WO2020043403A1

WO2020043403A1 - Driver assistance system

Info

Publication number: WO2020043403A1
Application number: PCT/EP2019/070163
Authority: WO
Inventors: Anja Petrich; Julian Fieres; Florian Ade; Lucas Hanson; Debora Lovison; Andrea Loi
Original assignee: Zf Friedrichshafen Ag
Priority date: 2018-08-27
Filing date: 2019-07-26
Publication date: 2020-03-05
Also published as: DE102018214420A1

Abstract

A method for detecting travel noise of rail vehicles comprising the following steps: detecting (St1) travel noise of a rail vehicle by means of a microphone which can be operated on a road vehicle; determining (St2) a response to the detected travel noise by means of a computing unit which can be operated on a road vehicle.

Description

Driver assistance system

FIELD OF THE INVENTION

The present invention relates to a method for detecting driving noises from rail vehicles and a training system for a vehicle control system for detecting driving noises from rail vehicles.

TECHNICAL BACKGROUND

DE102013004271 A1 shows a driver assistance system by means of a whistle tone detection for the detection of rail vehicles.

SUMMARY OF THE INVENTION

Against this background, the invention is based on the object of specifying a more reliable driver assistance system for warning of rail vehicles.

According to the invention, this object is achieved by a method for detecting travel noises of rail vehicles with the features of claim 1 and / or by a training system for a vehicle control for detecting travel noises by rail vehicles with the features of claim 9.

Accordingly, it is provided:

- A method for detecting travel noise of rail vehicles with the following steps: detecting a travel noise of a rail vehicle using a microphone that can be operated on a road vehicle; Determining a reaction to the detected driving noise by means of a computing unit which can be operated on a road vehicle; such as a training system for a vehicle control for detecting travel noises of rail vehicles, with at least one interface in order to receive training data, each of which has an audio signal and a target reaction signal, an evaluation unit which forms and sets up an artificial neural network is to feed the artificial neural network forward with training data in order to determine actual reaction signals, and to determine a changed topology of the artificial neural network, in particular weights, by feeding back the desired reaction signals in the artificial neural network, the topology Tet is to be stored in the vehicle controller for detecting travel noise from rail vehicles.

Motor vehicles in the sense of this patent application are motor-driven land vehicles. Rail vehicles are also included.

A road vehicle is a land vehicle that is primarily intended for use on roads. It is characteristic of all road vehicles that they can maintain or change their direction of travel on the ordered surface with the help of static friction and do not require any tracks for this. For example, motor vehicles are road vehicles.

Rail vehicles are vehicles of trains that run or are guided on one or more rails. Vehicle and rail are a coordinated system, usually referred to as a wheel-rail system. For example, trains are rail vehicles.

Driving noises arise due to the driving of a vehicle. These include e.g. Tire-road noise or wheel-rail noise, drive noise and aerodynamic noise. Wheel-rail noises are contact noises due to the wheel coming into contact with a rail.

Rolling noises and cornering noises are wheel-rail noises from rail vehicles. Rolling noises are generated when the wheels roll on the rail track. The material structure, wear and corrosion of certain surface and structural irregularities of the rail and wheel surfaces cause vertical excitations of the wheel sets and rails when they roll, which stimulate them to vibrate, which are emitted as sound into the surrounding air.

Curve noises such as curve squeaking occur when driving on narrow curves and switches, if the wheels cannot roll without slippage due to the radius difference between the two rails. In very narrow track arches, there is also a flange start.

Rail joints (area of abutting rail ends), heart pieces (elements in switch construction), flat spots (signs of wear on wheels of rail vehicles) as well as deposits (material that has been sanded down and often applied in front of a flat spot or also elsewhere) result in extensive impulsive noises that can be assigned to the wheel-rail system.

In the case of modern traction vehicles, there are also noises caused by the slight slip when accelerating and (electric) braking on slippery rails.

The contact between the wheel and the rail causes the tread lugs and the carcass to vibrate and emit airborne noise. In addition, air in the tire flap is displaced in the inlet and sucked in again in the outlet. This creates aerodynamic noises (so-called air pumping).

Drive noises arise from the operation of the engine, transmission and / or drive train of a vehicle as well as their ancillary units and attachments. In electric drives, they are caused by electromagnetic forces, in the bearings and by switching processes. Drive noise enters the vehicle interior and outside as airborne and structure-borne noise. At higher speeds, air turbulence is generated on the outer shell of a vehicle.

A microphone is a sound transducer that converts airborne sound as sound pressure fluctuations into corresponding electrical voltage changes as a microphone signal.

A camera is a photographic apparatus that can record static or moving images on a photographic film or electronically on a magnetic video tape or digital storage medium or transmit them via an interface.

Warning signals are perceptible to the driver of a vehicle and increase or focus their attention. Warning signals can be optical, acoustic or haptic signals.

A direction of travel of a vehicle can be determined by means of a cardinal direction, e.g. North. A direction of travel can also be specified in relation to a street, e.g. on the B30 towards Friedrichshafen.

A crossing between a rail and a road is also called a level crossing. A level crossing can be restricted or unrestricted. Barried level crossings can be restricted to multiple or single lanes. Lowered multi-lane barriers are road blocks. At single-track, level crossings, a vehicle can avoid a lowered barrier and thus pass through it, disregarding the traffic rules. Light signal systems for emitting a warning signal are often provided at unrestricted level crossings.

An interface is a device between at least two functional units on which an exchange of logical quantities, for example data, or physical quantities, for example electrical signals, takes place, either only unidirectionally or bidirectionally. The exchange can be analog or digital. The exchange can also be wired or wireless. A controller is an electronic module for control or regulation. Controls are used in the automotive sector in all conceivable electronic areas, as well as for the control of machines, systems and other technical processes.

A reaction to a traveling noise of a rail vehicle is, for example, a warning signal to a driver or a brake signal for braking the vehicle. Means for executing the signals are reaction units.

A computing unit is a device that processes incoming information and outputs a result resulting from this processing. Electronic circuits such as central processor units or graphics processors are computing units.

Computer program products generally comprise a sequence of instructions, by means of which the hardware is caused, when the program is loaded, to carry out a specific procedure which leads to a specific result.

An artificial neural network (KNN, English artificial neural network - ANN) is in particular a network of networked artificial neurons simulated in a computer program. The artificial neurons are typically arranged on different layers. The artificial neural network usually comprises an input layer and an output layer, the neuron output of which is the only one of the artificial neural network that is visible. Layers lying between the input layer and the output layer are typically referred to as hidden layers. Typically, an architecture or topology of an artificial neural network is initially initiated and then trained in a training phase for a specific task or for several tasks in a training phase. The training of the artificial neural network typically includes a change in the weight of a connection between two artificial neurons of the artificial neural network. The weight contains information on the strength of the consideration of an input of a neuron. The training of the artificial neural network can also include developing new connections between artificial neurons, deleting existing connections between artificial neurons, adapting threshold values of the artificial neurons and / or adding or deleting artificial neurons include.

An example of an artificial neural network is a flat artificial neural network (shallow neural network), which often contains only a single hidden layer between the input layer and the output layer and is therefore relatively easy to train. Another example is a deep neural network, which contains several nested, hidden layers of artificial neurons between the input layer and the output layer. The deep artificial neural network enables improved recognition of patterns and complex relationships

For example, the artificial neural network can be a single or multilayer feedforward network or a recurrent network. Feedforward networks have neurons that are fed exclusively forward, i.e. a neuron is fed exclusively by higher layers.

A recurrent network has bidirectionally connected neurons, i.e. a neuron is also fed to deeper layers. In this way, information from an earlier run can be taken into account in a later run of the KNN, which creates a memory.

A training system is a computing unit on which a KNN is trained. In this application, training data are data pairs from input data that are to be processed by the KNN, and target result data that are to be determined by the KNN. During the training, the KNN is adjusted on the basis of a comparison of the target result data with the result determined by the KNN, which results in a training effect.

The input data with which the KNN is fed in this application are noises or audio signals which encode the noises. The input data can contain aerodynamic noises and / or contact noises from rail vehicles.

An audio signal is an electrical signal that transports acoustic information.

The basic idea of the invention is to recognize a rail vehicle based on a driving noise. This is advantageous since the traveling noise of rail vehicles has characteristic components by means of which a large number of rail vehicles can be detected.

This is particularly advantageous compared to a signal tone detection, since this depends on various factors, e.g. may depend on legal regulations in different countries.

With the invention, a driver of a road vehicle can thus also be supported when looking at a level crossing, in that the perception radius for a rail vehicle of the method according to the invention is greater than that of a driver.

Advantageous refinements and developments result from the further subclaims and from the description with reference to the figures of the drawing. According to a preferred development of the invention, the reaction is a warning signal to a driver of the road vehicle. The driver of the road vehicle can thus be informed about a potential danger from a rail vehicle.

According to a preferred development of the invention, the reaction is a brake signal for braking the road vehicle. A collision of a road vehicle with a rail vehicle can thus be prevented in an emergency situation. This increases the safety of the occupants of both vehicles.

According to a preferred development of the invention, a rail vehicle is also detected by means of a camera which can be operated on the road vehicle. The detected driving noise can thus be checked using image data. Furthermore, a driver of a road vehicle can thus be supported when viewing a track section. This is particularly advantageous on level crossings that are difficult to see, since in particular a camera is often in a better position to see a railroad track than a driver.

According to a preferred development of the invention, a position and a direction of travel of the road vehicle are also determined. It is also determined whether there is an intersection between a rail and a road in the direction of travel of the road vehicle. In this way, it can be checked whether a potential hazard due to a level crossing is possible at all. Roads and railroad tracks often run parallel to one another for a longer distance without the road crossing over the railroad tracks.

According to a preferred development of the invention, it is also determined whether a system for emitting a warning signal and / or a system for installing a road block is provided at the intersection. Thus, for example, the response to the protection at a level crossing determined by the method can be adapted.

It goes without saying that driver assistance systems for a driver of a road vehicle for carrying out a method according to one of the preceding claims with a large number of microphones for recording audio signals from the surroundings of the road vehicle, a computing unit for evaluating the recorded audio signals with regard to the detection of travel noises from rail vehicles and for determining a reaction to a detected travel noise and with a reaction unit for carrying out a determined reaction based on the detected driving noise are advantageous.

The computer program product according to one embodiment of the invention executes the steps of a method according to the preceding description if the computer program product runs on a computer, in particular an in-vehicle computer. When the program in question is used on a computer, the computer program product produces an effect, namely the detection of rail vehicles, which represent a potential danger to a road vehicle.

According to a preferred development of the invention, the audio signal has acoustic information relating to an aerodynamic driving noise of the rail vehicle and / or to a contact noise due to a contact of the rail vehicle with a rail. Rail vehicles can thus be identified particularly reliably on the basis of characteristic driving noises.

According to a preferred development of the invention, the target reaction signal of the training data has a warning signal which is directed to a driver of the road vehicle.

It is advantageous if the warning signal is designed as a haptic, optical or audio warning signal. Haptic warning signals can, for example, vibrati on signals with which objects with which a driver is in contact are acted upon. For example, a steering wheel or an area of a driver's seat can be subjected to a vibration signal. Alternatively or additionally, the warning signal can be designed as an optical signal which is shown on a display, for example on a head-up display. As an alternative or in addition, audio warning signals, i.e. sounds, are also conceivable. It goes without saying that a method for training an artificial neural network of a driver assistance system is advantageous with the following steps: providing at least one pair of an audio signal and a target reaction signal; Feeding the artificial neural network forward with the at least one audio signal; Determining an actual reaction signal based on the forward feeding; Reverse feeding of the artificial neural network due to a difference between the actual response signal and the target response signal.

The computer program product according to one embodiment of the invention executes the steps of a method according to the preceding description if the computer program product runs on a computer, in particular an in-vehicle computer. If the program in question is used on a computer, the computer program product produces an effect, namely machine learning or training of a KNN for detecting travel noises from rail vehicles.

CONTENTS OF THE DRAWINGS

The present invention is explained in more detail below on the basis of the exemplary embodiments specified in the schematic figures of the drawings. It shows:

Figure 1 is a block diagram of an embodiment of the invention;

Figure 2 is a block diagram of an embodiment of the invention;

Figure 3 is a block diagram of an embodiment of the invention.

The accompanying drawings are intended to provide a further understanding of the embodiments of the invention. They illustrate embodiments and, in connection with the description, serve to explain principles and concepts. th of the invention. Other embodiments and many of the advantages mentioned result from the drawings. The elements of the drawings are not necessarily shown to scale with respect to one another.

In the figures of the drawings, elements, features and components that are the same, have the same function and have the same effect — unless stated otherwise — are each provided with the same reference symbols.

DESCRIPTION OF EXEMPLARY EXAMPLES

FIG. 1 shows a block diagram of a training system 10 in accordance with an embodiment of the invention. The training system 10 comprises an interface 12 and an evaluation unit 20 with a KNN 22. The KNN 22 comprises several neurons, which are shown in simplified form in FIG. 1 with 108a-f. The neurons 108a, b form an input layer 102, the neurons 108c, d, e form a hidden layer 104 and the neuron 108f forms an output layer 106.

The neurons 108a, b of the input layer 102 are fed forward with an audio signal 16 via the interface 12. The audio signal 16 is weighted in the neurons 108a, b of the input layer with initial weights. It can be provided that the audio signal 16 is broken down into several signal components and the signal components are weighted. It can be provided that one or more functions are applied to the weighted input data. The evaluation of the function forms the output value of a neuron 108a, b, which are transferred to the neurons 108c, d, e of the next lower layer, that is to say the hidden layer 104, as input values. It can be provided that the hidden layer 104 has a plurality of layers.

Similar to input layer 102, the input values passed to neurons 108c, d, e of the hidden layer are weighted and one or more functions are applied to the weighted input values. The evaluation of the functions applied to the weighted input values form the output values of neurons 108c, d, e. These output values are transferred to the neurons of the output layer 106 as input values. In FIG. 1, the neurons of the output layer 106 are exemplified as a neuron 108f. Neuron 108f calculates an output value from the input values that were transferred from neurons 108c, d, e to hidden layer 104, weighting the input values and applying one or more functions to the weighted input values. An actual reaction signal 24 can be derived from this output value. This process is also known as the forward feed of a KNN.

In a next step, the actual reaction signal 24 is compared with the target reaction signal 18, which was transmitted to the evaluation unit 20 via the interface 12.

In the next step, the topology of the individual layers 102, 104, 106 of the KNN 22 is adapted in such a way that the KNN 22 calculates the target reaction signal 18 for the transferred audio signal 16. The adaptation of the topology 26 can include changing the weights, adding connections between neurons, removing connections between neurons and / or changing functions that are applied to weighted input values. This process is also known as the reverse feed of a KNN.

FIG. 2 shows a block diagram of a method for training a KNN according to an embodiment of the invention. The method comprises steps S1-S4.

In step S1, a signal pair consisting of an audio signal 16 and a desired reaction signal 18 is provided.

In step S2, the KNN 22 is fed forward with the audio signal 16.

In step S3, an actual reaction signal 24 is determined on the basis of the forward feed S2. In step S4, the artificial neural network 22 is fed backwards due to a difference between the actual reaction signal 24 and the target reaction signal 18. A changed topology 26 of the KNN, in particular weights, is thereby determined in order to improve the determination of actual reaction signals on the basis of the forward feeding.

FIG. 3 shows a block diagram of a method for detecting travel noise from rail vehicles. In step St1, a traveling noise of a rail vehicle is detected by means of a microphone that can be operated on a road vehicle. In step St2, a reaction to the detected driving noise is determined by means of a computing unit that can be operated on a road vehicle.

Bezuaszeichen

10 training system

12 interface

14 training data

16 audio signal

18 target reaction signal

20 evaluation unit

22 artificial neural network

24 Actual reaction signal

26 topology

102 input layer

104 hidden layer

106 output layer

108a-f neurons

S1 -S4 process steps

St1, St2 procedural steps

Claims

claims

1. A method for detecting travel noise of rail vehicles with the following steps:

- Detecting (St1) a travel noise of a rail vehicle by means of a microphone which can be operated on a road vehicle;

- Determining (St2) a reaction to the detected driving noise by means of a computing unit which can be operated on a road vehicle.

2. The method of claim 1, wherein the response is a warning signal to a driver of the road vehicle.

3. The method according to any one of the preceding claims, wherein the reaction is a brake signal for braking the road vehicle.

4. The method according to any one of the preceding claims, wherein a rail vehicle is further detected by means of a camera which can be operated on the road vehicle.

5. The method according to claim 1, further comprising determining a position and a direction of travel of the road vehicle and determining whether there is an intersection between a rail and a road in the direction of travel of the road vehicle.

6. The method of claim 5, wherein it is further determined whether a system for emitting a warning signal and / or a system for establishing a road block is provided at the intersection.

7. Driver assistance system for a driver of a road vehicle for performing a method according to one of the preceding claims

a large number of microphones for recording audio signals from the surroundings of the road vehicle, - A computing unit for evaluating the recorded audio signals with regard to the detection of travel noises from rail vehicles and for determining a reaction to a detected travel noise and with

a reaction unit for carrying out a determined reaction on the basis of the detected driving noise.

8. Computer program product with program code means to carry out a method according to one of claims 1-6.

9. Training system (10) for a vehicle control system for detecting travel noises from rail vehicles, with

- at least one interface (12) in order to receive training data (14) which each have an audio signal (16) and a target reaction signal (18),

- an evaluation unit (20), which forms and is set up an artificial neural network (22),

feed the artificial neural network (22) forward with training data (14) in order to determine actual reaction signals (24), and

to determine a changed topology (26) of the artificial neural network (22), in particular weights, by feeding back the target reaction signals (18) in the artificial neural network (22),

wherein the topology (26) is set up to be stored in the vehicle controller for detecting travel noise from rail vehicles.

10. Training system according to claim 9, wherein the audio signal (16) has acoustic information on an aerodynamic driving noise of the rail vehicle and / or on a contact noise due to a contact of the rail vehicle with a rail.

11. Training system according to one of the preceding claims 9 or 10, wherein a target reaction signal (18) of the training data (14) has a warning signal which is directed to a driver.

12. A method for training an artificial neural network (22) of a driver assistance system according to claim 7 with the following steps:

- Providing (S1) at least one pair of an audio signal and a target reaction signal (18);

- Forward feeding (S2) of the artificial neural network with the at least one audio signal (16);

- Determining (S3) an actual reaction signal (24) based on the forward feed (S2);

- Reverse feeding (S4) of the artificial neural network due to a difference between the actual reaction signal (24) and the target reaction signal (18).

13. Computer program product with program code means to carry out the method according to claim 12.