Method for controlling a computing unit network and computing unit and computer program for its implementation

Abstract

The invention relates to a method for controlling a computing unit network comprising a plurality of functional units (1a, 1b, ..., 1n) and a decision unit (2), wherein each of the functional units (1a, 1b, ..., 1n) is configured to receive a plurality of requests, to determine, depending on the plurality of requests, one or more setpoint specifications on the basis of which the computing unit network is to be controlled, and to send the plurality of requests and the one or more setpoint specifications to the decision unit (2). In the method, the decision unit (2) receives (S100) a plurality of requests, a plurality of weighting factors, and a plurality of setpoint specifications from at least two of the plurality of functional units (1a, 1b, ..., 1n), wherein an associated weighting factor is received for each of the received requests.Subsequently, the decision unit (2) determines a plurality of measures depending on the plurality of target value specifications (S110). Subsequently, for each of the plurality of measures provided by the decision unit (2) and each of the plurality of requirements, a degree of fulfillment indicating a degree of fulfillment of the requirement by the measure is determined (S120). Using the weighting factors and the degrees of fulfillment, a measure to be implemented is determined from the plurality of measures (S130), and the computing unit network is controlled based on the measure to be implemented (S140).

Description

The present invention relates to a method for controlling a computing unit group and to a computing unit and a computer program for carrying it out.BACKGROUND OF THE INVENTIONArithmetic units or control units are partially entrusted with complex decisions which must be made on the basis of a plurality of decision criteria and measurement values. Especially technically challenging are decisions that must be made on the basis of conflicting setpoint values. In technical implementations, decision units that make the decisions are often referred to as a decision unit (arbitration unit).Such decisions can be made by software or software component running on the control device in so-called If-Then-Else structures. More complex decisions based on multiple decision parameters require complex decision trees or decision matrices.The creation of the decision matrices and decision trees is a very complex static process, which can usually only be carried out by specialists and function experts. In the design of the decision unit, it is assumed that the specialists are agreeable for each situation via the optimum solution option. In reality, however, the implementation is strongly dependent on the personal opinion of designers.Disclosure of the InventionAccording to the invention, a method for controlling a computing unit group and a computing unit and a computer program for carrying it out having the features of the independent patent claims are proposed. Advantageous embodiments are the subject matter of the dependent claims and of the following description.The invention is based on a computing unit group, for example for controlling a technical installation such as a vehicle, which has a multiplicity of functional units and a decision unit. Each of the functional units is configured to receive a plurality of requests, to determine one or more setpoint values on the basis of which the arithmetic unit group is to be controlled as a function of the plurality of requests, and to send the plurality of requests, weighting factors associated with the requests and the one or more setpoint values to the decision unit. The requests transmitted by different functional units can place requests for the same setpoint value, which are in particular conflicting.The requirements of the functional units can be defined, for example, across functional units from A1 to Ax, and the functional unit in this case sends only the indices with the associated weighting factors to the decision unit. It is also conceivable for the functional units to send the request in a standardized notation, for example ARXML format, and for the decision unit to collect requests of equal nature under an index A1 to An.The functional units are in particular setpoint requesters. The functional units can be hardware and / or software units. In the case of hardware units, these may be individual control units connected via communication lines, such as, for example, engine or vehicle control units (ECU, vehicle control unit, VCU), driver assistance systems (ADAS), electronic stability program (ESP), antilock system (ABS), measurement value transmitters, such as, for example, pedal value transmitters or temperature sensors, or software functions in the units mentioned.The essence of the invention is to specify a method for making decisions so that in situations in which different functional units send conflicting requirements or setpoint values to the decision unit, decisions can be made without the presence of a preprogrammed decision tree or decision matrices.For this purpose, the decision unit receives a plurality of requests, a plurality of weighting factors and a plurality of setpoint value specification from at least two of the plurality of functional units in the method, wherein an associated weighting factor is received for each of the received requests. Uniform setpoint values of different functional units are in particular combined. For example, the requirements and the associated weighting factors can be stored in a table. Weighting factors for a request that have not been sent by the functional unit but by other functional units are evaluated for the functional unit that has not sent the request, in particular with a value of 0 and optionally stored.Then, a plurality of actions is determined by the decision unit depending on the plurality of target value requests, and for each action of the plurality of actions and each of the requests, a degree of satisfaction indicating a degree of satisfaction of the request by the action is determined. To determine the degree of fulfilment, the measures are sent by the decision unit in particular to the functional units which determine the degree of fulfilment and send them back to the decision unit. Subsequently, a measure to be carried out is determined from the plurality of measures using the weighting factors and the degrees of fulfilment and the group of computing units is controlled on the basis of the measure to be carried out.Machine decisions, which are made, for example, by artificial intelligences, can be detected only with great difficulty or not at all in case of doubt. In contrast, by the method according to the invention, multidimensional decisions are made in the decision unit in a comprehensible manner on the basis of generally comprehensible principles via a quantitative decision method. This allows the decisions to be dynamically adapted in context-related fashion, wherein only one expert in the subfunctions and not by superordinate system experts is required for setting the function parameters. As a result, the method can be adapted in a simple manner and errors can be corrected in a short time with little effort.In one embodiment, a relative weighting factor is furthermore determined for each of the requests in order to determine a measure to be carried out. The relative weighting factor is determined by forming a quotient of the weighting factor of the request and a sum of all weighting factors of the requests of the same functional unit. This procedure can be represented in particular on the basis of the following equation: with the relative weighting factor W rel,n,i of the requirement n of the functional unit i, the weighting factor W n,i of the requirement n of the functional unit i and the sum of all weighting factors of the requirements 1 to m of the functional unit i. The measure to be carried out is then determined on the basis of the relative weighting factors and the degrees of fulfilment.The use of relative weighting factors advantageously makes it possible for a multiplicity of different requirements of one or more functional units and associated weighting factors to be able to be combined with one another in a meaningful manner.In one embodiment, the plurality of actions is determined by determining the plurality of target values as the plurality of actions, or by selecting the plurality of actions depending on the plurality of target values from among actions provided by the decision unit. In this case, predefined measures can be stored in a memory of the decision unit, which are selected by the decision unit as a function of the multiplicity of setpoint values specifications, or the decision unit can determine the multiplicity of measures as a function of the multiplicity of setpoint values specifications, i.e. determine optimized measures which fit the present decision situation in each situation.By using the plurality of setpoint values, the method can be carried out with little computing effort, as a result of which computing capacities for other measures become free.By determining the plurality of measures from measures provided by the decision unit as a function of the plurality of setpoint values, the measures used can be adapted to the present situation and therefore the control of the arithmetic unit group can be optimized.In one embodiment, in order to determine the measure to be carried out on the basis of the weighting factors or relative weighting factors and the degrees of fulfilment, an evaluation value is furthermore determined for each functional unit and each of the plurality of measures as a function of the determined (relative) weighting factor of the request and the degree of fulfilment of the request for the measure. For this purpose, a requirement evaluation value can be determined in particular for each functional unit, each requirement of the functional unit and each of the plurality of measures by multiplying the (relative) weighting factor of the requirement by the degree of fulfilment of the requirement. The request score may be represented by the following equation: the request score B n,i,j the request n of the entity i for the action j, the (relative) weighting factor W (rel,) n,i the request n of the entity i, and the satisfaction level EG n,i,j of the request n of the entity i for the action j.The evaluation value of the functional unit i for one of the plurality of measures is determined in particular as a sum over all requirement evaluation values B n,i,j of the requirements 1 to m of the functional unit i and can be specified by the following equation: with the evaluation value B i,j of the functional unit i of the measure j, the (relative) weighting factor W (rel,) n,i of the requirement n of the functional unit i and the degree of fulfilment EG n,i,j of the requirement n of the functional unit i for the measure j. The evaluation values can in this case be stored in particular in an i×j matrix.In one configuration, a useful value can then be determined for each of the plurality of measures as a function of the evaluation values for the measure, an expected value for each of the functional units, and an influencing factor for each of the functional units. The expected value specifies the threshold of the evaluation value from which the functional unit evaluates the performance of the measure as success for all requests sent to the functional unit. If the difference yields a negative value, the measure is considered to satisfy the requirements to an unsatisfactory extent. If, on the other hand, the difference has a positive value, the measure is considered to satisfy the requirements. The influencing factor specifies the importance of the respective functional unit or how strongly the evaluation of the functional unit is taken into account in the determination of the useful value in contrast to the other functional units. This means that system-critical functional units or functional units with safety functions have a higher influencing factor than functional units which merely fulfil comfort functions. In particular, the influencing factor should always have a value of less than 1.The determination of the useful value comprises, in particular for each evaluation value, the formation of a difference between the evaluation value of the functional unit for the measure and the expected value of the associated functional unit. The differences are then each potentiated with the influencing factor of the functional unit in order to obtain a functional-unit-specific individual value of use. The individual user values of the various functional units are subsequently multiplied with one another in order to obtain the user value for the measure to be carried out. Expressed by an equation, the useful value can be represented as follows: with the useful value NW j of the measure j, the evaluation value B i,j of the functional unit i for the measure j, the expected value E i of the functional unit i and the influencing factor p i of the functional unit i.In particular, the expected value can be fixed to the mean value of the value range that the weighting factors can assume. For example, when the weighting factors have values in a range of 0 to 10, the expected value E i for all the functional units may be set to a value of 5. A higher expected value can also be defined for system-critical or safety-relevant functional units. The influencing factor can conventionally be set to a fixed value of all functional units. The value of has been determined empirically as the value which can be used to best determine the measure that has the greatest benefit for all requirements received from the functional units. The influencing factor can also be determined, in particular, from the product of the constant value of a prioritization factor WF i for the functional unit i, which specifies the importance of the functional unit i. As a result, it is possible to define the weighting with which the difference between the evaluation value and the expected value of the respective functional unit flows into the overall result of the useful value. Thus, for example, functional units which are responsible only for comfort functions can have a lower prioritization factor than functional units which are responsible for critical functions. The prioritization factor of the functional units can assume, for example, a value between 0 (the functional unit is of no significance for the determination of the measure to be carried out and is intended to have no influence) to 1 (the functional unit is of great significance for the determination of the measure to be carried out and is intended to have a strong influence).Subsequently, in embodiments of the invention, a decision assessment is determined on the basis of the useful value for the measure to be carried out, which is in particular a complex number. In particular, the decision evaluation is determined as the sum of the real part and the imaginary part of the useful value. Represented by an equation, the decision evaluation of the measure to be carried out is as follows: with the decision evaluation EW j of the measure j to be carried out and the real part Re and the imaginary part Im of the useful value NW j of the measure j to be carried out.Subsequently, in embodiments of the invention, the measure to be carried out from the plurality of measures is determined on the basis of the decision evaluations of the plurality of measures. The measure to be carried out can be determined in particular as the measure with the highest decision evaluation.The above-mentioned advantages can be achieved in a particularly simple manner and with little computing effort by the described procedure.A computing unit according to the invention, e.g. the decision unit of a computing unit group, is configured, in particular by programming, to carry out a method according to the invention.The implementation of a method according to the invention in the form of a computer program or computer program product with program code for carrying out all method steps is also advantageous since this causes particularly low costs, in particular if an executing control device is also used for further tasks and is therefore present in any case. Finally, a machine-readable storage medium is provided with a computer program stored thereon, as described above. Suitable storage media or data carriers for providing the computer program are, in particular, magnetic, optical and electrical memories, such as hard disks, flash memories, EEPROMs, DVDs, among others. Download of a program via computer networks (Internet, intranet, etc.) is also possible. Such a download can be effected in a wired or wired or wireless manner (e.g. via a WLAN network, a 3G, 4G, 5G or 6G connection, etc.).Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawing.The invention is schematically illustrated in the drawing on the basis of an exemplary embodiment and is described below with reference to the drawing.Brief Description of the DrawingsFIG. 1 schematically shows the structure of a system having a multiplicity of functional units, a decision unit and an element of a computing unit group to be controlled, and FIG. 2 shows a flow chart of an embodiment of the method according to the invention.Embodiment of the InventionFIG. 1 schematically shows the structure of a system having a multiplicity of functional units 1 a, 1 b,..., 1 n, a decision unit 2 and an element 3 to be controlled of a computing unit group, for example in a vehicle. The system shown enables a method according to the invention to be carried out. FIG. 2 shows a flow chart of an embodiment of the method according to the invention. Both figures will be described together in the following.Each of the functional units 1 a, 1 b,..., 1 nis configured to receive one or more requests, to determine one or more setpoint values on the basis of which the arithmetic unit group is to be controlled as a function of the one or more requests, and to send the plurality of requests and the one or more setpoint values to the decision unit 2.The plurality of functional units 1 a, 1 b,..., 1 nand the decision unit 2 are connected to one another in a data-transmitting manner in such a way that the functional units 1 a, 1 b,..., 1 ncan transmit data to the decision unit 2 and receive it from the decision unit 2. The same applies to the decision unit 2 which can receive data from each of the plurality of functional units 1 a, 1 b,..., 1 nand transmit it to it. The transmitted data are, inter alia, requirements, setpoint values and weighting factors which are sent from the functional units 1 a, 1 b,..., 1 nto the decision unit 2, measures which are sent from the decision unit to the functional units 1 a, 1 b,..., 1 n, and degrees of fulfilment determined by the functional units 1 a, 1 b,..., 1 n, which are sent to the decision unit 2.In the method, the decision unit 2 receives requests, setpoint values presets from a plurality of, in particular at least two, of the plurality of functional units 1 a, 1 b,..., 1 nin step S 100. The requirements of different functional units 1 a, 1 b,..., 1 n relate, inter alia, to the same setpoint value or a plurality of the same setpoint values and, in particular, place conflicting requirements on the one setpoint value or the plurality of setpoint values. Furthermore, the functional units 1 a, 1 b,..., 1 ntransmit a weighting factor for each of the requests, which weighting factor is likewise received by the decision unit 2 in step S 100. Each of the weighting factors indicates which priority the request for the functional unit 1 a, 1 b,..., 1 nassociated with the weighting factor has.In step S 110, a plurality of measures are determined by the decision unit 2 depending on the plurality of target values. In this case, for example, the received plurality of setpoint values can be determined as a plurality of measures, the decision unit 2 combining setpoint values of the same nature of different or of the same functional unit 1 a, 1 b,..., 1 n. It is also conceivable that predefined measures are stored in a memory of the decision unit 2 and the decision unit 2 selects the plurality of measures from these predefined measures as a function of the plurality of setpoint values presets.In step S 120, a fulfillment degree indicating a degree of fulfillment of the request by the action is determined for each action of the transmitted plurality of actions and each of the requests, in particular by the functional units 1 a, 1 b,..., 1 n. To this end, the functional units 1 a, 1 b,..., 1 ncompare the request with each of the plurality of measures and determine the degree of fulfilment of the request by the measure that each functional unit 1 a, 1 b,..., 1 ntransmits subsequently again to the decision unit 2.In step S 130, a measure to be performed among the plurality of measures is determined by the decision unit 2 based on the weighting factors and the satisfaction degrees of each of the requests.For this purpose, in step S 131, a relative weighting factor is first determined for each of the requests, which is formed as the quotient of the weighting factor of the request (numerator) and the sum of the weighting factors of all requests (denominator) which were transmitted to the decision unit 2 by the same functional unit 1 a, 1 b,..., 1 n(cf. equation (1)). Subsequently, in step S 132, the measure to be carried out is determined on the basis of the relative weighting factors and the degrees of fulfilment.For this purpose, firstly in step S 132 a, an evaluation value is determined for each of the functional units 1 a, 1 b,..., 1 nand each of the plurality of measures as a function of the determined relative weighting factors of the requirements of the functional unit 1 a, 1 b,..., 1 nand the degrees of fulfilment of the requirements of the functional unit 1 a, 1 b,..., 1 ndetermined for the measures. For this purpose, first of all, requirement evaluation values for each of the requirements of the functional unit 1 a, 1 b,..., 1 ncan be determined on the basis of equation (2), i.e. as a product of the relative weighting factor of the requirement and of the degree of fulfilment of the requirement by the measure. Subsequently, all the requirement evaluation values for the plurality of measures are added to each other to obtain the evaluation value (see Equation (3)).Subsequently, in step S 132 b, a useful value is determined for each of the plurality of measures depending on the evaluation values of the measure, the expected values of the functional unit 1 a, 1 b,..., 1 nand the influencing factors of the functional unit 1 a, 1 b,..., 1 n. For this purpose, a difference is first formed between the evaluation value of the functional unit 1 a, 1 b,..., 1 nfor the measure and the expected value of the functional unit 1 a, 1 b,..., 1 n. The difference is potentiated with the influencing factor in order to determine individual values specific to the functional unit, i.e. the value of the measure for the requests sent by the functional unit 1a, 1b,..., 1n. Subsequently, all individual values of use for the respective measure are multiplied with one another in order to obtain the value of use of the measure (cf. equation (4)).Subsequently, in step S 132 c, a decision evaluation is determined for each of the plurality of measures depending on the usage values of the plurality of measures. For this purpose, in particular the real part and the imaginary part of the useful value of the measure are added to one another (cf. equation (5)).Finally, in step S 132 d, the action to be taken is determined based on the decision scores of the plurality of actions. Specifically, the one of the plurality of actions is determined as the action to be performed that has the largest decision evaluation.The method is explained in more detail below with reference to an example and numerical values. In the example, nine requests, referred to below as A1 to A9, are transmitted from three functional units, identified below by the numerals 1 to 3, to the decision unit. Furthermore, setpoint values are sent to the decision unit, from which the decision unit determines three different measures. Here, it can be assumed, for example, that each of the functional units 1 to 3 has sent a target value specification to the decision unit and the decision unit has determined the received target value specifications as measures.The following table shows the requirements, the associated weighting factors and the degrees of fulfilment EG for the measures 1 to 3 for the functional unit 1.A11061010A233106A35834A461050A58371In sum, the functional unit 1 sends five requests (A1 to A5) to the decision unit. The sum of the weighting factors yields a value of 32.The following table shows the requirements, the associated weighting factors and the degrees of fulfilment for the measures 1 to 3 for the functional unit 2.A19376A2108710A33939A62233A75862In sum, the functional unit 2 sends five requests to the decision unit. In this case, the functional unit 2, as well as the functional unit 1, sends the requests A1, A2, A3 to the decision unit. Furthermore, the functional unit 2 sends the requests A6 and A7 to the decision unit. The sum of the weighting factors yields a value of 29.The following table shows the requirements, the associated weighting factors and the degrees of fulfilment for the measures 1 to 3 for the functional unit 3.Weighting factorEG n,3,1EG n,3,2EG n,3,3A377109A48792A89338A9103210In sum, the functional unit 3 sends four requests to the decision unit. In this case, the functional unit 3, as well as the functional units 1 and 2, sends the requests A3 to the decision unit. Furthermore, the functional unit 3 sends the decision unit for the request A4, which is also sent by the functional unit 1. Furthermore, the functional unit 3 sends the requests A8 and A9 as the only functional unit to the decision unit. The sum of the weighting factors yields a value of 34.The following table shows, for each request and each functional unit, the relative weighting factor (given in %) which was calculated on the basis of equation (1).A131%31%0%A29%34%0%A316%10%21%A419%0%24%A525%0%0%A60%7%0%A70%17%0%A80%0%26%A90%0%29%Then, the evaluation values of the functional unit for each of the measures are determined from the equations (2) and (3).The evaluation values of the measures by the individual functional units are given in the following table.16,037,224,5626,146,146,7934,765,567,38Subsequently, the usage values for each of the plurality of measures must be determined. For this purpose, it is established, on the one hand, that each of the functional units has an expected value of 5, i.e. measures with evaluation values with a value of 5 or more are evaluated as positive by the respective functional unit and measures with evaluation values with a value of less than 5 are evaluated as negative by the respective functional unit. It is thus apparent from the above table that functional unit 1 evaluates measure 3 negatively and functional unit 3 evaluates measure 1. The functional unit 1 has a prioritization factor of 1, the functional unit 2 a prioritization factor of 0.5 and the functional unit a prioritization factor of 0.95.Functional unit 1 therefore has an influencing factor of functional unit 2 an influencing factor of and functional unit 3 an influencing factor of Since the value of the prioritization factor is less than 1, a complex number results for the decision factors of the measures which are evaluated negatively by at least one of the functional units (i.e. have an evaluation value of less than 5 in the case of an expected value of 5).The utility value NW j of each of the plurality of measures calculated from Equation (4) and the decision values EW j, which are calculated as the sum of the real part and the imaginary part of the utility value of the measure (see Equation (5)), are shown in the following table.NW j-0.27 + 0.23·i1,29 + 0·i-1.06 + 0.71·iEW j-0,041,29-0,35Therefore, as the measure to be performed in this example, the measure j=2 is used because it has the highest decision value.

Claims

Method for controlling a computing unit group having a plurality of functional units (1a, 1b,..., 1n) and a decision unit (2), wherein each of the functional units (1a, 1b,..., 1n) is configured to receive a plurality of requests, to determine, as a function of the plurality of requests, one or more setpoint value specifications on which the computing unit group is to be controlled, and to transmit the plurality of requests, weighting factors belonging to the requests and the one or more setpoint value specifications to the decision unit (2), the method comprising: receiving (S100), by the decision unit (2), a plurality of requests, a plurality of weighting factors and a plurality of setpoint value specifications of at least two of the plurality of functional units (1a, 1 b,..., 1 n), wherein a respective weighting factor is received for each of the received requests, determining (S 110), by the decision unit (2), a plurality of measures depending on the plurality of setpoint values specifications, determining (S 120), for each measure of the plurality of measures and each of the plurality of requests, a fulfillment degree indicating a degree of fulfillment of the requests by the measure, determining (S 130), from the plurality of measures, a measure to be performed using the weighting factors and the fulfillment degrees; controlling (S 140) the computational unit group on the basis of the measure to be performed.The method according to claim 1, wherein the determining (S130) one measure to be performed among the plurality of measures using the weighting factors and the satisfaction degrees further comprises: determining (S131), for each of the requests, a relative weighting factor by forming a quotient of the weighting factor of the request and a sum of all the weighting factors of the requests of the same functional unit (1a, 1b,..., 1n), determining (S132) the measure to be performed using the relative weighting factors as weighting factors.The method according to claim 1 or 2, wherein the determining (S120) of the plurality of actions comprises: determining (S121a) the plurality of target values as the plurality of actions, or selecting (S121b) the plurality of actions depending on the plurality of target values from actions provided by the decision unit (2).Method according to one of the preceding claims, wherein the determination (S130) of the measure to be carried out from the plurality of measures using the weighting factors and the degree of fulfilment further comprises: determining (S132a), for each of the functional units (1a, 1b,..., 1n) and each of the plurality of measures, an evaluation value depending on the weighting factors and the degrees of fulfilment, determining (S132b), for each of the plurality of measures, a useful value depending on the evaluation values of the measure, expected values of the functional units (1a, 1b,..., 1n) which each indicate a threshold of the evaluation value from which the functional unit (1a, 1b,..., 1n) evaluates the performance of a measure as satisfying the request, and influencing factors of the functional units (1a, 1b,..., 1n) which each indicate the influence of the functional units (1a, 1b,..., 1n) on the usage value, determining (S132c), for each of the plurality of measures, a decision evaluation depending on the usage values of the measures, determining (S132d) a measure to be performed from the plurality of measures based on the decision evaluations of the measures to be performed.Method according to the preceding claim, wherein the evaluation value is determined as the sum of the products of the weighting factor of each functional unit request with the associated degree of fulfilment of the functional unit request for the action.The method according to claim 4 or 5, wherein the determining (S132b) the utility value comprises: determining, for each of the plurality of actions, a difference between the evaluation value of the functional unit (1a, 1b,..., 1n) for the action and the expectation value of the functional unit (1a, 1b,..., 1n), potentiating each difference with the influence factor to obtain a single utility value for each of the functional units (1a, 1b,..., 1n), and multiplying all single utility values associated with a action to obtain the utility value of the action.The method according to any one of claims 4 to 6, wherein the determining (S132c) a decision score comprises: determining a real part and an imaginary part of the utility of the measure, and determining the decision score of the measure as a sum of the real part and the imaginary part of the utility of the measure.Method according to one of Claims 4 to 7, wherein the measure with the highest decision rating is determined as the measure to be carried out.Computing unit which is configured to carry out all method steps of a method according to one of the preceding claims.A computer program that causes a computing unit to perform all method steps of a method according to any one of claims 1 to 8 when executed on the computing unit.A machine readable storage medium having stored thereon a computer program according to claim 10.

Images