DE102021004068A1 - Method for operating an assistance system for speed-dependent distance control of a vehicle - Google Patents

Abstract

The invention relates to a method for operating an assistance system (2) for speed-dependent distance control of a vehicle (1). According to the invention, it is provided that - a respective object detected on the vehicle side is assigned to an object class and a position in an image plane of detected optical signals, with at least vehicles (3) traveling ahead being classified as objects, - based on image information and the vehicle (3) traveling ahead. assigned position a sequence of the vehicles ahead is extracted, - output light signals (S) from direction indicators (3.1), brake lights (3.2) and / or other lighting units of the vehicles ahead (3) extracted from the optical signals by means of a filter and a corresponding preceding vehicle (3) are assigned as light signal information, at least one spatial and temporal light signal information of the preceding vehicles (3) as a model-side egg to at least one based on a model (M) trained by means of a machine learning method for backwater forecasting Input information for predicting a point in time of a manually or automatically initiated braking intervention is supplied and - by means of the assistance system (2), depending on the predicted point in time of the manually or automatically initiated braking intervention of the vehicle in front (3), a deceleration of the vehicle (1) is automatically initiated.

Description

The invention relates to a method for operating an assistance system for speed-dependent distance control of a vehicle with an environment sensor system, by means of which signals are continuously recorded, with the aid of which an environment and objects located in it are detected.

From the DE 10 2007 049 516 A1 a system for the multimodal determination of objects in a field of view in front of and / or behind a vehicle is known. A first environment detection is carried out by means of a first sensor, the sensor output signal is fed to a first sensor signal analysis method with object detection and a further environment detection is carried out by means of a second sensor and the sensor output signal of the second sensor is fed to a second sensor signal analysis method with object detection. It is also provided that certain objects detected by the sensors and determined by the sensor signal analysis method are fed to an object confirmation and situation analysis module for verification, further vehicle information and the lane prediction being taken into account for verification. The object confirmation and situation analysis module determines relevant objects, with the relevant objects and the driver information initiating measures to increase passive safety when a dangerous situation is detected by the hazard computer.

The invention is based on the object of specifying a method for operating an assistance system for speed-dependent distance control of a vehicle.

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

Advantageous refinements of the invention are the subject matter of the subclaims.

In a method for operating an assistance system for speed-dependent distance control of a vehicle with an environment sensor system, optical signals are continuously recorded by means of this system, with the aid of which an environment and objects located in it are detected. According to the invention, it is provided that a detected respective object is assigned to an object class and a position in an image plane of the detected optical signals, with at least vehicles traveling ahead being classified as objects. Based on image information and the position assigned to the respective preceding vehicle, a sequence of the preceding vehicles is extracted and output light signals from direction indicators, brake lights and / or other lighting units, such as. B. a reversing light, a rear fog light, etc., the vehicles ahead are extracted from the optical signals by means of at least one filter and assigned to a corresponding vehicle ahead as light signal information. Furthermore, the method provides that at least one local-temporal light signal information of the vehicles driving ahead is fed to a model trained by means of a machine learning method for backwater wave prediction as model-side input information for predicting a point in time of a manually or automatically initiated braking intervention and by means of the assistance system in Depending on the predicted point in time of the manually or automatically initiated braking intervention of the vehicle in front, a deceleration of the vehicle is initiated automatically.

By using the method, road traffic safety can be significantly increased by expanding an assistance function of the assistance system through additionally estimated information relating to the dynamics of several vehicles driving ahead, thus ensuring a largely permanently stable distance control.

The method can be used to reduce the energy required for driving, in particular fuel and / or electrical energy, by means of an optimized automatic distance control system, since a braking process can be largely prevented by a relatively early reduction in vehicle acceleration.

The method can thus increase driving comfort for occupants, since a number of comparatively strong braking maneuvers or, if necessary, emergency braking maneuvers can be reduced due to the optimized automatic distance control system.

In addition, by using the method, a driving experience and the well-being of vehicle occupants can be increased through preventive, safety-increasing information processing, in particular with digital means.

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

• 1 schematisch ein Fahrzeug mit einer Vorrichtung zur Durchführung eines Verfahrens zum Betrieb eines Assistenzsystems zur geschwindigkeitsabhängigen Abstandsregelung,
• 2 schematisch das einer Reihe vorausfahrender Fahrzeuge folgende Fahrzeug und
• 3 schematisch anhand von Abbildungen einen Ablauf des Verfahrens.

Show:

• 1 schematically a vehicle with a device for carrying out a method for operating an assistance system for speed-dependent distance control,
• 2 schematically the vehicle following a number of vehicles in front and
• 3 schematically a sequence of the process based on images.

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

1 shows a vehicle 1 with a device for performing a method for operating an assistance system 2 for speed-dependent distance control. In 2 the vehicle follows 1 a number of vehicles in front 3 and in 3 a sequence of the procedure is shown on the basis of 3 figures A1 to A3.

The vehicle 1 comprises the device, which has an environment sensor system with at least one in front of the vehicle 1 facing camera 4th , a control unit 5 and the assistance system 2 having. In addition, the vehicle is 1 in particular via a radio connection with a data center in terms of data technology 6th connected with the other vehicles 7th , for example a vehicle fleet, are also connected in terms of data technology. In the vehicle 1 there is also a driver 8th .

Generally is a function of the assistance system 2 for speed-dependent distance control of the vehicle 1 in relation to a vehicle in front 3 known.

Here the driver provides 8th of the vehicle 1 a target speed at which the vehicle 1 , if available, a vehicle in front 3 follows at a given distance. This forms the distance between the vehicle 1 to the vehicle in front 3 a controlled variable of the assistance system 2 .

Decelerates the vehicle in front 3 or a driving speed of the vehicle 1 is higher than that of the vehicle in front 3 , the distance between the vehicle is reduced 1 to the vehicle in front 3 . If the distance then falls below a specified safety distance, the assistance system 2 a braking intervention is initiated automatically, so that the distance between the vehicles 1 , 3 enlarged again.

Relatively abrupt braking interventions by the assistance system 2 because of a reduction in the distance to the vehicle in front 2 To prevent as far as possible, whereby driving comfort and safety in road traffic can be increased, the method described below is provided.

While the vehicle is in motion 1 are by means of the camera 4th the environmental sensor system continuously captures optical signals, in particular image data, which the control unit 5 for evaluation and further processing. Based on the evaluation, the camera 4th detected objects recognized and localized.

In addition, the control unit 5 manual braking intervention by the driver 8th as well as automatic braking interventions by the assistance system 2 communicated. In addition, the control unit receives and exchanges 5 further information with the assistance system 2 , in particular with its further control unit, not shown in detail.

In the control unit 5 objects detected by cameras are recognized and localized using methods of computer-controlled vision, for example by using deeply folded neural networks, ie assigned to an object class and a position, in particular with regard to center point and size, in an image plane. Computer-aided vision enables z. B. by means of semantic segmentation, continue to assign each pixel of the image plane to an object category.

The camera 4th the environment sensor system is designed, for example, as a mono camera, stereo camera and / or as a so-called time-of-flight camera, wherein the environment sensor system can alternatively or additionally include at least one lidar-based and / or radar-based sensor to determine a class, a position based on captured optical signals , a size and a shape of objects in front of the vehicle 1 to investigate.

According to the in 2 shown embodiment are as objects of the vehicle 1 vehicles in front 3 recognized based on the captured optical signals and classified accordingly.

Based on the image data, the position and a pixel-perfect shape of the vehicles in front 3 it is possible to place a local order in relation to the vehicles in front 3 to extract. For this purpose, a geometric mapping property of detected objects, that is to say detected vehicles traveling ahead, can be used 3 , depending on the respective distance to the camera 4th be used. That is to say, a perspective illustration of vehicles driving ahead at a distance 3 takes place on smaller image areas in the image plane. Thus, the order of the vehicles in front 3 can be estimated on the basis of their relative sizes, as in 2 and in the first figure A1 in 3 is shown.

As an alternative or in addition, the detected shapes of the vehicles traveling ahead can be used 3 Information about the order of these will be extracted. For example, a vehicle driving further ahead can 3 from a subsequent vehicle in front 3 be partially covered, whereby the order can be recognized. By comparing object shapes, in particular vehicle shapes, it can be functionally determined which object is covering which, in particular if not only individual images but image sequences are evaluated in a recorded video and thus a large number of occlusions are present at different times.

In a further alternative or additional embodiment, depth information can be used to establish a sequence of the objects, ie the vehicles driving ahead 3 to extract. Individual recordings from mono cameras can be made at different times, taking into account a relative movement of the camera 4th can be used between individual recordings, stereo cameras, time-of-flight cameras, lidar-based and / or radar-based sensors.

On the basis of recognized vehicle areas in an image plane of the recorded image data, perspective size relationships, overlap lines and / or depth information, the sequence of the vehicles ahead is determined 3 determined. Light signals S are detected and the respective vehicle 3 assigned.

Light signals S, such as from a direction indicator 3.1 and / or brake lights 3.2 vehicles in front 3 are made from image data of a recorded in 2 and the first figure A1 in 3 Image B shown, extracted and the individual vehicles in front 3 assigned. For this purpose, machine learning methods can be used to generate direction indicators 3.1 and brake lights 3.2 to recognize and localize in the image data, for example by means of deeply folded neural networks, semantic segmentation, etc.

Based on this information, for example, classifications, e.g. B. on the basis of deeply folded neural networks, and / or color filters can be applied to these image areas in order to activate direction indicators 3.1 of deactivated direction indicators 3.1 to distinguish as well as a state of brake lights 3.2 vehicles in front 3 to recognize.

1. a) von Fahrzeugnummem in der Reihenfolge,
2. b) frühere diskredierte Zeitpunkte sowie
3. c) binär kodierten Lichtsignalen S, insbesondere in Bezug auf einen Zustand der

The information can be in a binary 3-dimensional field of dimensions

1. a) of vehicle numbers in the order,
2. b) earlier disputed times as well as
3. c) binary coded light signals S, in particular with regard to a state of the

brake-lights 3.2 , are aggregated, as shown in a second figure A2 in 3 is shown.

Alternatively, light signals S and combinations of light signals S can be coded together so that the information can be represented in a spatio-temporal 2-dimensional matrix.

Using machine learning methods, e.g. B. by means of neural networks, decision trees, etc. is in the control unit 5 trains a model M for backwater wave prognosis, which a spatial-temporal light signal information, z. B. in field or matrix form, the vehicles in front 3 and, if necessary, further information, for example a current driving speed of the vehicle 1 , a distance to a vehicle in front 3 , a time of day, a geolocation, etc. are used as input features and a point in time of a manual or automatic braking intervention that has been carried out is used as the target variable.

If a prediction accuracy of the model M exceeds a predefined threshold value, the model M can be used to predict a braking intervention and thus a regulation of the assistance system 2 can be optimized for speed-dependent distance control.

Since the assistance system is informed that it will carry out a braking intervention at a specific future point in time, this information can be stored in the assistance system 2 be used to prepare the braking intervention and / or already through z. B. Reducing a vehicle acceleration a relative distance to a vehicle immediately ahead 3 to anticipate. That is to say, by means of an assistance system designed in this way 2 Braking can be initiated in a predictive manner or, in general, predictive control can be implemented, as shown in a third figure A3 in 3 is shown.

The model M can be trained with training data and / or with data from other vehicles, in particular during development 7th , ie a vehicle fleet, trained or retrained. If the model M is based on data from other vehicles 7th trained or retrained, training data with the data center 6th communicated and the trained or retrained model M is the vehicle 1 and the other vehicles 7th the vehicle fleet from the data center 6th to the Made available, for example distributed. Distributed learning methods can be used to further train the model M in the vehicle fleet.

In one possible embodiment of the method, an individual driver braking behavior in the event of backwater waves can be learned using the model M and imitated by the assistance system. For this purpose, the M model is only provided with examples of manual braking interventions due to vehicles driving ahead 3 a clearly identified driver 8th trained further. Thus the assistance system learns 2 an individual braking behavior of the driver 8th and imitates this. Since the assistance system 2 then behaves as the driver does 8th behavior, may result in greater acceptance by the driver 8th for the assistance system 2 can be achieved.

Another possible embodiment provides that an analytical functional approximation of the backwater wave prognosis is derived from the spatial and temporal light signal information. However, probability distributions lead to the estimated relative positions and relative speeds of the vehicles in front 3 to probabilistic coupled differential equations whose solutions can also only be approximate.

In the first figure A1 the 3 is that by means of the camera 4th recorded image B, in the second figure A2 is the binary field over time t and in the third figure A3 is the model M for predicting the times of the braking intervention by the driver 8th or the assistance system 2 shown. The sequence of the method described above is shown on the basis of Figures A1 to A3.

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed 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

• DE 102007049516 A1 [0002]

Claims (1)

A method for operating an assistance system (2) for speed-dependent distance control of a vehicle (1) with an environment sensor system, by means of which optical signals are continuously recorded, based on which an environment and objects located in it are detected, characterized in that - a detected respective object an object class and a position in an image plane of the detected optical signals, with at least vehicles (3) traveling ahead being classified as objects, - based on image information and the position assigned to the respective vehicle traveling ahead (3), a sequence of vehicles traveling ahead (3) - Light signals (S) output by direction indicators (3.1), brake lights (3.2) and / or further lighting units of the vehicles ahead (3) are extracted from the optical signals by means of at least one filter and a corresponding vehicle ahead (3 ) are assigned as light signal information, - at least one local-temporal light signal information of the vehicles (3) ahead as model-side input information for predicting a point in time of a manually or automatically initiated brake intervention is fed to a model (M) trained by means of a machine learning method for backwater wave forecasting and - a deceleration of the vehicle (1) is automatically initiated by means of the assistance system (2) depending on the predicted time of the manually or automatically initiated braking intervention of the vehicle (3) in front.

Images