阅读说明：本技术 接收并分析雷达信号的雷达系统和方法 (Receive and analyze the radar system and method for radar signal ) 是由 P·格布哈特 于 2019-05-14 设计创作，主要内容包括：接收并分析雷达信号的雷达系统和方法。适用于自动化车辆的雷达系统包括至少一个接收元件,所述至少一个接收元件被配置成对由所述雷达系统的视场中的对象反射的反射雷达信号进行检测。所述接收元件还被配置成生成检测信号,所述检测信号表示被所述接收元件检测到的所述反射雷达信号。此外,对所述检测信号执行自动编码操作的自动编码装置被嵌入所述接收元件中,以使所述接收元件输出编码形式的检测信号。(Receive and analyze the radar system and method for radar signal.Radar system suitable for automated vehicle includes at least one receiving element, at least one described receiving element is configured to detect the radar signal reflected of the object reflection in the visual field by the radar system.The receiving element is further configured to generate detection signal, and the detection signal indicates the radar signal reflected detected by the receiving element.In addition, the autocoding device for executing autocoding operation to the detection signal is embedded into the receiving element, so that the detection signal of the receiving element exports coding form.)

1. a kind of radar system (10) suitable for automated vehicle, the radar system (10) include:

At least one receiving element (14), at least one described receiving element are configured to the view by the radar system (10) The radar signal reflected (20) of object (19A, 19B) reflection in is detected,

Wherein, the receiving element (14) is configured to generate detection signal (21), and the detection signal is indicated by the reception The radar signal reflected (20) that element (14) detects；

Autocoding device (22), the autocoding device execute autocoding operation to the detection signal (21), and The autocoding device is embedded into the receiving element (14), so that the receiving element (14) exports coding form (30) Detection signal (21), the coding form is preferably lossy coding form (30).

2. radar system (10) according to claim 1, which is characterized in that the radar system includes:

Transmission device (25), the transmission device are communicated with the receiving element (14)；And

Controller (26), the controller are communicated with the transmission device (25),

Wherein, code detection signal (30) is sent to the controller from the receiving element (14) by the transmission device (25) (26), and

Wherein, the controller (26) includes decoding apparatus (27), and the decoding apparatus executes decoding operate, so as to the volume Code detection signal (30) is decoded.

3. radar system (10) according to claim 2, which is characterized in that

First nonlinear function (103) is applied to the detection radar signal (21) by the autocoding device (22), so as to The detection radar signal (21) is encoded, and

The decoding apparatus (27) applies the second nonlinear function (203), to carry out to the code detection radar signal (30) Decoding.

4. radar system (10) according to claim 3, which is characterized in that

First nonlinear function (103) is different from second nonlinear function (203).

5. radar system (10) according to claim 3 or 4, which is characterized in that

First nerves network (101) obtains first nonlinear function (103) applied in autocoding operation Parameter, and

Nervus opticus network (201) obtains the parameter for second nonlinear function (203) applied in the decoding operate.

6. radar system (10) according to claim 5, which is characterized in that

The controller (26) includes discriminator (29), as the autocoding device (22) and the decoding apparatus (27) A part of training process, parameter and the second non-linear letter of the discriminator to first nonlinear function (103) The parameter of number (203) is assessed.

7. radar system (10) according to claim 6, which is characterized in that

The controller (26) includes that at least one analyzed detection radar signal (21) applies algorithm (31),

Wherein, the discriminator (29) uses the application algorithm (31) during the training process, so as to described first The parameter of nonlinear function (103) and the parameter of second nonlinear function (203) are assessed.

8. radar system (10) according to claim 7, which is characterized in that

It is described at least one using algorithm (31) include for landmarks categorizations, pedestrian classification and/or bicycle classification at least one A algorithm.

9. a kind of receiving element (14) for detecting radar signal (20), the receiving element (14) includes embedded encoded device (22), the embedded encoded device encodes the detection radar signal (21) exported by the receiving element (14).

10. a kind of reception and the method for analyzing radar signal (20), the described method comprises the following steps:

Reflection by least one receiving element (14) to object (the 19A, 19B) reflection in the visual field by radar system (10) Radar signal (20) is detected；

Detection radar signal (21) is carried out certainly by the autocoding device (22) being embedded in the receiving element (14) Dynamic coding；

Code radar signal (30) are transmitted to controller (26)；

The code radar signal (30) is decoded by the decoding apparatus (27) of the controller (26)；And

At least one for including by the controller (26) analyzes decoding radar signal (40) using algorithm (31).

11. according to the method described in claim 10, it is characterized in that,

It is described at least one using algorithm (31) include for landmarks categorizations, pedestrian classification and/or bicycle classification at least one A algorithm.

12. method described in 0 or 11 according to claim 1, which is characterized in that

Autocoding step the following steps are included: the first nonlinear function (103) is applied to the detection radar signal (21), To be encoded to the detection radar signal (21), and

Decoding step is the following steps are included: using the second nonlinear function (203), to the code detection radar signal (30) It is decoded,

Wherein, first nonlinear function (103) is preferably different from second nonlinear function (203).

13. according to the method for claim 12, which is characterized in that

Using first nerves network (101), to obtain first nonlinear function applied in autocoding operation (103) parameter, and

Using nervus opticus network (201), to obtain second nonlinear function (203) applied in the decoding operate Parameter.

14. according to the method for claim 13, which is characterized in that

As a part of the autocoding device (22) and the training process of the decoding apparatus (27), pass through the control The discriminator (29) that device (26) includes, parameter and second nonlinear function to first nonlinear function (103) (203) parameter is assessed.

15. according to the method for claim 14, which is characterized in that

At least one applies algorithm (31) described in during the training process, so as to first nonlinear function (103) parameter of parameter and second nonlinear function (203) is assessed.

Technical field

The present invention relates to the radar system for being suitable for automated vehicle and it is related to for receiving and analyzing radar signal Method.

Background technique

Now, radar system is used for many automobile applications, for example, for including auto-cruising control (ACC) and/or certainly Senior Officer's auxiliary system (ADAS) of dynamic emergency braking (AEB).This system common radar sensor (as it is preceding to or side To radar sensor) according to so-called experiment result list offer sensor information.These lists include to radar sensor The characteristic for the physical detection that data-signal executes.In general, experiment result list includes the characteristic of about 64 to 128 physical detections. Single testing result includes distance, angle, amplitude and the speed letter for reflecting the target of the radar signal sent by radar system Breath.

Although experiment result list is sufficient to traditional driver assistance system, more complicated application (as to pedestrian, Bicycle or terrestrial reference are classified) need to be already present on more detailed information in the original signal of radar sensor.Therefore, It needs a greater amount of data transmission to data processing unit, with the application complicated for these.For suitable for such application Data transmission, using be referred to as compressed data cube (CDC) another data format, without the use of experiment result list. It is original to the three-dimensional of the radar sensor of target by applying threshold value corresponding with the noise level of radar signal for CDC Data are compressed.Due to the sparsity of radar original signal, compressed data cube realizes suppressing for these data Contracting.This compression is necessary, because can not be by the way that currently available Data Transport Protocol transmits not in automotive system The data cube of compression.

Therefore, the intracorporal data compression of compressed data cube is based on the assumption that lower than the noise level arbitrarily defined Signal do not include any relevant information.This hypothesis is not necessarily correct, because some algorithms can reduce the influence of noise simultaneously Disclose the information for being lower than noise level.In addition, the object classification algorithm based on study or training process can be applied more effectively In unmodified original signal.In addition, the compression based on threshold value corresponding with noise level may be by harmful artefact (artefact) it is introduced into the signal of transmission.

Summary of the invention

The object of the present invention is to provide the radar system and method that receive and analyze radar data, the radar system and side Method is included in radar original signal and be that all relevant informations needed for specific application provide sufficiently high transfer rate.

The purpose is realized by a kind of radar system.According to the present invention, the radar system includes that at least one connects Element is received, at least one described receiving element is configured to the reflection thunder to the object reflection in the visual field by the radar system It is detected up to signal.The receiving element is further configured to generate the reflection thunder for indicating to be detected by the receiving element Up to the detection signal of signal.In addition, autocoding device executes autocoding operation to the detection signal, and described automatic Code device is embedded into the receiving element.Therefore, the detection signal of the receiving element exports coding form.

The coding form of the detection signal is preferably lossy coding form.This means that according to code detection signal reconstruction Data may be different from initial detection signal.However, data caused by due to the detection signal is encoded and reconstructed Losing for example to reach minimum by training process.

Since the autocoding device is embedded into the receiving element, it is thereby achieved that believing the reflection radar Number or the original radar signal compact unit that is received and encoded.It can easily realize in automotive vehicles applications this compact Reception and coding unit, and in the case where being not necessarily to additional code device, volume has been provided in the reception and coding unit Code data.

The output of the receiving element can also be expressed as " intermediate representation " of the detection signal.Due to being carried out to data Coding, therefore the amount of data to be transmitted would generally reduce.Therefore, compared with the receiving element of not outputting encoded data, by institute The information for stating radar signal reflected offer can be transmitted in the output section of reception device with higher transfer rate.In described Between " dimension " (that is, maximum amount of data or sample number) for indicating therefore can be according to being used to transmit accordingly connecing for encoded signal Mouthful available transmission bandwidth specify.In addition, will not introduce when the detection signal is encoded and transmitted because any fixed Justice threshold value and caused by artefact.

Due to being encoded to the initial detection signal, the detection signal is exported in the receiving element When, the detection signal is compression.Compressibility factor can be defined as the sample and the coding of the initial detection signal Detect the sample number of signal or the ratio of corresponding data collection.Therefore, the compressibility factor is the amount for reducing data to be transmitted A kind of measurement, or be the measurement for the data transfer rate for enhancing the output section of the receiving element conversely speaking,.

It should be noted that the initial detection signal is the signal for example directly generated by the antenna of the receiving element. That is, these data are considered as not yet being exposed to further process step (as being normally used for creation for example apart from more One or more Fourier transforms of Pu Letu) initial data.

Advantageous embodiments of the invention are described in the appended claims, specification and attached drawing.

According to Advantageous embodiments, the system also includes the transmission device communicated with the receiving element and with institute State the controller that transmission device is communicated.The code detection signal is sent to by the transmission device from the receiving element The controller, and the controller includes decoding apparatus, which executes decoding operate, so as to the code detection Signal is decoded.

Since the amount of data of receiving element output is reduced because of coding, can by standard transmission device (for example, CAN bus) it is used for data transmission.This transmission device is generally used for automobile application or automated vehicle.Accordingly, with respect to transmission The requirement (for example, about data transfer rate) of device is reduced because of the coding.The decoding for including by the controller Device is preferably the peer device of the autocoding device, because the decoding apparatus must be relative to the initial detection The minimum deflection of signal reconstructs the detection signal.

Specifically, the first nonlinear function is applied to the detection radar signal by the autocoding device, so as to right The detection radar signal is encoded, and the decoding apparatus applies the second nonlinear function, to believe code detection radar It number is decoded.Reconfiguration request due to the decoding apparatus about the detection signal, first nonlinear function and Second nonlinear function is depending therefrom.Therefore, number can be assessed and/or controlled by the characteristic of the nonlinear function According to the characteristic (for example, about inevitable loss of data) of transmission.For example, this can by nonlinear function institute according to The parameter of Lai Yu is controlled to execute.Preferably, first nonlinear function is different from second nonlinear function.

According to another embodiment, it is non-that first nerves network obtains described first applied in autocoding operation The parameter of linear function, and nervus opticus network obtains the ginseng for second nonlinear function applied in the decoding operate Number.It can realize the first nerves network and the nervus opticus network on development system, which is by institute It states and uses before radar system is for example mounted on automated vehicle.It alternatively, can be in one as the radar system The first nerves network and the nervus opticus network are realized on the partial controller.

Since first nonlinear function and second nonlinear function are depending therefrom, therefore, because reconfiguration request, The first nerves network and the nervus opticus network can combine with one another.Specifically, convolutional neural networks are very suitable to locate Manage radar initial data.Therefore, the parameter set of first nonlinear function and second nonlinear function can be considered as By the parameter set of the first nerves network and the nervus opticus network provision.

Preferably, the controller includes discriminator, the training as the autocoding device and the decoding apparatus A part of process, the discriminator comment the parameter of first nonlinear function and second nonlinear function Estimate.Therefore, the task of the discriminator is in this way to first nonlinear function and the second non-linear letter Several parameters are assessed: so that the difference between the initial detection signal and the reconstruction signal is minimum.In test environment In (that is, before realizing the radar system on vehicle), the discriminator connects during training process from the receiving element Receive the un-encoded initial detection signal or original signal and the code detection signal and decoding (that is, reconstruct) detection Signal.

During the training process, the parameter of first nonlinear function and second nonlinear function is by described First nerves network and the nervus opticus network change, until the initially signal and the reconstruction signal each other enough one It causes.It can specify that the residual error (residual deviation) between the initially signal and the reconstruction signal, to terminate The training process.Furthermore it is possible to define the residual error in this way: it is reduced as far as the amount of data to be transmitted, or Person maximizes the relevant information obtained by the data.

Specifically, the controller include analyze the detection radar signal at least one using algorithm, and it is described Discriminator applies algorithm using described during the training process, so as to first nonlinear function and described second non- The parameter of linear function is assessed.Preferably, it is described at least one using algorithm include for landmarks categorizations, pedestrian classify And/or at least one algorithm of bicycle classification.In addition, can be by other by least one described result for applying algorithm to provide Automobile application (for example, by Senior Officer's auxiliary system (ADAS)) is further processed.

Instead of the original signal of directly more described initially signal and reconstruct, institute can be used during the training process It states using algorithm, is assessed with the parameter to first nonlinear function and second nonlinear function.For example, can be with Will for identification object or classify subjects into (for example, identification terrestrial reference or pedestrian) algorithm be applied to the initial detection signal and The reconstruction signal, and if recognition result has enough consistency, terminate the training process.

Through these means, the training can be customized and optimized for the required application of the radar signal is detected Process.In addition, inevitably loss of data therefore can be relative to institute caused by the autocoding and due to decoding operate For example, identifying object, rather than any level (such as noise level) need to be provided to limit data to be transmitted using minimizing Amount.Therefore, inevitable loss of data " transfer " is arrived to the extraneous data of required application by the training process.

Specifically, before the radar system is mounted on automated vehicle, by application mass data collection (that is, It is from known object, using the detection signal of original form and using the detection signal of coding form and decoded form) come Execute the training process.However, during the operation of the radar system of automated vehicle, it is preferable that can repeat to instruct Practice process, so that the parameter of parameter and second nonlinear function to first nonlinear function is verified.For example, When the transmission device allow transmit mass data when (for example, when vehicle does not move), training process can be triggered.

The invention further relates to can be used in the above-mentioned radar system for detecting the receiving element of radar signal.It is described Receiving element includes embedded autocoding device, the detection of the embedded autocoding device to being exported by the receiving element Radar signal is encoded.

In addition, the present invention relates to a kind of methods for receiving and analyzing radar signal.By at least one receiving element to by The radar signal reflected of object reflection in the visual field of radar system is detected.By be embedded into the receiving element from Dynamic code device carries out autocoding to detection radar signal.Then, by code radar signal be sent to controller and pass through by The decoding apparatus that the controller includes is decoded.Algorithm is applied finally by least one for including by the controller Decoding radar signal is analyzed.

Due to autocoding step and decoding step, it is possible to reduce from the receiving element to the to be transmitted of the controller The amount of data.It is thus possible to improve the transfer rate of relevant information.Further, since not using any restrictions of data (as making an uproar Vocal level), therefore can make artefact occur in reconstruct radar signal or decoding radar signal to reach minimum.

Detailed description of the invention

Referring to Advantageous embodiments and attached drawing, present invention is described in an illustrative manner.It shows:

Fig. 1 is the radar system based on vehicle of embodiment according to the present invention；

Fig. 2 is the schematic diagram of the data compression and data reconstruction that are executed by the system of Fig. 1；And

Fig. 3 is to instantiate the flow chart of the training process of the system execution for Fig. 1.

Specific embodiment

Fig. 1 instantiates the radar system based on vehicle, hereinafter referred to system 10.System 10 is including radiated element 12 and extremely A few receiving element 14, receiving element 14 includes mutiple antennas 16, for example, first antenna 16A and the second antenna 16B.Transmitting Element 12 emits radar signal 18, and radar signal 18 is by the first object 19A and/or the second object 19B in the visual field of system 10 It is reflected into radar signal reflected 20.The first antenna 16A of receiving element 14 and the second antenna 16B carries out radar signal reflected 20 Detection, in order to provide the detection signal 21 for indicating radar signal reflected 20.

In addition, receiving element 14 includes autocoding device 22.In other words, autocoding device 22 is embedded in reception member In part 14, to provide compact unit together with antenna 16.Autocoding device 22 is by (joining the first nonlinear function 103 See Fig. 3) it is applied to detection signal 21 to execute autocoding operation to detection signal 21.Therefore, autocoding device 22 exports Code detection signal 30.In short, include embedded autocoding device 22 14 exports coding of receiving element detection signal 30 and It is not initial detection signal 21.

Due to being encoded to detection signal 21, as shown in the left side of fig 2, data pressure is executed to these signals Contracting.As radar sensor, the typical sample rate of antenna 16 is 100MHz to 1000MHz, and typical resolution is 16.Detection letter Numbers 21 single sweep operation is that autocoding device 22 provides input X, and input X is made of 1000 to 10000 samples.By In having carried out autocoding through device 22, therefore, code detection signal 30 is acquired as so-called intermediate representation Z, in this Between indicate Z include much less data volume, for example, only hundreds of samples.In this example, due to autocoding, this causes to examine The compressibility factor for surveying signal 21 is about 10.

System 10 further includes the transmission device 25 communicated with receiving element 14 and is communicated with transmission device 25 Controller 26.Code detection signal 30 is sent to controller 26 from receiving element 14 by transmission device 25.Controller 26 includes: solution Code device 27, discriminator 29 and the application algorithm 31 that be applied to detection signal.In addition, controller 26 includes to radiated element 12 provide transmitting signal 28 to generate the device (not shown) of transmitting radar signal 18.

For example, transmission device 25 is the CAN bus used in typical automotive system.Transmission device 25 is by code detection Signal 30 is sent to the decoding apparatus 27 of controller 26 from receiving element 14.Decoding apparatus 27 is by by the second nonlinear function 203 (referring to Fig. 3) are applied to code detection signal 30 to execute decoding operate.

The decoding operate executed by decoding apparatus 27 is instantiated on the right side of Fig. 2.By the way that the second nonlinear function 203 is applied In code detection signal 30 (or intermediate representation Z), decoding or the reconstruction signal 40 of the output X' as decoding apparatus 27 are generated.By In nonlinear function 103,203 is applied to initially signal and encoded signal, therefore, it may occur that inevitable loss of data, That is, decoding or reconstruction signal 40 may deviate initial detection signal 21.However, not by any limitation (for example, with noise electricity Equal corresponding threshold value) it is applied to signal.On the contrary, optimizing the parameter of nonlinear function 103,203 by training process.

In the first operator scheme of system 10 (that is, during training process), the output X'(of decoding apparatus 27 is that is, solution Code signal 40) it is used as the input of discriminator 29, discriminator 29 is to first applied in coding step and decoding step and The parameter of two nonlinear functions is assessed.This comment is executed by the way that initial detection signal 21 is also sent to controller 26 Estimate.

In the first evaluation form, decoded signal 40 and initial detection signal 21 are compared, and change first The parameter of the parameter of nonlinear function 103 and the second nonlinear function 203, until initial detection signal 21 and decoded signal 40 Between deviation reach minimum.It additionally include that (it is usual using algorithm 31 in appraisal procedure in second of evaluation form Applied to detection signal 21).That is, detection signal 21 and decoded signal 40 will be respectively applied to using algorithm 31, then Implementing result compares.By this method, the identification step executed by discriminator 29 can be applied to mention for detection signal 21 Algorithm 31 is applied in confession.

In the identification step of two kinds of forms, the comparison result of initially signal and reconstruction signal is used to assess and control One and second nonlinear function parameter.Inhibit the data without or with smaller correlation automatically by identifying as a result,.It is right The parameter of first nonlinear function 103 and the parameter of the second nonlinear function 203 execute assessment, until initially signal and reconstruct are believed Deviation between number (that is, original signal or apply result) reaches minimum.In detail, by training process as shown in Figure 3 come Execute assessment.

According to the training process, by the output (that is, detection signal 21) of antenna 16 be fed to first nerves network 101 with Just code detection signal 30 is generated, which determines the ginseng for executing the first nonlinear function 103 of coding step Number.These encoded signals 30 are fed to nervus opticus network 201 again, which determines that second is non-linear The parameter of function 203.After the second nonlinear function 203 of application, reconstruct detection signal 301 is supplied to discriminator 29.Separately Outside, initial detection signal (that is, copy of detection signal 21) is also supplied to discriminator 29.

Reconstruct detection signal 301 and the initially comparison result of detection signal 303 generate the first feedback from discriminator 29 105 and second feed back 205, and the first feedback 105 and the second feedback 205 are respectively by first nerves network 101 and nervus opticus network 201 use.Due to this feedback, the node of neural network 1 01,201 is updated, and the parameter of first function 103 and second The parameter of function 203 is corrected.Therefore, the first learning procedure of training process is completed.

For example, by another group of detection signal 21 being provided for known object 19A, 19B come the study of repetition training process Step reaches minimum until initially detecting the deviation between signal 303 and reconstruct detection signal 301.

In the second operator scheme of system 10 (referring to Fig. 1), by the output data of decoding apparatus 27 (that is, decoded signal 40 (referring to fig. 2)) it is supplied directly to using algorithm 31, that is, complete the first nonlinear function 103 and the second nonlinear function After the training process of 203 parameter.Training process executes in test environment first, that is, realizes by radar system 10 Such as in automated vehicle before.It, can be by not available attached usually in automated vehicle in such test environment Computing hardware is added to be easily performed the data transmission of initial detection signal 21,303.However, the execution of second operator scheme can To be interrupted in later phases, and can be with repetition training process (for example, utilizing the data set of reduction), so as to non-linear to first The parameter of function 103 and the parameter of the second nonlinear function 203 are verified.

Since compared with initial detection signal 21 and decoded signal 40, code detection signal 30 or intermediate data indicate Z It (referring to fig. 2) include therefore the data of quantity reduction compared with un-encoded and decoded data transmission, code detection are believed Numbers 30 increase from the transfer rate that receiving element 14 is sent to controller 26.In addition, by including the ginseng determined by training process Several nonlinear functions executes the inhibition of non-relevant data.As a result, because of the autocoding and decoding by nonlinear function Caused by inevitably loss of data adapts to automatically needed for apply.Therefore, artefact is not introduced, because not setting arbitrarily Set threshold value for example corresponding with noise level.On the contrary, radar system 10 of the invention ensures decoded signal 40 or reconstruct detection Signal 301 includes used using maximum information needed for algorithm.

Reference signs list

10 radar systems

12 radiated elements

14 receiving elements

16 antennas

16A first antenna

The second antenna of 16B

18 transmitting radar signals

The first object of 19A

The second object of 19B

20 radar signal reflecteds

21 detection signals

22 autocoding devices

25 transmission devices

26 controllers

27 decoding apparatus

28 transmitting signals

29 discriminators

30 code detection signals

31 apply algorithm

40 decoded signals

101 first nerves networks

103 first nonlinear functions

105 the first feedbacks from discriminator

201 nervus opticus networks

203 second nonlinear functions

205 the second feedbacks from discriminator

301 reconstruct detection signals

303 initially detect signal