a sensor fusion of data which with the single sensor in succession during the Parksuchfahrt (and / or of several similar sensors parallel or in chronological order).

In accordance with preferred embodiments of the present invention, a grid-based population condition map is used. Here, the vehicle environment is divided into a two-dimensional grid of rectangular cells to create the occupation state map. Each cell is provided with a measure (parameter), which takes one of three possible states: "unknown", "free" and "occupied". This measure indicates that the respective cell in the evaluation of the data from the sensor fusion with sufficient probability (probability higher than a predetermined threshold) is occupied by an obstacle or free space. If the occupation status of a cell can not be determined with sufficient probability from the existing data, then this is marked with the parameter value "unknown". If, in particular, no data information is available about a cell, it will be marked "unknown". A high value of a occupation probability determined from the sensor fusion data thus indicates that the cell contains an obstacle and therefore can not be traveled on. A small value means that the cell is free. The population probabilities of the grid cells are accumulated over time in accordance with a particular sensor model. The occupancy state map is ultimately obtained as a representation of the vehicle environment in which each cell is provided with the above three states. Since the population state map is generated from the sensor fusion, it is also referred to as a sensor fusion card. This card is then used to identify possible parking spaces (candidates for parking spaces).

When candidates for parking spaces have been identified, a suitable parking space (target parking space) is either automatically selected by the automatic parking system (this may, for example, typically be the nearest suitable parking space) or manually selected by the driver. Then the route is planned and the vehicle is automatically controlled in the selected parking space.

The principle of the present invention is shown schematically in FIG 1 shown. On the left side are vehicle position data as input data 2 (from a navigation system, for example via GPS), as well as sensor data from a camera 4 , an ultrasound device 6 , a lidar device 8th and a radar device 10 shown. On the basis of this, the occupancy status card (sensor fusion card) 12 created. On the basis of the cast condition card 12 takes place in the block 14 illustrated parking space detection and the subsequent infrastructure planning (according to the selection of a particular parking space not shown here as a target parking space). block 16 illustrates the vehicle control. Since in this case the vehicle position and thus the relevant area of the vehicle environment changes, there is a feedback to the sensor fusion card 12 ,

As already stated, the map can be created and updated based on different sensor types. The most common cases here are 1.) the provision of sensor data only from ultrasound and / or radar sensors (frequency signals) and 2.) the provision of sensor data from cameras and / or lidar devices, or from data of all the mentioned sensor types.

In the first case, due to the design of the sensor, only information about the front of an object but no depth information can be obtained. Like, for example 2 can be seen, "see" side of the vehicle mounted sonar devices only the front of a parked vehicle, while they can not detect the sides of the vehicle. As a result, the sonars can not precisely detect the orientation and the depth of the parking space, which leads to a high error rate and usually requires additional vehicle movements in order to recognize a possible parking position or to recognize it with sufficient accuracy.

The situation is better in the second case. In this case, a three-dimensional structure along the side edges of the object can also be reconstructed.

Therefore, within the scope of the invention, different solutions have been developed to carry out a determination of free parking spaces, depending on the amount of available data in the occupancy state map. These will be described in more detail below with reference to the following drawings. By way of example, it is assumed here that parking spaces are arranged on a parking space perpendicular to a direction of travel in the parking search traffic ("parking transverse to the direction of travel"). However, the invention is applicable in an analogous manner in the case of parking parallel to the direction of travel or oblique parking.

First, with reference to the 2 to 4 an exemplary solution is described for the case that only information concerning the vehicle side of the search for a parking space motor vehicle during movement in the parking search side of objects (hereinafter referred to simply as "front") are available, as is the case with the use of ultrasound and / or radar sensors is the case. In this case, an object which is detected (more precisely: the front side is detected) usually a parked in a parking space of the parking lot vehicle, so that in In the following description (as well as in the accompanying drawings) specifically this case is considered by way of example.

Thus, only information about the front of objects (parked vehicles) is used here. This is in 2 in the form of black dots, each one labeled as "occupied" grid cell 25 and a straight line drawn approximately through these points 26 , which is about the course of the front of an object (here: vehicle front of an already parked vehicle 22 ), shown. Since no depth information is available, it can be assumed in this case that the cells located behind the "front lines" 26 are marked as "unknown". In the example shown it is assumed that already two vehicles 22 park in a section of a parking lot so as to be substantially parallel to each other but at a small angle to their respective central axes 28 stand. The rectangle shown in the middle 24 is intended to illustrate a determined candidate for a parking space, the procedure being as follows:

In a first step, a grouping of occupied cells 25 in the grid representation, for example, obtained by ultrasound and / or radar measurements, detected, and detected due to their location in the occupation map of the vehicle environment relative to the seeking vehicle as a possible front of a parked vehicle already. By approximately a straight line 26 is laid ("fitted") by these points, becomes a linear representation of the front of the object 22 (parked vehicle).

In the following second step, the bisector becomes perpendicular 28 the line 26 as the main axis of the found object 22 established.

The goal of the subsequent third step is to determine the main axis (central axis) of a free area suitable as a candidate for a parking space. Depending on the situation (corresponding to the occupation information taken from the sensor fusion card), this can be done in different ways.

In particular, as in 2 exemplified, it has been determined that two vehicles are already parked, being between the two front lines 26 a sufficiently large free space, possibly as a candidate 24 is eligible for a parking space, so is the central axis 29 a candidate to be accepted in free space 24 for a parking space corresponding to the average orientation of the major axes 28 the already parked vehicles 22 set to be from the endpoints adjacent to the free space L1 and L2 the lines 26 is about the same. In particular, the bisector of the two main axes 28 determined at their intersection formed angle and as the main axis 29 of the candidate 24 considered for a parking space. This case is called "Case 1" on the left side of the 3 shown. On the right side of the 3 is represented as "Case 2" the case that two already parked vehicles 22 parallel to each other. In this case, the middle axis of the candidate 24 for a parking space as parallel to the orientation (main axis) of the already parked vehicles 22 and equal complained of both vehicles (distance d) set.

On the other hand, if only a parked vehicle 22 is determined in the relevant area, then the central axis 29 of the candidate 24 for a parking space (hereinafter also referred to as "target axis") as a parallel to the central axis 28 of the already parked vehicle 22 established. The distance of the target axis 29 from the central axis 28 of the object is here according to known vehicle data of the own vehicle 20 chosen, so that your own vehicle 20 at a sufficient distance to the already parked vehicle 22 can be set up when its centerline of the fixed axis 29 equivalent. This case is in 4 illustrated. Here is the front line of the candidate 24 perpendicular to the target axis 29 and by the point TF , which is the projection of the endpoint L1 on the target axis 29 was determined.

Finally, the case should be considered that in a relevant area of the vehicle environment no further objects, such as already parked vehicles, are detected. In this case, the target axis becomes 29 as perpendicular to the direction of travel of the search vehicle 20 set, wherein the the track of the search vehicle 20 facing front of the candidate 24 a distance ("lateral distance") from the vehicle side of the search vehicle 20 Has. This situation is in 5 illustrated. When the entire car park or a relevant area thereof is in the area of a vehicle seeking a parking space 20 has neither markings, nor other vehicles are parked, so the vehicle 20 Park everywhere, with a certain distance between the track and the parking position, just the already mentioned "side clearance" is observed. In the described manner, a suitable parking space can be determined even in the rare case, if neither markings nor already parked vehicles give a basic orientation. The lateral distance is a configurable value and in particular depends on the width of the roadway available for the search run (for example, the distance of the front sides from parked vehicles in a parking situation such as in 11 ). A typical value for example, the side clearance may be 1.5m or a similar value.

In the following fourth step the candidate becomes 24 for a parking space ultimately set so that its orientation of the target axis 29 corresponds and its front line (perpendicular to the target axis 29 ) is substantially aligned with the front line of one or both of the already parked vehicles. As in 2 represented here is the center TF the frontline of the candidate 24 for a parking space in the middle of the projections of the candidate 24 facing endpoints of the lines 26 on the target axis 29 , If there is only one object on a page, as in 4 , so the center of the projection corresponds to the one endpoint L1 on the target axis 29 , The distance between L1 and or L2 and the page of the free area (candidate 24 ) is a configurable value, for example 0.3m or 0.5m. However, the invention is not limited to the values exemplified here and elsewhere, but any other meaningful value in the given circumstances, for example, a value in the range 0.3m to 0.5m or outside this range, is also within the scope of Invention possible. For the depth of the candidate 24 Due to the lack of any depth information, this is assumed to be the length of one's own vehicle 20 equivalent. The same applies to the width of the candidate 24 ,

Thus, in the case described, bearing in mind that there is no information about the pages of existing objects (parked vehicles) 22 there, the goal orientation 29 and the positioning of the candidate 24 a parking space only from the orientation 26 the front of already parked vehicles 22 (or other corresponding objects). This can lead to larger errors if the objects found 22 are asymmetric or the data provided by the data, for example ultrasound or radar sensors, are of low density. Therefore, it is usually necessary to drive several times in the parking search traffic along the relevant parking area in order to determine suitable parking spaces.

The following is with reference to the figures 6 to 8th Consider a case when (in addition to or alone) information from one or more cameras and / or one or more lidar sensors is available. In this case, the target position (position of a candidate 24 for a parking space) from the information about the front, the sides of the vehicle and the depth of free space between existing objects (already parked vehicles) determined as generally in 6 is shown. The procedure is as follows:

In a first step, the population information in the sensor fusion card is evaluated. Since it is assumed that the sensors used additionally also depth information, in particular side information of the vehicle 20 supply, here is assumed to be not just a group of cells corresponding to a front line of an object 22 but also a group of cells corresponding to a sideline of an object 22 is specified as "busy". In the presentation of the 6 correspond to the drawn without filling circles 25 (only along a front line of a parked vehicle 22 ) based on information from ultrasound or radar sensors as occupied cells of the occupation map and the points drawn with black filling 65 (along the front line as well as along the sideline) correspond to the occupation marks due to camera or lidar information.

Correspondingly, straight lines become again 26 (Front line) and 62 (Sideline) approximated by the cells determined to be occupied ("fitted"). The corresponding intersections C1 and C2 are called the corners of parked vehicles (objects) 22 certainly. The depth of the space between the points C1 and C2 is determined with the aid of the nearest object in the depth direction of the potential parking space, if such exists (such as the hatched object shown 60 in 6 ), or it will be a standard maximum length, about 10m or one on the length of your own vehicle 20 based value, assumed. According to the assumed depth of the space, the remaining ("back") corners of the parked vehicles, C3 and C4 , determined. The one represented by the polygon (shown as thick dotted line) with the four corners C1 . C2 . C3 and C4 limited space is a free space potentially available for parking within which a candidate 24 for a parking space is set. The depth of free space is shown in the drawings of the 6 . 8th . 9 and 10 illustrated by a double arrow.

The goal of the subsequent second step is to determine the target axis 29 , According to various possible situations, different cases are possible here again.

In particular, as in 6 by way of example, it has been determined that two vehicles are already parked with a not insignificant difference in their orientations, being between the two front lines 26 a free space, possibly as a candidate 24 is eligible for a parking space, so is the central axis 29 according to the average orientation of the two side lines C1C4 and C2C3 determined as shown in the left illustration in 7 is shown in more detail as "Case 1". In particular, the Intersection of the two side lines 62 the parked vehicles 22 determined, and the bisector of the two side lines 62 included acute angle as the target axis 29 established.

The case of two parallel parked vehicles 22 is in the right-hand illustration of 7 shown as "Case 2". This becomes the target axis 29 as the parallel to the two sidelines 62 the objects 22 set in the middle between the two sidelines 62 that is, at the same distance d ' to the sidelines 62 located.

The case of only one already parked object 22 is in 8th shown. This becomes the target axis 29 as parallel to the determined sideline 62 of the one object assumed. The distance between the target axis 29 and the sideline 62 of the object 22 is a configurable value that is again determined to have a sufficient safety margin (for example, between 0.25m and 0.5m) between the point C1 and the sideline of the candidate 24 guaranteed if for the width of the candidate 24 again the inherent width of the searching motor vehicle 20 Is accepted.

Of course, even with the presence of more precise sensors, it may happen that no objects are detected. Again, this is related to 5 described method applicable, so that can be dispensed with a new description here.

In the subsequent third step, the center becomes TF the frontline of the candidate 24 as the center of the projections of the points C1 and C2 on the target axis 29 determined and thereby achieved that the front line as the perpendicular to the target axis 29 through the point TF substantially with the already parked objects 22 flees. Exists only on one side of the candidate 24 an object 22 , so the projection defines the one corresponding corner ( C1 in 8th ) the center TF of the candidate's front line 24 , The line C3C4 is perpendicular to the target axis and their distance from the point TF corresponds to the depth of the candidate 24 for a parking space assumed free space.

With orientation, given by the target axis 29 and the center TF the front line is the candidate 24 clearly defined for a parking space. As already stated, the width and length of a parking space corresponding to the dimensions of the own vehicle 20 accepted. If the available space for a candidate 24 is not sufficient for this, such free space is not identified and presented as a candidate.

So if, as in the last example, the side edges 62 of objects 22 As a rule, a potential candidate may be known 24 for a target parking space already be determined with a single pass by with sufficient accuracy. Therefore, it becomes possible to perform the automatic parking maneuver in most practical cases without requiring several parking searches in the parking lot. Only if, for example, errors in the travel and speed measurement (for example, too many braking maneuvers) or system errors occur, for example because of problems on vehicle components, it may be possible that the position of the target parking space must be updated with respect to a current vehicle position, which more than Passage is required, but even then the invention should not require more than three passages for automatic parking.

The following are further variants of a determination of candidates 24 for parking spaces in the event that additionally there are lane markings recognized by vehicle sensors with reference to the drawings of 9 and 10 explained.

It should be noted that in this case the target axis 29 is to be determined according to the orientation of the marking lines of the parking space, if they are not obscured by adjacent objects. In this case, the free space bounding points become the end points of the parking space marks C1 ' . C2 ' . C3 ' and C4 ' defined (see the lines 90 in 9 ).

In case of occlusion, however, the ascertained side lines become 65 of objects 22 Prioritized, ie this is a higher priority for the definition of the orientation of the target axis 29 granted. In this case, the marker lines 90 the parking space is not used, as it is safer the objects 22 for the estimation of the free space to use (see 10 ).

The further steps of the procedure for the determination of candidates 24 for parking spaces according to embodiments of the present invention using parking space markings are analogous to those previously described, so that a further detailed description is omitted here.

The following is a description of a representation of the vehicle environment as it may be presented based on the occupation map determined by the sensor fusion and after performing the steps described above for determining possible candidates for parking spaces in an exemplary case.

An example of such a representation (visualization) is in 11 given. Here are the driveway of a vehicle 20 in the parking search traffic as well as freely determined areas in black. Of objects 22 , such as occupied areas already parked vehicles are shown in crosshatch. Thick white boundary lines represent a boundary as freely determined spaces outside the path of the park search traffic. Gray areas are those whose occupation status could not be determined with sufficient certainty even after all steps described above have been performed, the occupation status of which is thus unclear.

Among the free areas that are potential candidates for parking spaces, in preferred embodiments of the present invention, ultimately only those are selected as "valid" for display and for selection as a target parking space satisfying further criteria. In particular, the size of an area determined as free must be the size of one's own vehicle 20 taking into account additional uncertainties. Examples of possible candidates are in the presentation of 11 as white areas 24 characterized. From these "valid candidates", the automatic parking system can now either automatically select a parking space to be used for automatic parking, or prompt the driver for an appropriate user input to select a parking space to use. In the case of an automatic selection, for example, the valid candidate closest to a current vehicle position for which a practicable travel path for the parking can be determined is selected.

For a better understanding, it should be mentioned that the differently represented areas in 11 are not identical to the cells of the occupation state map, the representation of the 11 merely as a mathematical model. While the presentation in 11 is about 10m x 10m in area, the lattice cells of the population state map, which is determined based on sensor fusion and serves as a basis for further processing, are significantly smaller (for example, on the order of 5cm (centimeters) x 5cm).

A simplified illustration of a portion of an exemplary visualization of an area of a parking lot is shown in FIG 12 shown. In this case, areas shown in crosshatching (with vehicle outlines indicated below) are objects 22 , in particular already parked vehicles. These are additionally limited in the chosen representation by thick black lines. Free areas 24a . 24b and 24c as valid candidates for parking spaces are shown in white or filled with dots. These areas are bounded by thin black lines. Broken lines are also park markings 90 shown.

Of the plurality of "valid" candidates for parking spaces, one is now automatically or manually selected by the driver as the target parking space. In the example shown, the candidate becomes 24b selected as the target parking space. Once the selection has been made, a corresponding change in the visual representation is made, illustrating the determination of a valid candidate as "selected". In the picture, this is that the filling of the white area is omitted by dots and the corresponding area is displayed in pure white.

It should be noted that a colored visualization preferably takes place in a real system. For example, in a preferred embodiment, those shown in the drawings of FIGS 11 and 12 areas shown in cross-hatched form in brown, the areas shown in white in green and the areas provided with a dot-shaped "hatching" in 12 as well as the white border the in 11 shown in yellow. Accordingly, a candidate becomes 24 for a parking space in the case of its selection visualized by a striking color change from yellow to green. In addition, the thick black boundary lines of objects in 12 preferably visualized by another color, for example red. The own vehicle 20 is shown in blue or violet, for example. Of course, the above-mentioned colors are merely examples, and they can be freely replaced by other colors that are sufficiently distinguishable from each other for the visualization.

As already stated, such a representation of a vehicle environment or a section thereof is the result of a processing of all available information regarding boundaries of obstacles, free space and parking markings taken during at least one parking search run of the own vehicle 20 through the environment. A free space is considered a valid candidate for a parking space if and only if the space is larger than the size of the own vehicle. The target parking space 24b is ultimately chosen so that it can be used with maximum efficiency and practicability for parking, or is manually selected by the driver from the proposed candidates 24 selected. The presentation in the 12 additionally visualizes a route planned by the automatic parking assistance system 120 from a current vehicle position into the parking space.

Another form of visualization in the form of a three-dimensional representation is in 13 given. Such a representation is preferably selected as the vehicle approaches and enters the selected parking space during the automatic parking process. It becomes a representation of a virtual viewpoint above the own vehicle 20 chosen, and at the same time the planned route 120 for parking, in the present case backwards.

To visualize the target parking space 24b becomes a three-dimensional "wall" corresponding to the lateral and rear boundaries ( C1C4 . C4C3 and C2C3 in 6 ). The height of the "wall" is configurable. It can, for example, be chosen so that it provides a significant representation of the vehicle environment 1m equivalent.

To visualize obstacles, a three-dimensional wall is also shown around them, following the determined boundaries.

With the chosen representation, the driver can track how the vehicle 20 in automatic parking mode on the parking space 24b to and enters this.

In preferred embodiments, for example, the three-dimensional "walls" in brown, the own vehicle and the planned route in gray and the destination position can be displayed in green. Again, of course, other suitable colors are equally possible.

Automatic parking systems according to embodiments of the present invention are further adapted to visualize different stages of a parking operation. In particular, such stages may be: "search," "parking spot found," "automatic parking in progress," and "parking completed."

A corresponding representation can be made both in the context of a two-dimensional visualization as in 11 as well as in the context of a three-dimensional visualization such as in 13 respectively. Other suitable representations are possible within the scope of the present invention. In addition, an information of the driver in particular on the completion of the parking operation additionally or alternatively as acoustic information, or via a vibration signal.

In the following, the flow of a method for automatically parking a motor vehicle according to an embodiment of the present invention will be described with reference to the representation in the flowchart of FIG 14 described.

The procedure begins with a search trip (step S10 ). Here, the vehicle for which a parking space is searched (hereinafter for the sake of simplicity also "own vehicle") with or without driver (the latter in the case of autonomous driving at level 4 or 5 ) over the parking lot. During the search run, the vehicle environment data is detected by the at least one sensor of the motor vehicle. This is done continuously or at predetermined intervals. All available information is collected, collected and processed as well as visualized as described above, such as in a two-dimensional representation according to FIG 11 , As a process stage "Search" is visualized during this phase.

Below (step S20 ) automatically detects parking spaces by the automatic parking system. This can be done according to embodiments of the invention either on both sides of the own vehicle, or only on one side of the vehicle, which is predetermined by a turn signal indicator.

Once a candidate for a parking space has been found and identified as "valid" or found, the driver is in step S30 notified accordingly. This is preferably done both via a corresponding visualization, as well as additionally via an acoustic message or voice message. The notification preferably also contains information about the vehicle side (right or left) on which the parking space was found.

In the subsequent step S40 The driver stops the vehicle to park in reverse. According to preferred embodiments of the invention, the automatic parking is done backwards both when parking transverse to the direction of travel and in the direction of travel. By way of exception, it can be packed forward during inclined parking.

In the following step S50 the valid parking spaces are visualized. This is preferably done as above with reference to the illustrations of 11 to 13 has been described.

Subsequently, in the step S60 the selection of a parking space from several candidates. This is done either automatically by the automatic parking system or manually by the driver (configurable in the system). Alternatively, it is also possible to first issue a request to the driver to determine whether he wants to manually select the parking space or whether the system should automatically select a parking space. When the parking space is selected, this is visualized, for Example of a color change, from yellow to green as related to 12 described.

In the subsequent step S70 the planning of the travel path takes place so that the own vehicle reaches the target parking space. This is visualized in the form of a planned trajectory, for example in gray (see 12 and 13 ). In addition, preferably also a current setting of the steering, or the corresponding direction of travel) with a different color, such as green, visualized.

In the exemplary embodiment of the 14 takes place in the following optional step S80 a repeated request for a confirmation from the driver that the automatic parking process should now be executed. The driver now again has the opportunity to either confirm the automatic parking procedure or to cancel the procedure (to park manually or to discard the selected parking space altogether). However, the last step is optional, and it can be packed automatically without asking the driver again. Of course, this is necessary especially when there is no driver in the car when driving autonomously.

Alternatively (or additionally), a corresponding query can also be made immediately after the step S60 take place (not shown in the picture), so it can be queried whether a particular automatically selected parking space should either be accepted or discarded. Preferably, this query and / or the query takes place after step S80 in that a button of a touch screen is displayed and prompts for actuation by the driver. Alternatively, query and confirmation can of course be made acoustically and by voice control. Other possibilities are also known to the person skilled in the art.

In the following step S90 the automatic parking of the vehicle takes place in the target parking space. Here, the visualized stage changes to "automatic parking is running". If necessary, the gearshift is automatically actuated by the system.

In the final step S100 a renewed change of the visualized stage of the parking operation after "parking process completed" takes place as soon as the final position of the vehicle in the target parking space is reached. In addition, an acoustic or voice message can inform the driver about the successfully completed automatic parking.

In summary, the present invention relates to an automatic parking system and a corresponding automatic method for parking a motor vehicle, in which possible candidates for free parking spaces are automatically determined and selected. This is done on the basis of a population state map in which sensor data of one or more vehicle sensors are fused. The data are prepared accordingly and visualized for a driver so that he can follow the process at any time and if necessary intervene. For a selected parking space, a route is planned and, optionally after confirmation by the driver, the automatic parking process is carried out.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been 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.

Cited non-patent literature

• "Dynamic Map Building for an Autonomous Mobile Robot" by John J. Leonard, Hugh F. Durrant-Whyte [0004]
• Ingemar J. Cox, The International Journal of Robotics Research, Vol. 4, August 1992, pp. 286-298 [0004]
• "Sensor Fusion for Flexible Human-Portable Building Scale Mapping" by Maurice F Fallon, Hordur Johansson, Jonathan Brookshire, Seth Teller [0004]
• John J. Leonard, Proceedings oft he 2012 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4405-4412 [0004]

Claims (15)

An automatic parking system for a motor vehicle (20), comprising a memory device for storing a population state map (12) of a vehicle environment of the motor vehicle (20), the population state map (12) comprising a two-dimensional grid of cells, each associated with a parameter value representing one of the possible population states "free", "busy" and "unknown", a calculation device (14) configured to determine free areas of the vehicle environment on the basis of the occupation state map (12) as candidates (24) for parking spaces, a visualization device configured to visualize the vehicle environment, including the determined candidates (24) for parking spaces and for visualizing a parking space (24b) selected from the determined candidates (24), a planning device (14) configured to determine a planned travel path (120) from a current vehicle position into the selected parking space (24b), taking into account given vehicle parameters and a control device (16) configured to automatically control the motor vehicle (20) in the selected parking space (24b) along the planned travel path (120). Automatic parking system after Claim 1 wherein the occupancy status map (12) is the result of a sensor fusion of data determined by at least one vehicle sensor (4, 6, 8, 10) of the motor vehicle (20). Automatic parking system after Claim 2 in which the occupancy status card (12) contains occupancy information at least with respect to cells in the region of a front line (26) of objects (22) located during the data acquisition by the sensors (4, 6, 8, 10) along one side of the motor vehicle (20) and Calculating means (14) is arranged to determine a free area as a candidate (24) for a parking space by a front line (26) of an object (22) as a designated by a plurality of occupied cells along the side of the motor vehicle (20 ) defines an orientation of the object (22) according to a predetermined algorithm based on the front line (26), a central area (29) of a free area as a candidate (24) for a parking space based on the thus determined orientation of an object (22) or two adjacent objects (22) defines and a front line of the free area perpendicular to the fixed central axis (29) so f determines that it is substantially aligned with the front line (26) of the one or two adjacent objects (22). Automatic parking system after Claim 3 wherein the predetermined algorithm determines the orientation of the object (22) as a normal (28) to the front line (26). Automatic parking system after Claim 3 in that the occupancy status card (12) additionally includes occupancy information regarding cells in the range of side edges (62) of objects (22) located during the data acquisition by the sensors (4, 6, 8, 10) along one side of the motor vehicle (20) and the predetermined one Algorithm determines a sideline (62) of an object (22) as a cell characterized by a plurality of busy in a direction substantially perpendicular to the front line (26) and orienting the object (22) as the orientation of the sideline (62 ). Automatic parking system after Claim 5 wherein said calculating means (14) additionally uses information about lane markings (90) of parking spaces taken by a vehicle sensor (4, 6, 8, 10) for defining the center axis (26) of a free area as a candidate (24) for a parking space, if these road markings (90) are not covered completely or partly by objects (22) located along one side of the motor vehicle (20) during data acquisition by the sensors (4, 6, 8, 10). Automatic parking system after one of the Claims 2 to 6 , wherein the data recording by the vehicle sensors (4, 6, 8, 10) takes place during the journey, in particular while driving through a parking lot. Automatic parking system after one of the Claims 1 to 7 wherein the selection of the parking space from the determined candidates (24) takes place automatically. Automatic parking system after one of the Claims 1 to 8th , wherein the visualization device is set up to additionally visualize the planned route (120). A method for automatically parking a motor vehicle with an automatic parking system, comprising storage means for storing a population status map (12) of a vehicle environment of the motor vehicle (20), the occupancy state map (12) comprising a two-dimensional grid of cells, each associated with a parameter value which accepts one of the possible occupation states "free", "busy" and "unknown", and wherein the method comprises the steps of automatically determining (S20) free areas of the vehicle environment on the basis of Occupancy state map (12) as parking space candidate (24), visualizing (S50) the vehicle environment including the determined candidates (24) for parking spaces, selecting (S60) a parking space from the determined candidates (24) and visualizing the selected parking space (24b), Scheduling (S70) a travel path (120) from a current vehicle position into the selected parking space (24b) taking into account given vehicle parameters and automatically controlling (S90) the motor vehicle into the selected parking space (24b) along the planned travel path (120). Method according to Claim 10 , wherein furthermore a respective run through process section is visualized. Method according to Claim 10 or 11 further comprising the step (S10) of detecting vehicle surroundings data by at least one vehicle sensor (4, 6, 8, 10) of the motor vehicle (20), the occupancy state map (12) detecting the result of sensor fusion from the at least one vehicle sensor (4, 6, 8, 10) is determined. Method according to one of Claims 10 to 12 wherein the at least one vehicle sensor comprises an ultrasonic sensor (6), a radar device (10), a camera (4) or a lidar sensor (8). Method according to one of Claims 10 to 13 , further comprising the step of issuing a message to a driver of the motor vehicle (20) requesting to accept a parking space (24b) selected in the step (S60) of selecting a parking space and the step of accepting the selected parking space (24b) by a user input of the driver. Method according to one of Claims 10 to 14 , further comprising the step (S100) of informing the driver of the successful completion of the automatic parking operation by outputting a corresponding message upon completion of the step (S90) of automatically controlling the motor vehicle (20) in the selected parking space (24b).

Images