Method for tidying up a data set comprising at least one time series, device for performing, vehicle and computer program

Abstract

The invention relates to a method for processing a data set (DS) comprising at least one time series. This is for example measurement values acquired at specific points in time by sensors. The measured values (int(0)-int(n-1)) and the corresponding points in time (t(0)-t(n-1)) at which the data acquisition took place are stored as data elements (DP(0)-DP(n-1)) of a time series. The method performs a compression of the data set (DS) in such a way that a step of rounding the acquired data (int(0)-int(n-1)) is performed and subsequently a decimation of the data elements (DP(0)-DP(n-1)) of the data set (DS) comprised in the time series is performed. In the transmission of the data set (DS) there is, in addition to the advantages of a reduction in the data volume, a saving in storage capacity and transmission capacity, the possibility of performing a data analysis on the compressed data without prior decompression.

Description

Technical Field

[0001] This invention relates to the technical fields of data compression and data analysis. In vehicles, an increasing amount of data is being collected and transmitted to external computing centers capable of performing data analysis. This invention also relates to a correspondingly designed device for data compression and / or data analysis, a vehicle equipped with such a device, and a correspondingly designed computer program. Background Technology

[0002] With the development of new technologies ranging from private transportation to autonomous driving, it is conceivable that in the near future, an increasing amount of data will have to be exchanged between vehicles and between vehicles and computing centers using databases, improved vehicle sensors, and other means at the system level. This necessitates a massive expansion of communication networks, as vehicles must exchange ever-increasing amounts of data with each other via vehicle-to-vehicle (V2V) communication and with backend servers via vehicle-to-infrastructure (V2X) communication.

[0003] Therefore, an increasing amount of data is being transmitted from vehicles to the "cloud." The bandwidth available for V2X communication is limited. Therefore, one solution to this conflict is to compress the data in the vehicles and only transmit the compressed data to the backend server (in the "cloud").

[0004] The known compression method used for this purpose is the ZIP method, which can compress large datasets of any amount of data without loss.

[0005] A method for lossless payload compression is known from US 2018 / 253559 A1. For payload compression, entropy coding such as Huffman coding and LZ77, corresponding to Lempel-Ziv 77, is used. In a variant, selected portions of the payload in a data block are encrypted using a Huffman-coded specification according to the AES method (which corresponds to the Advanced Encryption Standard), indicating the start and end of the encryption restrictions. This allows for the selective application of higher levels of security, as desired in a particular application.

[0006] The known solutions have drawbacks. This is known within the scope of this invention. The problem with the known solutions is that while they achieve compression of the raw data, requiring less storage space to store the compressed data and less bandwidth in the communication systems used to transmit it, the analysis of the compressed data is associated with overhead. Therefore, a data decompression step needs to be performed before analysis can occur. This can lead to significant additional overhead, for example, in the case of large datasets collected throughout the operating time for applications in vehicles, machines, and devices. If the data is analyzed in a backend server, there is usually sufficient computing power available to perform decompression in addition to other server tasks. However, this also incurs costs. However, the situation is different if the analysis itself should be performed in the vehicle or machine. For applications in vehicles or machines, for example, it is necessary to monitor the wear condition of wear parts during operation. Here, information regarding the necessity of maintenance should be output to the driver or machine operator as soon as possible. This is absolutely necessary in the case of safety-related wear parts. However, in vehicles, electronic control devices equipped with microcontrollers with limited computing power are used. As mass-produced products, these control devices are also under increasing price pressure. Furthermore, control devices operate in real-time, thus they cannot execute, or can only execute, specific data analysis methods with additional computing power, the computational overhead of which depends at least linearly on the number of data points to be analyzed. Limitations may also arise from the need to transmit data through the vehicle's bandwidth-limited internal communication network. Additionally, the reconstructed data, at best, has the same size as the original data. In the case of a 1-bit status signal (e.g., driver's door closed / open) transmitted via the vehicle's communication network (e.g., CAN bus (Controller Area Network)), analysis using the floating-point data type "double precision, 64 bits" means a 64-fold increase in memory compared to the original size, or an 8-fold increase compared to an "8-bit unsigned integer" representation. The implementation and execution of compression algorithms in vehicle control devices require permanent memory for program code and volatile memory for runtime and computation-time variables. Some compression methods are not considered for application in control devices due to their resource requirements, or they can only be used to compress a limited number of signals.

[0007] However, for cloud applications, reducing the overhead associated with data analytics is also of interest. In automotive applications, raw data is compressed within the vehicle and then decompressed again on the backend server. In the case of high-volume automakers, this results in a massive backlog of data that must be stored and processed on the backend. When millions of vehicles are produced annually, it's easy to imagine the need to build enormous computing centers to store and process all this data. This could be advantageous if the required overhead could be limited.

[0008] In addition, methods known from EP 2 645 318 A1 and EP 3 522 577 A1 are used in which data values ​​that are recorded successively are persistently stored or transmitted to the host only if the difference between the data values ​​recorded successively and the last stored / transmitted data value is greater than a pre-given threshold. Summary of the Invention

[0009] Therefore, there is a need to develop an improved compression algorithm that uses fewer resources and can be variably set in terms of compression level and loss rate, while reducing the overhead for decompression or data analysis. This is the technical problem that this invention aims to solve.

[0010] The technical problem is solved by the method for organizing a data set containing at least one time series according to claim 1, the apparatus for performing the method according to claims 13 and 16, the vehicle according to claim 15, and the computer program according to claim 17.

[0011] The dependent claims include advantageous extensions and improvements of the invention as described below, corresponding to these measures.

[0012] In a general configuration, the present invention relates to a method for organizing a data set comprising at least one time series, wherein the time series includes data collected at specific points in time and the corresponding points in time at which the data was collected as data elements. The points in time at which the data was collected may be represented, for example, as absolute time or as incremental time (Deltazeit) at absolute points in time, or as an index in the case of equidistant data records in time. The method compresses these data sets in a specific manner, wherein the steps of rounding the collected data and subsequently extracting the data elements of the data set contained in at least one time series are performed. The data compression method provides advantages known from other known data compression methods, such as reducing the amount of data to be stored or transmitted. However, the method additionally provides the particular advantage that the compressed data set can be analyzed without having to perform a decompression algorithm. This is a significant advantage because data often needs to be analyzed not only for archiving but also for maintenance purposes, for identifying safety-critical developments, for adapting functionality to user behavior and preferences, for warranty purposes when necessary, and for determining incident events.

[0013] Specifically, an advantageous solution for the extraction step is to remove data elements whose rounded data is indistinguishable from the rounded data of their immediate predecessor in time. Therefore, high compression rates >10 can be achieved across multiple time series where the acquired measurements have hardly changed.

[0014] A common application scenario involves time series corresponding to measurement sequences, where the measurements are collected as integers or floating-point numbers. In this case, performing a conversion to integers before applying the compression method is advantageous for the compression approach.

[0015] Advantageously, in order to convert to integers, the decimal places of the floating-point numbers of the time series data elements are converted to integer decimals by decimal point shifting, and then converted to integers in a further step.

[0016] An additional advantage is that an algorithm for rounding integers is performed when preparing to extract data elements from the time series. The strength of the rounding operation can be predetermined here. This has a decisive impact on the degree of compression.

[0017] In a particularly advantageous variation, rounding operations are prepared according to the following rules:

[0018] rnd(x)=AND(BSR(int(x),nbit-1),1)

[0019] Here, rnd(x) corresponds to the rounding information of the integer selected for subsequent rounding operations, where x corresponds to the data element index, int(x) corresponds to the integer, nbit-1 indicates which bit position of the integer should be selected to calculate the rounding information rnd(x) according to the mentioned rules, and AND corresponds to the logical AND operation between the entry and the integer value "1", which is located at the bit position determined by right-shifting (BSR (Bit Scan Reverse) function) the integer int(x) by nbit-1 bit positions. The degree of compression can be set by selecting the parameter nbit.

[0020] The actual rounding operation can be performed in a favorable manner according to the following rules:

[0021] comp(x)=BSR(int(x),nbit)+rnd(x)

[0022] Here, `comp(x)` corresponds to the reduced resolution of the integer `int(x)` produced by rounding, and `nbit` specifies how many bits to shift the integer to the right before adding it to the rounding information `rnd(x)`. By choosing parameter `nbit = 0`, the compression algorithm can be set to lossless compression. In this case, there is no need to calculate the rounding information `rnd(x)`, as it is zero.

[0023] After the preparation steps, data elements can be extracted from the data set. Advantageously, the rounded value comp(x) is compared with comp(x+1), and subsequent data elements with the same value comp(x+1) are discarded from the time series. Therefore, in cases where data always repeats, these data elements can be eliminated from the data set, achieving a high compression rate.

[0024] Compression algorithms can also be implemented directly on the floating-point representation of the data. The floating-point number is then divided by the smallest distinguishable increment in the reduced representation, rounded, and then multiplied again by the smallest distinguishable increment in the reduced representation. Typically, the execution time in the integer representation is less than that in the floating-point representation, so an integer implementation is preferred.

[0025] In another implementation variant, the compression algorithm described above can be implemented within time windows. A time window can, for example, have a fixed length of 1 second, or it can be started or ended by an external event such as "opening the vehicle" or "locking the vehicle." Typically, a new window opens directly after the previous window closes. Then, for example, a constant signal is included in each window containing data elements.

[0026] The window length defines the maximum achievable compression ratio and the frequency at which compressed data groups are transmitted. Longer windows require a higher maximum compression ratio.

[0027] The compressed data set can be stored in the computing unit where compression was performed. Another advantageous aspect of the method, if data analysis is to be performed at an external location, is to transfer the compressed data set from the device to an external computing device where the data set was compressed.

[0028] This can be advantageously implemented using public cellular mobile radio communication systems or other mobile radio communication systems. With the introduction of 5G mobile radio communication systems, more and more vehicles, machines, and devices are connecting to the Internet on this basis. As another example of a mobile radio communication system, WLAN p systems based on the IEEE 802.11p standard are mentioned, which are also specifically used for V2V and V2X communications.

[0029] What is particularly advantageous for data analysis is that it allows for the analysis of time series data extracted from a dataset without the need to decompress the dataset.

[0030] For box-plot analysis of compressed datasets, this can be performed by sorting the extracted time series in ascending order of the reduced-resolution comp(x) information, and selecting eigenvalues ​​for box-plot analysis from the compressed datasets, particularly the minimum, maximum, median, lower quartile, and upper quartile, taking into account the time difference between each data element and the next.

[0031] Data compression and data analysis can be performed on a common device or on separate devices.

[0032] Advantageously, the device used to perform the method has at least one computing unit designed to acquire or receive measurements from at least one connected sensor at successive time points. Additionally, it is advantageous that the computing unit is designed to form a time series using the measurements and associated measurement time points, and further perform at least the following steps: rounding, and subsequently extracting the measurements of data elements contained in at least one time series. Particularly in vehicles, sensors may not be directly connected to the device used to perform the method, but rather transmit data regularly to the device via a communication bus, such as a CAN bus. However, this can also be done in other application areas. As another example, automation technology is mentioned, in which sensors can be connected to a fieldbus system. Measurements from sensors are not necessarily required. The calculation results can also be used to build a time series, which are archived as data content in the memory of the computing unit, i.e., ultimately as any information digitally present in the vehicle. As an example, calculated average vehicle consumption is mentioned.

[0033] Furthermore, it is advantageous that the computing device or another computing unit of the apparatus is designed to perform box plot analysis on the compressed data set by sorting the extracted time series in ascending order of the reduced-resolution comp(x) information, and selecting eigenvalues ​​for box plot analysis, particularly the minimum, maximum, median, lower quartile, and upper quartile, from the compressed data set, taking into account the time difference between each data element and the next. In this variant of the invention, both data compression and data analysis occur within the apparatus. Vehicles can advantageously be equipped with such an apparatus.

[0034] Therefore, another configuration of the invention is a vehicle having the device according to the invention.

[0035] Furthermore, another variation of the invention involves an apparatus for performing the method, the apparatus having at least one computing device designed to receive compressed data sets. Here, the computing device is designed to perform time-series analysis on the extracted data sets without decompressing them. This apparatus is preferably used in a backend server to which the compressed data sets are transmitted.

[0036] The computing device can be designed here to perform box plot analysis on the compressed data set by sorting the extracted time series in ascending order of the reduced-resolution comp(x) information, and selecting eigenvalues, particularly the minimum, maximum, median, lower quartile, and upper quartile, from the compressed data set, taking into account the time difference between each data element and the next.

[0037] Finally, another embodiment of the invention is a computer program designed to perform, when run in a computing unit, the steps of a method for organizing a data set containing at least one time series according to one of the methods of the invention. Attached Figure Description

[0038] Embodiments of the present invention are shown in the accompanying drawings and are described in more detail below with reference to the drawings.

[0039] In the attached diagram:

[0040] Figure 1 This demonstrates the principle of vehicle-to-vehicle communication via mobile radio and a connection to the Internet;

[0041] Figure 2 The structure of a data group with time series and individual data elements is shown;

[0042] Figure 3 The traditional process is illustrated by compressing a data set and transmitting it to a backend server connected via the Internet, and then analyzing the data set after decompression.

[0043] Figure 4 The overall scheme for the compression process of data sets is shown, which allows for subsequent analysis of the data sets after they are transmitted to the backend server without prior decompression.

[0044] Figure 5 A block diagram of the vehicle's onboard electronics is shown;

[0045] Figure 6 The original data of the time series of an exemplary data set and its conversion to integers are shown;

[0046] Figure 7 It shows Figure 6 A graphical representation of the raw data for the example time series;

[0047] Figure 8 The results of rounding information calculated according to the overall scheme and a table of rounding results are shown in an example data set;

[0048] Figure 9 It shows Figure 6The process of extracting rounded data elements from a time series example;

[0049] Figure 10 The compressed data set after extracting the data elements from the time series is shown in tabular form.

[0050] Figure 11 It shows Figure 6 A graphical representation of the time series data retained in the data set after compression;

[0051] Figure 12 It shows Figure 6 A graphical representation of the compressed time series data compared to the original data;

[0052] Figure 13 An example of applying box plot analysis to the original dataset is shown;

[0053] Figure 14 It shows the application to the raw data Figure 6 A graphical representation of the box plot analysis of the example time series;

[0054] Figure 15 An example of applying box plot analysis to data analysis of compressed datasets is shown; and

[0055] Figure 16 This shows the application to the compressed data. Figure 6 The example is a graphical representation of the box plot analysis of the time series. Detailed Implementation

[0056] This description illustrates the principles disclosed in this invention. Therefore, it should be understood that those skilled in the art will be able to design different arrangements, which, although not explicitly described herein, embody the principles disclosed in this invention and should also be protected within the scope of this invention.

[0057] Figure 1A system architecture for vehicle communication via mobile radio is illustrated. The vehicle is indicated by reference numeral 10. A passenger car is shown. However, any other vehicle can also be considered a vehicle. Examples of other vehicles include: buses, commercial vehicles (especially trucks), agricultural machinery, construction machinery, camping vehicles, motorcycles, bicycles, scooters, wheelchairs, rail vehicles, etc. The use of this invention in vehicles will be universally applicable to land vehicles, rail vehicles, watercraft, and aircraft. Furthermore, the use of this invention is not limited to vehicles. In fact, this invention can generally be used in all areas of electrical engineering. Further examples include machines and devices, small electronic devices, consumer electronics, white goods, medical devices, etc. This list is not exhaustive. The connection of devices to the “cloud” is also extending to an increasingly wider range of fields. A typical keyword in this regard is the term “Internet of Things” (IoT), which represents a technological trend of an increasing number of devices in industry, commerce, and homes connecting to the internet using newer communication technologies such as 5G.

[0058] The vehicle 10 is equipped with an on-board communication module 160 with a corresponding antenna unit, so that the vehicle can participate in different types of vehicle communication, such as vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2X). Figure 1 The vehicle 10 is shown to be able to communicate with a mobile radio base station 210 of a mobile radio provider.

[0059] Such a base station 210 can be an eNodeB base station of an LTE (Long Term Evolution) mobile radio provider or a 5G (5th Generation Mobile Radio System) mobile radio provider. Base station 210 and related equipment are part of a mobile radio communication network with multiple mobile radio units, each of which is served by base station 210.

[0060] Base station 210 is typically located near the main road on which vehicle 10 travels. Additionally, vehicle 10 is equipped with an onboard communication module 160. The onboard communication module 160 corresponds to an LTE communication module, which allows vehicle 10 to receive mobile data (downlink) and transmit such data in the uplink direction (uplink). However, it also supports V2V and V2X communication via 5G mobile radio systems. There, the corresponding radio interface is called the PC5 interface. Regarding the LTE mobile radio communication system, the LTE E-UTRAN (Evolved UMTS Terrestrial Radio Access Network) consists of multiple eNodeBs, which provide the E-UTRA user plane (PDCP / RLC / MAC / PHY) and control plane (RRC). The eNodeBs are interconnected via a so-called X2 interface. The eNodeBs are also connected to the EPC (Evolved Packet Core) 200 via a so-called S1 interface.

[0061] Figure 1 As shown in this general architecture, base station 210 is connected to EPC 200 via an S1 interface, and EPC 200 is connected to the Internet 300. Backend server 320 is also connected to the Internet 300, and vehicle 10 can send and receive messages to and from the backend server. Backend server 320 can be located in the computing center of a vehicle manufacturer or other mobile service provider, or in the computing center of an administrative agency, such as a traffic dispatch center. Finally, road infrastructure station 310 is also shown. This road infrastructure station can be represented, for example, by a roadside unit, often referred to in technical terms as a "Road Side Unit" (RSU) 310. For simplicity, it is assumed that all components are assigned Internet addresses, typically in the form of IPv6 addresses, so that packets transmitting messages between components can be routed accordingly. All the aforementioned interfaces are standardized. In this regard, refer to the corresponding published specifications for mobile radio communication systems.

[0062] Figure 2shows by way of example the structure of a data set containing a time series. A time series may correspond to a measurement sequence. Here, temporally successive measured values of physical parameters such as temperature, pressure, humidity, speed, duration, rotational speed, acceleration, length, voltage, resistance, light intensity, chemical composition and the like are acquired, for example. In the simplest case, the time series consists of a plurality of data elements, each of which contains a measured value and an associated acquisition time point. The acquisition time points usually follow one another at time-equidistant intervals. However, time series also arise if measured values are only acquired occasionally or periodically with a variable cycle duration. In Figure 2 a time series with n data elements is illustrated. The data elements of the time series are denoted DP(0) to DP(n-1). Each individual data element contains the acquired measured values int(0) to int(n-1) as integers and respectively associated timestamps t(0) to t(n-1), which specify the corresponding acquisition time points. A time series is characterized in that the data elements are sorted according to the acquisition time points. Usually, it is even required that the acquisition time points must be sorted in a strictly monotonically increasing manner. This requirement can be expressed by the relationship t(x)<t(x+1). Sensors do not always provide their acquired measured values as integers. It is often the case that a sensor generates an analog signal which is converted into a digital signal, for example via an analog input of a control device or a microcontroller. For this purpose, an analog-to-digital converter, which may likewise be part of the microcontroller, is usually used. A conversion table is then typically used to convert to the physical parameter. However, it is well known that physical parameters are usually signed and given in specific units, which may result in rational numbers or real numbers that are typically stored as floating-point numbers in a computer. Here, "single-precision" or "double-precision" representations typically exist, where "single-precision" floating-point numbers are stored in a 32-bit memory space, and "double-precision" floating-point numbers are stored in a 64-bit memory space. It is always possible to convert a floating-point number into an integer and vice versa. In this respect, reference is made to the following conversion formula for converting an integer into a floating-point number:

[0063] fp(x)=int(x)*f+o

[0064] wherein fp(x) corresponds to the floating-point number, f corresponds to a multiplication factor, and o corresponds to an offset value.

[0065] In Figure 3The diagram illustrates the conventional workflow of data acquisition and analysis within the vehicle, which also includes cloud transmission and storage, as well as cloud-based data analysis. The upper portion P10 contains the processing steps within the vehicle 10. The middle portion T200 contains the step of transmitting compressed data sets to a backend server 320 via mobile radio. The lower portion P320 contains the cloud processing steps. The steps within the vehicle 10 are described first. Measurement acquisition is not shown separately here. The step S10, storing data in the memory of the control device, is shown first. In step S12, the data set is read from the memory and fed to a data compression algorithm. This algorithm is executed in step S14. As mentioned at the beginning, a known ZIP compression algorithm can be used in step S14. In step S16, the compressed data set is forwarded to the corresponding onboard communication module 160 of the vehicle 10 for transmission via mobile radio.

[0066] There are applications that require evaluation of data sets stored in vehicle 10. Take moving box plot data analysis as an example. A box plot should quickly provide an impression of the area where the data is located and how the data is distributed within that area. Therefore, all values ​​of the so-called five-number generalization are determined: the median, the lower quartile (25th quartile), the upper quartile (75th quartile), and the two extreme values ​​(minimum and maximum). These values ​​can be well graphically represented as numerical lines using entries for the five values ​​of the five-number generalization. Box plot representation may be suitable for quickly acquiring the condition of worn parts. In step S18, the moving box plot analysis occurs in vehicle 10.

[0067] Another data analysis method that can be used in vehicles involves break / drift analysis. This corresponds to trend analysis. Creep wear can be identified through this. Take the clutch wear of vehicle 10 as an example. For this purpose, for example, the process from engaging first gear with the transmission to the vehicle starting to roll is measured. As the clutch wears more severely, the operating path required to disengage the clutch pedal during the rolling moment changes slowly. Break / drift analysis occurs in step S20. In step S22, the results of the data analysis are forwarded to the on-board diagnostic unit, which is responsible for notifying the driver of the wear condition of vehicle 10, for example, by outputting a warning message on the display unit or requesting immediate workshop access.

[0068] It is important to emphasize that traditional vehicle data analysis processes use uncompressed data. Therefore, in Figure 3 As shown, after step S12, the uncompressed raw data is forwarded to data analysis steps S18 and S22. Data analysis in the vehicle requires processing all the raw data.

[0069] After step S16, the compressed data set is transmitted via mobile radio to a computing center with a backend server 320. In step S26, the data set is archived there in compressed form. However, for data analysis, it is necessary to convert the data set back to a processable form (original form). For this purpose, a decompression algorithm is used in step S28. In step S30, the reconstructed data set is provided for subsequent data analysis. The reconstructed data set may also be stored in an archive memory for later use, as indicated in step S32. As already described in the vehicle-side data analysis, the moving box plot analysis in step S34 and the interruption / drift analysis in step S36 can also be performed in the computing center. The results are provided to the cloud diagnostic unit, see step S38. This unit can be responsible for sending the corresponding warning messages back to vehicle 10 via mobile radio, with the data originating from the vehicle. Other analysis methods that can be performed in the computing center are also shown. This includes similarity analysis performed in step S40 and pattern recognition in step S42. Further details on different analysis methods can be found in the references. In step S44, other algorithms may also be applied for data analysis.

[0070] Compared to traditional forms of data compression, data archiving, and data analysis, Figure 4 The diagram illustrates the process of data acquisition and analysis in a vehicle, as well as cloud processing, according to an improved method based on the invention. The same reference numerals denote... Figure 3 The same steps and components are used. The essential difference lies in the compression method for time-series data sets. A compression algorithm that allows for data analysis of the compressed data is used. The details of the new compression algorithm will be explained in more detail below. This algorithm allows for data analysis of the compressed data. Therefore, in Figure 4 As shown, in step S16, the compressed data is forwarded to the analysis step S18. Therefore, significantly less data must be forwarded, and less data must be analyzed, as explained in more detail below. This also facilitates archiving and analysis of data in the "cloud." Figure 4 Therefore, steps S28 to S32 can be omitted. Data analysis is performed on the specially compressed data in steps S34 to S42. It is possible that the data analysis step cannot be performed on the specially compressed data. For this reason, a decompression algorithm is set up to be executed in step S46. This is possible because the special compression algorithm is also designed to be reversible, thus allowing data recovery. Here, the compression algorithm can be set to either lossless or lossy compression. If the compression algorithm is set to lossy, the original data cannot be recovered without loss of accuracy.

[0071] Figure 5A block diagram of in-vehicle electronics is schematically shown, including the infotainment system of vehicle 10. In vehicles, particularly sedans, an infotainment system refers to a combination of in-vehicle radio, navigation system, hands-free calling device, driver assistance system, and other functions in a central operating unit. The term "infotainment" is a portmanteau of "information" and "entertainment." A touch-sensitive display unit 30, a computing device 40, an input unit 50, and a memory 60 are used to operate the infotainment system. The display unit 30 includes a display surface for displaying variable graphic information and an operating surface (touch-sensitive layer) arranged above the display surface for user input commands. This operating surface can be designed as an LCD touchscreen display.

[0072] Screen 30 can be easily viewed and operated by both the driver and the front passenger of vehicle 10. Furthermore, mechanical operating elements, such as buttons, adjustment knobs, or combinations thereof, such as push-button adjustment knobs, can be arranged in the input unit 50 below screen 30. Steering wheel operation of the infotainment system is also typically possible. This unit is not shown separately but is considered part of the input unit 50.

[0073] Display unit 30 is connected to computing device 40 via data line 70. The data line can be designed according to the LVDS standard, which corresponds to Low Voltage Differential Signalling. Display unit 30 receives control data from computing device 40 via data line 70 for controlling the display surface of touchscreen 30. Control data for input commands is also transmitted from touchscreen 30 to computing device 40 via data line 70. Input unit 50 is connected to computing unit 40 via data line 90.

[0074] The memory device 60 is connected to the computing device 40 via data line 80. The table catalog and / or symbol catalog are stored in the memory 60 along with the tables and / or symbols for any additional information that may be inserted.

[0075] Additionally, the projection unit 20 is connected to the computing device 40 via a data line 70. The projection unit can be designed as a head-up display and used to fade augmented reality information into the driver's field of vision.

[0076] Other parts of the infotainment system, including camera 150, radio 140, navigation device 130, telephone 120, and instrument cluster 110, are connected to the devices used to operate the infotainment system via data bus 100. A high-speed variant of the CAN bus according to ISO standard 11898-2 can be used as the data bus 100. Alternatively, an Ethernet-based bus system, such as BroadR-Reach, can also be considered. A bus system in which data transmission is achieved via optical waveguides can also be used. Examples mentioned include the MOST bus (Media Oriented System Transport) or the D2B bus (Domestic Digital Bus). Vehicle measurement unit 170 is also connected to data bus 100. This vehicle measurement unit 170 is used to acquire vehicle motion, particularly vehicle acceleration. This vehicle measurement unit can be designed as a conventional IMU unit, corresponding to an Inertial Measurement Unit. An accelerometer and a rotational rate sensor, such as a laser gyroscope or a magnetometer gyroscope, are typically present in the IMU unit. The vehicle measurement unit 170 can be considered as part of the odometer of the vehicle 10. But it also includes wheel speed sensors.

[0077] It is also mentioned here that camera 150 can be designed as a conventional video camera. In this case, the video camera records 25 full frames per second, which corresponds to 50 half frames per second in interlaced recording mode. Alternatively, a specific camera that records more frames per second can be used to improve the accuracy of object recognition in the case of fast-moving objects, or to record light in the spectrum outside the visible spectrum. Multiple cameras can be used for environmental observation. In addition, radar or lidar (LIDAR) sensors 152 and 154 are supplementarily or alternatively used to perform or broaden environmental observation. As already mentioned, vehicle 10 is equipped with a communication module 160 for inward and outward wireless communication.

[0078] Reference numeral 181 indicates engine control equipment. Reference numeral 182 corresponds to ESP control equipment, and reference numeral 183 indicates transmission control equipment. Other control equipment may be present in the vehicle, such as additional dynamics control equipment (for vehicles with electrically adjustable dampers), airbag control equipment, etc. The connection of all such control equipment categorized under the powertrain system is typically achieved using a CAN bus system (Controller Area Network) 104, which is standardized to the ISO standard, typically ISO 11898-1. For different sensors 171 to 173 in the motor vehicle that are no longer connected to individual control devices, it is also specified that they are connected to the bus system 104 and their sensor data is transmitted to the respective control devices via the bus. Examples of sensors in a motor vehicle include wheel speed sensors, steering angle sensors, acceleration sensors, rotational speed sensors, tire pressure sensors, distance sensors, knock sensors, air quality sensors, etc. Wheel speed sensors and steering angle sensors are particularly included in the vehicle's odometer. Acceleration sensors and rotational speed sensors can also be directly connected to the vehicle measurement unit 170.

[0079] Reference numeral 144 indicates an On-Board Diagnostic (OBD) interface. This OBD interface is used to connect diagnostic equipment while the vehicle is in the workshop. Fault data stored in the vehicle can then be read. The OBD interface 144 is connected to a gateway 142 via a communication bus 102. The gateway 142 is used to facilitate the exchange of information from different communication branches. For this purpose, the gateway performs format conversion, such as converting messages received by the gateway in the format of bus system 100 to the format of bus system 104. This also occurs accordingly in the forwarding between other bus systems of the on-board electronics.

[0080] Reference numeral 180 indicates another control device responsible for autonomous driving functions or one or more driver assistance systems. This allows for different driving speed levels with respect to autonomous driving. A corresponding high-performance computing unit must be incorporated into the control device 180 to enable the implementation of the desired driving speed levels individually.

[0081] The following considers the scenario where sensor 171, connected to CAN bus 104, acquires measurement values ​​and transmits them to gateway 142. The measured values ​​are stored in the memory of gateway 142. An example of such a measurement sequence is shown below. Figure 6The results are presented in tabular form. However, this is a hypothetical measurement sequence used to illustrate the principle of the compression method. The top row contains the index of the corresponding measurement value. Measurements are performed at equidistant time intervals of 10 ms. The second row contains the time specification t(x) for when the corresponding measurement value was acquired. The measurement sequence can begin based on a specific event. Measurement acquisition occurs within the range of 0 to 80 ms after the event occurs. The fourth row contains the measurement value fp(x) in decimal form with two decimal places. These numbers are transmitted as integers from sensor 171 to gateway 142 via CAN bus 104. The measurement values ​​are in the range of 0.0 to 0.08.

[0082] Figure 7 It is shown in a graphic representation Figure 6 The measured values ​​are shown in the table. Here, the corresponding measurement time points are plotted along the x-axis of the image in the range of 0 to 80 ms. The measured value fp(x) is plotted along the y-axis in the range of 0 to 0.08. Measurement points are indicated by a cross symbol. The dashed lines connecting the measurement points correspond to the step function.

[0083] The novel compression algorithm that enables data compression is processed by the computing unit in gateway 142. This algorithm is based on rounding measurements using bit shifting. This operation is also the name of the algorithm, which is simply called the BISCO algorithm, corresponding to "Binary Shift Compression".

[0084] If necessary, the measured values ​​are pre-converted to integers. In the case of signed numbers, the measured values ​​can be converted to "signed" integer format. In the case of unsigned measured values, the numbers can be converted to "unsigned" integers. This conversion can also be performed in gateway 142.

[0085] Many sensors already provide measurements in integer format, so in the case of this type of sensor, the step of converting to integers can be omitted. The compression algorithm operates on integers.

[0086] The BISCO algorithm is described here using the syntax of the C++ programming language and consists of 6 steps.

[0087] 1. set index x = 0

[0088] 2.calculate round information “rnd(x)”:rnd(x)=int(x)>>(nbit-1)&0x1

[0089] 3.calculate compressed value “comp(x)” with lower accuracy-“nbit” rightshift:comp(x)=(int(x)>>nbit)+rnd(x)

[0090] 4.if "comp(x)" is EQUAL "last" AND x>0:go to 6

[0091] 5.remember index “x” for transmission.Set “last” to “comp(x)”.

[0092] 6.Increase x=x+1,go to step 2and continue with next time“t(x+1)”andvalue“int(x+1)”

[0093] In step 1, the measured value index x is set to the initial value 0. In step 2, the rounding information rnd(x) is calculated, which should be inserted at bit position nbit-1 of the integer int(x). This rounding information rnd(x) is calculated by shifting the integer int(x) to the right by nbit-1 positions and performing a logical AND operation with the binary value "1". Here, by choosing the parameter nbit, the degree of rounding can be specified. This is equivalent to choosing the degree of compression for a compression algorithm. If nbit = 0 is chosen, it means that lossless compression should be performed. In step 3, the compressed value is calculated. Here, the integer int(x) is shifted to the right by nbit and added to the calculated rounding information rnd(x). The rounding operation process and the resulting rounded value comp(x) are described in... Figure 8 As shown in the table. Figure 8 In the case shown, nbit = 2. In step 2, the integer is shifted one position to the right (nbit - 1 = 1). This position is... Figure 8 Add borders to the table and represent it as rnd. The rounding information value rnd(x) is generated by an AND operation, as shown in... Figure 8As shown in the third column of the table. In step 3, the entry for the integer shifted two bits to the right is added to the rounding information rnd(x). The result comp(x) is given in decimal form in the fourth column of the table. In step 4, the element decimation of the measurement sequence is performed. Here, data elements whose rounded data comp(x) is indistinguishable from the directly preceding rounded data comp(x-1) are removed. Supporting values ​​here are: decimal 0, decimal 1, and decimal 2. Thus, these are the values ​​in the table where the rounded value comp(x) is distinguishable from the preceding value comp(x-1).

[0094] The compression algorithm's processes in steps 4 through 6 are in another... Figure 9 The table shows the supported values, which are given in the fifth row, "last". Only these values ​​are saved in the compressed data set and are either stored in vehicle 10 or forwarded to the onboard communication module 160 and from there transmitted to the backend server 320. The sixth row, "Tx", is marked with an 'x', indicating which data elements of the time series are saved and transmitted. The compressed time series is... Figure 10 The table shows this. However, the fully compressed data set for the compressed time series only contains rows t(c) and comp(c). For comparison, a floating-point number derived from the rounded comp(x) value of the data element is also entered in row fp(c). The calculation of the floating-point number fp(c) is performed based on comp(c) using the factor: fc = 2^nbit * f, which corrects for n-bit shifts. Here, the factor fc is derived from the factor f mentioned above and the selected n-bit description. When transmitting the compressed data set, if metadata has not yet been derived from other information for the application, metadata may also be optionally transmitted. To convert integers to floating-point numbers, information about the factor f or fc, the offset value o, and the n-bit description may be optionally transmitted together, and if necessary, the physical unit in which the sensor 171 acquired the value may also be transmitted together.

[0095] Compression is therefore achieved through a rounding step combined with a extraction step. Only the supporting values ​​are retained as a remainder in the time series.

[0096] Figure 11 A graphical representation of the values ​​retained in the compressed time series is shown. The retained support values ​​are represented by circular symbols.

[0097] The comparison between the original measured values ​​and the support values ​​of the compressed representation is shown in Figure 12 The original measurements are shown in the image. Figure 12 This is referred to as the "original signal" (RS) and is represented by a cross symbol. From this, it can be clearly seen that the retained support values... Figure 12The term "compressed signal" (CS) does not always correspond to the original measurement value, but rather represents a rounded value that may lie between the original measurement values.

[0098] The following section explains how data analysis can be performed on datasets compressed in this way without prior decompression. For this purpose, a box plot analysis example will be used.

[0099] First, the results of the box plot analysis, taking the original measurements as an example, are as follows: Figure 13 As shown in the diagram. As previously mentioned, boxplot analysis aims to illustrate the two extremes, median, and two quartiles Q25 and Q75 of the measurement sequence. All five values ​​are within... Figure 13 The description is in column 3 of the table.

[0100] exist Figure 14 The image shows the five results of the boxplot analysis [BRPS] (Box Plot Raw Signal) in graphical form. The characteristic boxplot analysis values ​​are shown in... Figure 14 It is identified by a diamond symbol.

[0101] Figure 15 The box plot analysis is illustrated in tabular form as follows: How does it work on a compressed dataset? Importantly, the time difference dt to the next data element is considered, which can be easily calculated from the measurement time points recorded together in the compressed dataset. Therefore, in the box plot analysis, the difference time is first calculated and entered together into the data elements. Then, the data elements are sorted according to the ascending size of the compressed integer comp(c). The minimum value is then located in the first row of the time series sorted in this way. Next, it is analyzed which value lies at 25% of the relative time (lower quartile Q25). The total time covered by the measured value corresponds to 80 ms. This is derived by summing the values ​​in column dt. Therefore, the Q25 value lies at 20 ms in the first interval between the supporting values ​​"0" and "1", because according to... Figure 14 In the table, the value "0" has a duration of 30ms. Therefore, the Q25 value is also entered for the value comp(c) = 0. The median is the value that is exactly located at the "midpoint" of the relative time when the measured values ​​are sorted by size, that is, at 0.5 of the relative time. Therefore, the median corresponds to the value "1", see [link to relevant documentation]. Figure 15 The table below illustrates this. The determination of the Q75 value is in principle exactly the same as that of the Q25 value. Therefore, the Q75 value in the example also corresponds to the value "1" because it is positioned at 75% of the total time represented (corresponding to 60ms). This is still at the value "1" because the maximum value "2" is reached after 70ms, but this corresponds to 87.5% of the total time of 80ms. Therefore, the 75% value is still "1".

[0102] Figure 16 The results of boxplot analysis [BPCS] (Box Plot Compressed Signal) on the compressed dataset are also shown. The sorting and searching overhead for boxplot analysis of the compressed dataset is correspondingly reduced. Thousands of measurements can be easily recorded for the actual measurement sequence, significantly reducing the overhead for analyzing the original dataset and thus significantly reducing the overhead for sorting and searching within the compressed dataset.

[0103] Other analytical methods can also be performed on the compressed dataset. Examples include similarity analysis, pattern recognition, and trend analysis. This also includes simple algorithms for preparing information for graphical representations, such as bar charts and pie charts. Another example is histogram representation, which considers, for instance, the relative durations of different categories.

[0104] If the analysis method cannot be used with the compressed integer, a conversion to floating-point numbers can first be performed according to the formula fp(c) = comp(c) * fc + 0, as described above. If the analysis method requires values ​​that are equidistant in time and / or the values ​​of all necessary signals at each time point t, this representation can be generated from the compressed representation (decompression). For this purpose, it is preferable to determine the last value that is valid at the desired time point for each signal. In principle, these values ​​can also be calculated using other algorithms (e.g., interpolation, linear or nonlinear regression, based on a cubic spline function, etc.).

[0105] Data analysis can also be performed by gateway 142. If a significant deviation from the target value is identified, gateway 142 can send a message to computing unit 40, which then generates a warning message, which is displayed to the driver via display unit 30 and / or 20. Alternatively, data analysis can be performed in backend server 320, where compressed data sets are transmitted.

[0106] All examples and conditional statements mentioned herein should be understood as not being limited to the specific examples cited. Thus, those skilled in the art will recognize, for example, that the block diagrams shown herein represent conceptual views of exemplary circuit arrangements. Similarly, it can be seen that the flowcharts, state transition diagrams, pseudocode, and the like shown are different variations used to represent processes substantially stored on a computer-readable medium and therefore implementable by a computer or processor.

[0107] It should be understood that the proposed methods and related devices can be implemented in various forms of hardware, software, firmware, dedicated processors, or combinations thereof. Dedicated processors may include application-specific integrated circuits (ASICs), reduced instruction set computers (RISCs), and / or field-programmable gate arrays (FPGAs). Preferably, the proposed methods and devices are implemented as a combination of hardware and software. The software is preferably installed as an application program on a program memory device. Typically, this relates to a computer platform-based machine having hardware such as one or more central processing units (CPUs), random access memory (RAM), and one or more input / output (I / O) interfaces. An operating system is also typically installed on the computer platform. The various processes and functions described herein may be part of an application program or part of an implementation via an operating system.

[0108] This disclosure is not limited to the embodiments described herein. There is room for various adjustments and modifications that those skilled in the art would consider to be part of this disclosure.

[0109] List of reference numerals

[0110] 10 vehicles

[0111] 20 Head-up Display

[0112] 30 Touch-sensitive display units

[0113] 40 computing units

[0114] 50 input units

[0115] 60 memory units

[0116] 70 Data lines of the display unit

[0117] 80 Data lines of memory unit

[0118] 90 Input Unit Data Lines

[0119] 100 data bus

[0120] 102 Diagnostic Bus

[0121] 104 CAN bus

[0122] 106 Ethernet bus

[0123] 110 Combination Instrument

[0124] 120 phone number

[0125] 130 Navigation Device

[0126] 140 Radio

[0127] 142 Gateway

[0128] 144 On-board diagnostic interface

[0129] 150 cameras

[0130] 160 Communication Module

[0131] 170 Vehicle Measurement Unit

[0132] 171 Sensor 1

[0133] 172 Sensor 2

[0134] 173 Sensor 3

[0135] 180 Control unit for autonomous driving function

[0136] 181 Engine Control Equipment

[0137] 182 ESP control device

[0138] 183 Transmission control equipment

[0139] 200 Group Core Evolution

[0140] 300 Internet

[0141] 310 Roadside Unit

[0142] 320 Backend Server

[0143] BCDA Data Analysis of BISCO Compressed Data Sets

[0144] DP(0)-DP(n-1) data elements

[0145] DS Data Group

[0146] int(0) - int(n-1) integers

[0147] P10 Process steps in the vehicle

[0148] Process steps in the P320 backend server

[0149] S10 stores the raw data group

[0150] S12 Read raw data group

[0151] S14 executes the compression algorithm

[0152] S15 executes the BISCO algorithm

[0153] S16 forwards compressed data groups

[0154] S18 performs box plot analysis on the raw data.

[0155] S20 performs interruption / drift analysis on the raw data.

[0156] S22 outputs a warning message

[0157] S24 transmits the compressed data set to the cloud.

[0158] S26 Archives the compressed data set.

[0159] S28 executes the decompression algorithm

[0160] S30 forwarding reconstruction data group

[0161] S32 archives the reconstructed data set.

[0162] S34 Box Plot Analysis

[0163] S36 Execution Interruption / Drift Analysis

[0164] S38 generates a warning message and transmits it to the vehicle.

[0165] S40 performs similarity analysis

[0166] S42 Execution Mode Recognition

[0167] S44 performs other analysis methods on the reconstructed data.

[0168] S46 executes the decompression algorithm

[0169] S48 forwards the reconstructed data set to another analysis method.

[0170] t(0)-t(n-1) time points

[0171] T200 transmits data packets via mobile radio.

Claims

1. A method for organizing a data set containing at least one time series using a computing unit, wherein, The time series corresponds to a measurement value sequence, wherein the time series includes data collected at a specific time point and the corresponding time point of data collection as data elements. The feature is that it involves performing steps of rounding the collected data and subsequently extracting the data elements contained in the at least one time series from the data group, and storing the extracted time series of the data group in the computing unit and / or transmitting it to an external computing device. The extraction step includes: checking the changes in the rounded data of the time series data elements, and removing data elements whose rounded values ​​are equal to the rounded values ​​of their directly adjacent preceding data elements in time. Specifically, the measured values ​​are collected either as integers or as floating-point numbers and then converted to integers. Specifically, for the data elements of the time series, the floating-point numbers are converted to integers, the offset "o" is subtracted, and the result of the subtraction is divided by the factor "f", where f corresponds to the multiplication factor and o corresponds to the offset value.

2. The method according to claim 1, wherein, In order to prepare for the extraction of data elements from the time series, an algorithm based on rounding integers is executed.

3. The method according to claim 2, wherein, Prepare for rounding operations according to the following rules: ,in, This corresponds to the rounding information of the integer chosen for subsequent rounding operations, where x corresponds to the data element index, where, Corresponding to the integer, where, It specifies which bit position of the integer should be selected to calculate the rounding information according to the mentioned rules. Where AND corresponds to a logical AND operation between an entry and the integer value "1", the entry being located by right-shifting the integer. of The bit position determined by the bit position, where BSR represents the bit scan inversion function.

4. The method according to claim 3, wherein, The rounding operation is performed according to the following rules: ,in, Corresponding to the integer produced by rounding The portion with reduced resolution, of which, Corresponding to the rounding information The instruction specifies how many bits the integers should be shifted to the right before addition.

5. The method according to claim 4, wherein, For the data element extraction step of the time series, the rounded values ​​are... and The values ​​are compared with each other, and subsequent identical values ​​are discarded from the time series. .

6. The method according to any one of claims 1 to 5, characterized in that, Perform analysis on the extracted time series of the data set without decompressing the data set.

7. The method according to claim 6, wherein, Perform box plot analysis on the compressed dataset, in the manner of reducing resolution. The information is sorted by ascending order of size, and the eigenvalues ​​of the box plot analysis are determined from the compressed data set, taking into account the time difference between each data element and the next.

8. The method according to claim 7, wherein, The eigenvalues ​​analyzed by the box plot are the minimum, maximum, median, lower quartile, and upper quartile.

9. An apparatus for performing the method according to any one of the preceding claims, comprising at least one computing unit configured to perform measurement value acquisition or receive measurement values ​​acquired by at least one connected sensor at successive time points, characterized in that, The computing unit is designed to form a time series using the measured values ​​and associated measurement time points, and the computing unit is designed to perform at least the steps of rounding and subsequently extracting data elements contained in at least one time series from the data group, and to store the extracted time series of the data group in the computing unit and / or transmit it to an external computing device.

10. The device according to claim 9, wherein, The computing device or other computing device is designed to perform box plot analysis on the compressed data set, in a manner that reduces the resolution. The information is sorted by ascending order of size, and the eigenvalues ​​for the box plot analysis are determined from the compressed data set, taking into account the time difference between each data element and the next.

11. The device according to claim 10, wherein, The eigenvalues ​​analyzed by the box plot are the minimum, maximum, median, lower quartile, and upper quartile.

12. A vehicle, characterized in that, The vehicle (10) has the equipment according to any one of claims 9 to 11.

13. An apparatus for performing the method according to claim 6, comprising at least one computing device configured to receive compressed data sets, characterized in that, The computing device is designed to perform analysis on the time series of extracted data sets without having to decompress the data sets.

14. The device according to claim 13, wherein, The at least one computing device is designed to perform box plot analysis on the compressed data set, in a manner that reduces resolution. The information is sorted by ascending order of size, and the eigenvalues ​​for the box plot analysis are determined from the compressed data set, taking into account the time difference between each data element and the next.

15. The device according to claim 14, wherein, The eigenvalues ​​analyzed by the box plot are the minimum, maximum, median, lower quartile, and upper quartile.

16. A computer program, characterized in that, The computer program is designed to execute, when run in a computing unit, the steps of the method for organizing a data set containing at least one time series according to any one of claims 1 to 8.

Images