Procedure for operating a control unit, control unit and vehicle

Abstract

Method for operating a control device (CD), in particular a control device (CD) for detecting living beings or objects in an area, wherein the control device (CD) has a processor (1) and a working memory (3), a first interface (7) for receiving measured values ​​(x_i) and a second interface (9) for outputting a result (E), wherein the method comprises the following steps: - providing a respective measured value (x_i) and a decision value (β_i) to the working memory (3), wherein, based on the respective decision value (β_i) and the measured value (x_i), a partial result (p_i) is calculated using the processor and the partial result (p_i) is provided to the working memory (3), wherein the working memory (3) is configured to receive the respective decision value (β_i), the respective partial result (p_i) and the respective measured value (x_i);Whereby for i=1 to i=N the following step is performed:- Calculating a further partial result (p_i+1) from a further measured value (x_i+1), the further decision value (β_i+1) and the partial result (p_i) from the respective preceding step (i), Whereby after i=N steps the partial result (p_N) is provided as the result (E) to the second interface (9).;

Description

The invention relates to a method for operating a control device and to a control device. Furthermore, the invention relates to a vehicle. Control devices are ubiquitous in modern vehicles and serve to control and / or regulate almost all functions of a vehicle. Despite the use of fast microprocessors, computing capacity in control systems is often insufficient or expensive. Therefore, the purpose of the invention is to simplify the operation of a control device and thus save computing power and working memory. The problem is solved by a method according to claim 1. Furthermore, the problem is solved by a control device according to claim 10 and by a vehicle with such a control device. Advantageous designs and further training are subject to dependent claims. The method serves to operate a control device, in particular a control device for detecting living beings or objects in a given area. The control device has a processor and a working memory, a first interface for acquiring measured values ​​and a second interface for outputting a result, wherein the method, in particular an evaluation of the measured values ​​by the control device, comprises the following steps: - Provision of a respective measured value and a decision value to the working memory, wherein, based on the respective decision value and the measured value, a partial result is calculated with the aid of the processor and the partial result is provided to the working memory, wherein the working memory is configured to receive a decision value, the respective partial result and the respective measured value;Where, for i=1 to i=N, the following step is performed: - Calculating a further partial result from another measured value, the further decision value, and the partial result from the preceding step. Where, after i=N steps, the respective partial result is provided to the interface for outputting the result. In the procedure described here, i denotes a running variable that runs from 1 to N and is incremented by one at each step. The measured value is advantageously provided by a sensor. The respective i-th partial result advantageously approximates the corresponding N-th partial result. Furthermore, a probability can be calculated from the N-th partial result, where the probability indicates, for example, whether a living being is located in the area. The processor is advantageously designed as a microprocessor, which is preferably connected to the main memory via a bus system. Alternatively, the main memory can also be integrated into the processor. The respective interface can also be designed as a common interface, suitable for both outputting the result and receiving the measured values. Advantageously, multiple measured values ​​and multiple decision values ​​can also be stored in the working memory. The invention takes into account that modern evaluation routines perform a large number of calculations, and the data generated or required in these operations can occupy a significant portion of main memory. To minimize the amount of main memory required, the invention proposes making the respective calculations executable serially and, furthermore, eliminating the need to "retain" (i.e., store) unused values ​​in main memory. Advantageously, a decision value (β_i) is a number or a vector that can be provided using a machine learning algorithm. The decision value advantageously serves to transform the respective measured value (x_i) into a contribution to the result in the form of a partial result, such that the product of the respective measured value and the respective decision value yields a partial result. The partial results thus contribute to the overall result. The result is advantageously a number indicating whether the measured value corresponds to a given situation or not. In other words, the result can be a probability of whether a living organism is detectable in a given area. The result (as the Nth partial result) is advantageously the sum of the partial results p_1 to p_N-1. The result can be transformed into a probability using, for example, a (logistic) function or another sigmoid function. The result can indicate whether or not a situation exists based on the measured values. For example, sensor readings can detect whether an object or person is located in a certain area. An advantageous application of the invention is the detection of whether a person, in particular a small child, is in the passenger compartment of a vehicle, preferably in the passenger compartment of an automobile. Advantageously, it is possible to determine what type of living being is present. The result can advantageously be evaluated to determine whether a small child, an adult, or a pet has been detected by means of a sensor. A measured value is advantageously provided by a sensor and is advantageously a measurement at a specific point in time. Advantageously, a measured value can also be an n-tuple of individual measured values. Advantageously, the measured value can also correspond to a complex quantity such as the sum over the time-dependent changes of a signal. The invention described here advantageously reduces the utilization of working memory and allows it to be designed much smaller. This saves considerable effort and costs in the control unit. In an advantageous embodiment of the invention, each subsequent partial result is calculated by multiplying the measured value by the respective decision value, and the product of this value is then added to the respective partial result from the preceding step. Advantageously, the respective decision value is assigned to the respective measured value. Using decision values ​​is particularly advantageous for easily calculating the respective partial results. This simplified calculation significantly reduces the computational effort and thus the processor load compared to calculations using a neural network. In a further advantageous embodiment of the invention, the decision values ​​(β_i ; i=1,...,N) are determined using a learning algorithm. The decision values ​​are advantageously values ​​from a decision matrix. Alternatively, the decision values ​​β_i can be the parameters of a function, in particular a linear function: The decision matrix is ​​advantageously used to calculate the result from the measured values ​​x_i. By providing only a few numbers / vectors to the main memory and / or the processor, the control unit can be designed to be significantly less powerful or to be operated in a much more energy-efficient manner. Therefore, the invention allows for a reduction in storage space and computing power in the control unit. Using the method described here, up to 99% of storage space and a significant portion of computing power can be saved compared to conventional methods according to the current state of the art. In a further advantageous embodiment of the invention, the method serves to evaluate measured values ​​from at least one sensor, in particular an ultra-wide-band sensor. The sensor is advantageously configured as a UWB receiver and / or a UWB transmitter. The control unit is advantageously used to evaluate the output of the UWB sensor or multiple UWB sensors. The method is advantageously used to detect people in a given area. By evaluating the measured values ​​of one (or more) UWB sensors, the otherwise very complex evaluation can be carried out in a resource-saving manner. In a further advantageous embodiment of the invention, after the calculation of the next partial result, the respective measured value and the respective decision value are deleted from the main memory or from a cache of the processor, in particular from the preceding (calculation) steps. Optionally, the respective partial result that is no longer needed can also be deleted from the main memory after the calculation of the next partial result. By deleting the no-longer-needed values ​​from RAM, the amount of RAM required for the method described here can be significantly reduced. The method presented here can advantageously reduce the required RAM area to one-hundredth of its original size. In a further advantageous embodiment of the invention, the control unit has an additional memory, wherein, at each step, the respective decision value is provided from this additional memory to the main memory. The provision of the decision values ​​occurs either directly from the additional memory to the main memory or via the processor, which advantageously provides the measured values ​​and / or decision values ​​necessary for the respective step to the main memory. The additional memory is advantageously designed as non-volatile memory such as a hard disk or a ROM chip. This additional memory is advantageously used to store decision values ​​and optionally as a buffer for measured values. It is advantageous for the additional memory to be implemented as ROM memory. The additional memory can either be integrated into the control unit or connected to the control unit via an interface. In a further advantageous embodiment of the invention, the respective decision value β_i (i=1,...N) is stored in the additional memory in the form of a decision matrix or a so-called decision vector. The decision matrix is ​​advantageously provided by a machine learning algorithm. A particularly simple form of the decision matrix can be implemented as a decision vector β = (β1,...βN), βi∈ ℝ, provided that the decision matrix can be diagonalized. The method also allows for an evaluation of the measured values ​​based on the respective decision value, without requiring the resources previously necessary for this purpose. In a further advantageous embodiment of the invention, the calculation is carried out in such a way that in one step a plurality of measured values ​​and a plurality of decision values ​​are provided to the working memory and a (further) partial result is provided accordingly for the corresponding plurality of measured values ​​and decision values. Advantageously, such a partial result can be calculated by summing the respective products of the respective measured value x_i and the corresponding decision value β_i. In other words, a partial result is calculated as follows: By calculating such a partial result in one step, the computation time is significantly reduced because the transfer of each partial result to memory does not need to be performed individually for each i=1 to N (or k=1,...,i). Advantageously, values ​​of 3 to 10, especially 4 to 7, and preferably i = 5, have proven beneficial for k. Furthermore, depending on the processor design, the steps can be executed in parallel using a multi-core processor. In a further advantageous embodiment of the invention, the presence or position of a living being in an area, in particular in the passenger compartment of a vehicle, is determined with the aid of the control device. The procedure or control device is advantageously used to determine whether a small child or a pet, e.g., a dog, has been "forgotten" in a car. A small child or a pet can therefore be understood as a living being. The area in question can also be the rear seat of a car. Such a so-called "child presence detection" can be carried out particularly easily and in a resource-saving manner with the help of the invention. Description of the control unit The control unit comprises a processor and a working memory, and optionally an additional memory, as well as a first interface for recording measured values ​​and a second interface for outputting a result, wherein the control unit is configured and designed to execute a procedure according to the above description. The control unit is advantageously very simple in design and, in particular due to its low memory requirements, can be provided cost-effectively. The use of such a control unit in a vehicle is advantageous. Especially in vehicle-specific control units, the working memory is limited due to cost considerations and operational reliability. Therefore, the use of a dedicated vehicle control unit is particularly well-suited for the aforementioned application. The vehicle is advantageously designed with such a control unit as either an automobile or a rail vehicle. It is advantageous for the vehicle with the control unit to be used for passenger transport. The invention is further described and explained below with reference to figures. Figure 1 shows an exemplary control device, and Figure 2 shows a possible process diagram. Fig. 1 shows an exemplary control unit SE: The control unit SE comprises a processor 1, a main memory 3, and a further memory 5. Furthermore, the control unit SE comprises an interface 7 for acquiring measured values ​​x_i and an interface 9 for outputting a result E. The additional memory 5 comprises a region containing the decision values ​​β_1 to β_N. The decision values ​​β_i are advantageously stored in the additional memory 5 in the form of a decision matrix. The additional memory 5 is advantageously implemented as non-volatile memory, e.g., a ROM chip. The decision values ​​β_i have been advantageously provided using a learning algorithm and are advantageously provided using one of the interfaces 7, 9 for the additional memory 5 or a main memory 3. The control unit SE is designed and configured to calculate partial results p_i from the respective measured values ​​x_i using the decision values ​​β_i. The result E can be calculated by adding the partial results p_i. Processor 1 and main memory 3 are used for this purpose. The control unit SE performs the following steps: - In a first step, an initial measured value x_1 is provided to processor 1. The first measured value x_1 is provided either directly from interface 7 or from main memory 3. A first decision value β_1 is then provided to processor 1. The respective decision value β_i is preferably stored in the additional memory 5 and is advantageously transferred from the additional memory 5 to main memory 3 and / or to processor 1. In the second part of the first step i (steps are numbered i here), processor 1 multiplies the first decision value β_1 by the first measured value x_1. The result of the multiplication is a partial result p_i, which is provided to and stored in main memory 3. In each subsequent step (i=2,...,N), the partial result p_i from the preceding step is added to the product of the respective further decision value β_i+1 and the respective further measured value x_i+1.In this way, a further partial result p_i+1 is determined in each case. This step is performed for i = 2 to N, where the number of N is determined by the number of available measured values ​​x_i or the number of available decision values ​​β_i. In a final step, the last calculated partial result p_N is advantageously converted into a probability, for example by substituting it into a logistic function, and / or the previous partial result p_N is provided as result E to the second interface 9. Advantageously, after calculating each (further) partial result p_i+1, the values ​​p_1, x_i, and β_i from the previous calculation are deleted from the working memory of the respective step i. Therefore, for each calculation, only three values ​​are stored in the working memory 3: the respective measured value x_i, the respective decision value β_i, and the respective partial result p_i (and optionally the further partial result p_i+1). This allows the working memory 3 to be chosen to be particularly small and inexpensive. Fig. 2 shows a possible procedure. The calculation is shown, where each measured value x_i is assigned a partial result p_i, where for all i = 1,...N, the following holds: p_i+1 = p_i + (β_i * x_i). The final partial result p_N (for i=N) corresponds to the result p_N = E. Advantageously, the final result can be converted into a probability using a logistic function. In an advantageous application of the invention, the respective measured value x_i is a measured value of a sensor, in particular a UWB sensor. Advantageously, in each step i, a measured value x_i and a corresponding development value β_i are transferred to the working memory 3 or to the processor 1. The embodiment shown here demonstrates a particularly simple method for operating a control device SE, especially for detecting small children in a passenger compartment of a vehicle. In summary, the invention relates to a method for operating a control unit SE, a control unit SE configured for this purpose, and a vehicle with such a control unit SE. The method serves to evaluate measured values ​​x_i, wherein the measured values ​​x_i can be measurement curves of a UWB sensor. The evaluation advantageously serves to determine whether a living being is located in an area such as a passenger compartment. The method comprises steps i, which are advantageously executable sequentially, and in each step i, the respective measured value x_i is combined with a decision value β_i, in particular multiplied, and added to a partial result p_i from a preceding step i. The step i described above is advantageously performed N times sequentially. Optionally, a probability for the state of affairs to be determined, such as the probability of a living being being located in the area, is derived from the last partial result p_N.Advantageously, after each step, the data no longer needed is deleted from the main memory using a processor, so that the required area of ​​the main memory for carrying out the procedure is particularly small.

Claims

Method for operating a control device (CD), in particular a control device (CD) for detecting living beings or objects in an area, wherein the control device (CD) has a processor (1) and a working memory (3), a first interface (7) for receiving measured values ​​(x_i) and a second interface (9) for outputting a result (E), wherein the method comprises the following steps: - providing a respective measured value (x_i) and a decision value (β_i) to the working memory (3), wherein, based on the respective decision value (β_i) and the measured value (x_i), a partial result (p_i) is calculated using the processor and the partial result (p_i) is provided to the working memory (3), wherein the working memory (3) is configured to receive the respective decision value (β_i), the respective partial result (p_i) and the respective measured value (x_i);Whereby for i=1 to i=N the following step is performed:- Calculating a further partial result (p_i+1) from a further measured value (x_i+1), the further decision value (β_i+1) and the partial result (p_i) from the respective preceding step (i), Whereby after i=N steps the partial result (p_N) is provided as the result (E) to the second interface (9).; Method according to claim 1, wherein the calculation of the respective further partial result (p_i+1) is carried out by multiplying the measured value (x_i+1) with the respective decision value (β_i+1), the product being added to the respective partial result (p_i) from the preceding step (i). Method according to one of the preceding claims, wherein the decision values ​​(β_i ; i=1,...,N) are determined or have been determined using a learning algorithm. Method according to one of the preceding claims, wherein the method is for evaluating measured values ​​(x_i) from at least one sensor, in particular an ultra-wide-band sensor. Method according to one of the preceding claims, wherein after the calculation of the further partial result (p_i+1) the respective measured value (x_i), optionally the respective partial result (p_i) and the respective decision value (β_i) are removed from the working memory (3). Method according to one of the preceding claims, wherein the control device (SE) has a further memory (5), wherein in the respective step (i) the respective decision value (β_i) is provided from a further memory (5) to the working memory (3). Method according to claim 6, wherein the respective decision value (β_i) is stored in the further memory (5) in the form of a decision matrix. Method according to one of the preceding claims, wherein the presence or position of a living being in an area, in particular in the passenger compartment of a vehicle, is determined by means of the control device (SE). Method according to one of the preceding claims, wherein the calculation is performed in such a way that in one step a plurality of measured values ​​and a plurality of decision values ​​(β_i) are provided to the working memory (3) and accordingly a partial result (β_i) is provided for the corresponding plurality of measured values ​​(x_i) as well as a plurality of decision values ​​(β_i). Control device (SE) comprising a processor (1) and a working memory (3) and optionally a further memory (5), further comprising a first interface (7) for recording measured values ​​(x_i) and a second interface (9) for outputting a result (E), wherein the control device (SE) is configured and designed to execute a method according to one of the preceding claims. Vehicle, emanating a control device (CD) according to the preceding claim.

Images