DE102016003261A1 - Method for self-localization of a vehicle in a vehicle environment - Google Patents

Abstract

The invention relates to a method for self-localization (SL) of a vehicle in a vehicle environment, in particular in a multi-storey car park, wherein
- Environment data of the vehicle environment are recorded and
An actual position of the vehicle in the vehicle environment is determined,
characterized in that a digital three-dimensional route map (WK) is created and stored during at least one learning trip, wherein
For generating the three-dimensional route map (WK), a three-dimensional coordinate system is defined which comprises position coordinates of the vehicle in a driving plane of the vehicle environment and a path length of a route traveled in the vehicle environment,
- A temporal position profile of the vehicle is determined by means of path length measurement, and
- The path length of the distance traveled is determined based on the determined position history.

Description

The invention relates to a method for self-localization of a vehicle in a vehicle environment according to the preamble of claim 1.

From the prior art is, as exemplified in DE 10 2013 018 721 described a method for detecting at least one parking space for a motor vehicle. In this case, at least a portion of the environment of the motor vehicle is detected by at least one distance-measuring sensor of the motor vehicle by at least one primary signal emitted by the sensor and at least one of a reflection of the primary signal at least one of the cars different object resulting echo signal is detected. The sensor data characterizing the echo signal and provided, for example, by the distance-measuring sensor are registered and accumulated in cells of a digital occupancy grid which characterizes the partial area of the environment as a function of the position of the motor vehicle relative to the detected partial area.

Furthermore, from the DE 10 2013 019 804 A1 a method for detecting a movement of an object, in particular in an environment of a vehicle, known from radar data. In this case, the object is detected simultaneously with at least two radar sensors in an angular range, wherein based on at least three object points arranged at different positions on the object, assuming a rigid object, motion information of the object is determined.

In addition, from the DE 10 2013 003 117 A1 a method for self-localization of a vehicle and for detection of objects in an environment of the vehicle known. Here, image features and comparison features described by feature signatures pixel-based areas in the images are used.

The invention is based on the object to provide a comparison with the prior art improved method for self-localization of a vehicle in a vehicle environment, in particular in a multi-storey car park. The object is achieved with the features specified in claim 1.

Advantageous embodiments of the invention are the subject of the dependent claims.

In a method for self-localization of a vehicle in a vehicle environment, in particular in a multi-storey car park, environmental data of the vehicle environment are detected and an actual position of the vehicle in the vehicle environment is determined.

According to the invention, it is provided that a digital route map is created and stored during at least one learning run, wherein a three-dimensional coordinate system is defined for generating the route map, which includes position coordinates of the vehicle in a driving plane of the vehicle environment and a path length covered in the vehicle environment distance. In this case, a temporal position profile of the vehicle is determined, wherein the path length of the distance traveled is determined on the basis of the determined position profile.

The inventive method allows the creation of a three-dimensional route map for a multi-storey car park by means of any distance-measuring sensors. Thus, for example, two-dimensional measuring sensors or possibly also the use of different sensors are possible, which can be dispensed with a costly 3D sensor. Furthermore, the method allows the extension of a scope of a parking assistant, as well as multi-level car parks can be automated drive. In addition, the method allows a highly accurate localization without navigation satellite systems, such. B. Global Positioning System (short: GPS) and Differential Global Positioning System (DGPS), which are usually limited in parking garages available.

The three-dimensional route map can be used advantageously for the parking garages, which are used regularly. For this purpose, the learning journey to create the three-dimensional route map is preferably carried out by the vehicle which is to use the relevant parking garage later.

Embodiments of the invention are explained in more detail below with reference to drawings. Showing:

1 1 is a schematic diagram of a method for self-localization of a vehicle in a vehicle environment;

2 schematically a block diagram of a method step for creating a digital map of a vehicle environment and

3 schematically a block diagram of a method step for self-localization of the vehicle in accordance with 2 created digital map.

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

1 shows a simplified flow diagram of a method for self-localization SL (see 2 and 3 ) of a vehicle not shown here in a vehicle environment.

The method serves to localize the vehicle in a parking garage, in particular in a multi-storey car park, along a three-dimensional route map WK (see 2 ) by means of distance-measuring sensors. This enables the extension of a parking assistant integrated in the vehicle for the highly automated start-up of frequently used parking spaces. The vehicle environment thus represents here a multi-storey car park.

The method can generally be subdivided into three method steps S1 to S3. In an initial method step S1, the vehicle carries out at least one learning journey in the vehicle environment in which the vehicle surroundings are detected and a route to be traveled is learned. When capturing the vehicle environment, environmental data is captured and stored in an angular range of 360 °. For this purpose, a plurality of sensors, in particular distance-measuring sensors, such as. As radar sensors or Lidarsensoren arranged, the environmental data capture and store.

Furthermore, during the learning run, an X position of the vehicle, a Y position of the vehicle and a path length of a traveled distance are detected or determined along a trajectory, wherein these coordinates span a three-dimensional coordinate system. By means of the three-dimensional coordinate system, a digital, three-dimensional path map WK can subsequently be generated for the detected vehicle environment.

The detection or determination of the X position and Y position of the vehicle can be determined, for example, by means of a coupled navigation of a speed and yaw rate measurement of an electronic stability program (ESP for short). Alternatively or additionally, the X position and Y position of the vehicle can be determined via a radar odometry, as exemplified in the aforementioned DE 10 2013 019 804 A1 is described.

T

= Trajektorie;

X

= X-Position des Fahrzeugs;

Y

= Y-Position des Fahrzeugs und

N

= Anzahl der Zeitschritte.

The determination of the path length takes place via a mathematical determination of the traveled distance, wherein a two-dimensional trajectory (= traveled distance) is determined by storing the current vehicle position for each time step as follows: T = [(X ₁ , Y ₁ ) ^T , ..., (X _N , Y _N ) ^T ] (1) With:

T

= Trajectory;

X

= X position of the vehicle;

Y

= Y position of the vehicle and

N

= Number of time steps.

mit:

TD

= Weglänge

t

= Zeitschritt.

Based on the determined trajectory, the path length of the trajectory can be determined as follows:

With:

TD

= Path length

t

= Time step.

The term ∥T _t - T _t-1 ∥ determines the so-called L ² -norm (Hilbert space) in two consecutive time steps within the trajectory. By determining the path length ambiguities of two-dimensional trajectories, especially in multi-storey car parks, at least reduced, preferably eliminated.

By means of these three determined coordinates, the aforementioned three-dimensional coordinate system can be generated, based on which a three-dimensional path map WK is created for the vehicle environment. The creation of this map WK will be in 2 described in more detail.

In a further method step S2, a self-localization SL of the vehicle in the created three-dimensional route map WK is possible, wherein the learned trajectory can be traveled. Radar sensors, ultrasonic sensors, cameras or laser scanners can be used for self-localization SL.

In a last method step S3, the environment data are updated after each successful trajectory search, so that the method can be adapted to daily and seasonal fluctuations.

2 shows a block diagram for creating the three-dimensional map WK in the context of the above-described initial method step S1.

The block diagram comprises an input field E, in which sensor measurement data SM and proper movement data ED are entered for creating the three-dimensional route map WK. The eigenmotion data ED can for example be provided by an ESP or an IMU (short for intertial measurement unit).

In addition or as an alternative to the sensor measurement data SM, so-called medium-level representations MLR are provided, which include, for example, landmarks, cluster analyzes or so-called NDT data.

Landmarks, for example, a local allocation grid is created, which is a virtual map of the detected vehicle environment. This occupancy grid is also referred to as an "occupancy grid" and comprises a plurality of individual cells. The sensor measurement data SM are entered into these cells as a function of a position of the vehicle and accumulated in the cells. Subsequently, distinctive point targets in the allocation grid are extracted.

The extraction takes place, for example, by means of grouping of cells in which at least part of an object or objects located in the detected vehicle environment is / are characterized by the registered sensor measurement data SM. The groupings are extracted from the allocation grid and analyzed in detail by determining scattering centers on the basis of the sensor data. These scattering centers represent places or areas of the actual vehicle environment in which emitted sensor signals were particularly strongly reflected. This exemplary procedure is from the aforementioned DE 10 2013 018 721 A1 known.

The distinctive point targets extracted in this way are then placed around the circles or circular regions and statistics are extracted from the assignment grid along several circular regions and stored in a descriptor. A similarity of second landmarks is determined by a descriptor comparison.

On the basis of the sensor measurement data SM and the medium-level representations MLR, a mapping KA of the vehicle environment can take place in the context of the initial method step S1, from which the three-dimensional route map WK is created together with a self-localization SL of the vehicle, the eigenmotion data ED being included here , The three-dimensional route map WK is arranged here in an output field A.

3 shows a flowchart for self-localization SL of the vehicle.

In this case, measurement vectors m _{j of} the sensor measurement data SM and of the medium-level representations MLR are provided as input data. By means of a measure of similarity AM along the trajectory of the three-dimensional route map WK, a coarse positioning takes place within the three-dimensional route map WK. The coarse positioning describes a section of the trajectory.

In the coarse positioning area, a vehicle _pose P Fzg is then determined by means of a RANSAC (English: RANdom SAmple Consensus) algorithm RAVS. B. distance and angle of the vehicle, the three-dimensional map WK describes.

As input data of the RANSAC process step RAVS, three arbitrary measurement vectors m _{j are} defined and from this the vehicle _pose P _{Fzg is determined} with respect to the coarse positioning region of the three-dimensional route map WK. Subsequently, a distance A _{j of} all predetermined measurement vectors m _{j is determined} for this vehicle _pose P _Fzg and, finally, a weighting GW of the results is carried out. The vehicle _pose P _Fzg can also be smoothed with a temporal filter. The result of the flowchart shown is the self- _localization SL of the vehicle with the parameters Vehicle P P _Fzg and _Weglänge .

The vehicle _pose P _Fzg determined in the flowchart can be used to automatically regulate a vehicle along the determined trajectory. It is advantageous here that due to the position determination within the three-dimensional route map WK automated trips in complex car parks with multiple levels are possible.

LIST OF REFERENCE NUMBERS

• A

  output field

  A _j

  distance

  AT THE

  similarity

  e

  input box

  ED

  Proper motion data

  GW

  weighting

  KA

  mapping

  MLR

  Medium-level representations

  P _vehicle

  Fahrzeugpose

  Rav

  RANSAC algorithm

  S1 to S3

  step

  SL

  self-localization

  SM

  Sensor measurement data

  WK

  three-dimensional route map

  m _j

  measurement vector

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 patent literature

• DE 102013018721 [0002]
• DE 102013019804 A1 [0003, 0020]
• DE 102013003117 A1 [0004]
• DE 102013018721 A1 [0031]

Claims (4)

Method for self-localization (SL) of a vehicle in a vehicle environment, in particular in a multi-storey car park, wherein - environmental data of the vehicle environment are detected and - an actual position of the vehicle in the vehicle environment is determined, characterized in that a digital three-dimensional path map (WK) is generated and stored during at least one learning run of the vehicle, wherein - for the creation of the three-dimensional route map (WK) a three-dimensional coordinate system is defined, which includes position coordinates of the vehicle in a driving plane of the vehicle environment and a path length traveled in the vehicle environment distance, - a temporal Position course of the vehicle is determined and - the path length of the distance traveled is determined based on the determined position profile. Method according to Claim 1, characterized in that current position coordinates of the vehicle in the driving plane are detected in each case at a defined time step in order to determine the temporal position profile of the vehicle. A method according to claim 1 or 2, characterized in that the vehicle environment is detected in an angular range of 360 °. Method according to one of the preceding claims, characterized in that the environment data are updated each time the vehicle in the vehicle environment.

Images