Patent Application
20190165996
The invention relates to a method for operating a real-time-capable simulation network having multiple network nodes for computing a simulation model, wherein the network nodes are connected to one another by means of a serial data bus, and the network nodes exchange data by means of data bus messages, wherein at least one event-driven task of the simulation model is implemented on a first network node, wherein a nondeterministic triggering event is detected by a second network node, wherein the second network node communicates the detected triggering event to the first network node by transmission of a detection signal, and the first network node computes the event-driven task after receiving the detection signal.
Methods of the type described above for operating a real-time-capable simulation network have been in use for some years, in particular in control unit development for distributed systems. Fields of application typically reside in the area of very complex technical systems, as for example in automotive applications, in aerospace, and in other demanding industrial areas where extensive use is made of open-loop and closed-loop controls.
In the applications that are of interest here, multiple network nodes—it is not unusual for there to be several tens or more—are connected to one another through a serial data bus, or even several serial data buses, and jointly carry out a control task in the broadest sense of the term. The network nodes are typically connected to a physical technical process; they acquire information through sensors about states of the process and exert influence on the connected physical technical process through connected actuators.
The network nodes usually have varying designs, although they can also have identical characteristics; typically, the network nodes are set up for certain specialized tasks, however. Some network nodes essentially provide massive computing capacity in order to compute control algorithms and manage the processing of relatively large quantities of data, where this generally must be carried out subject to real-time requirements. The real-time requirement is a simple result of the fact that the networks considered here are, at least in part, in direct use with a real physical technical process. Microprocessors, microcontrollers, and even hard-wired functional components such as FPGAs (Field Programmable Gate Arrays) are used as computing units for network nodes of this type.
Other network nodes implement more of a measurement or actuation specialization in which the emphasis is not so much on a particular computing capacity, but rather on providing suitable I/O interfaces, for example in the form of analog interfaces with suitable analog-to-digital converters and digital-to-analog converters, PWM inputs and outputs, digital inputs and outputs, and if applicable interfaces for specific bus protocols. The dividing line between the different network nodes with different specializations is fluid; network nodes specialized for computing can also have I/O interfaces, just as network nodes specialized for sensing or actuation typically have suitable computing units.
Taken as a whole, the above-described networks constitute distributed discrete-time sampled-data systems.
In order to develop networks and network components of this type, individual network nodes or even large parts of the network having multiple network nodes are simulated in real time. In this context, a distinction is made largely between the application cases of Rapid Control Prototyping (RCP) and hardware-in-the-loop simulation (HIL simulation). In Rapid Control Prototyping, several—or even all—network nodes of the actual network are replaced by development hardware, which is generally more powerful than the production network nodes that will actually be used later. The network nodes are then typically programmable by means of a graphical programming environment and a connected automatic code generator, so that various functionalities can be implemented and can be tested in the actual physical environment within a very short time.
In hardware-in-the-loop simulation, exactly the opposite application case is tested, which is to say the production control units existing at the end of development, in the form of network nodes, are tested in a simulated hardware environment, wherein the simulated environment includes a part of the network with simulated network nodes and at least a part of the connected physical process. The method considered here for operating a real-time-capable simulation network applies equally to both application cases.
As already explained, the network with its network nodes constitutes a distributed discrete-time sampled-data system. Viewed from the standpoint of time, there are two types of tasks that must be processed by the various network nodes. There are the tasks that run, as it were, in time with the sampled-data system, and therefore run completely deterministically, and there are other tasks that are unforeseeable in time, and thus are processed nondeterministically, and cannot be planned into a time scheme. Tasks of the first type are often called “timer tasks” and are processed periodically. For example, there may be tasks that are executed every millisecond or even only every 50 milliseconds, or indeed with other period lengths that are an integer multiple of the smallest sampling period length. Examples of such timer tasks, which is to say “time-driven” tasks, are the periodic computation of deviations and controlled variables, as well as the acquisition of measurement data from the process and the output of controlled variables to the connected process.
Nondeterministic applications, which thus are nonperiodic and consequently cannot be predictably executed or computed, are also called “event-driven” tasks or nonperiodic tasks. The computation of such tasks is triggered by similarly nondeterministic events (influencing of the physical process by operators, alteration of boundary conditions, etc.). The present case is concerned with the handling of this type of nondeterministic events, which trigger the computation of an event-driven task associated therewith.
It is known to the applicant from practice that a nondeterministic triggering event is detected by the second network node, and the second network node first communicates the detected triggering event to the first network node, which has an event-driven task connected to this detected triggering event. The first network node in turn notifies all other network nodes from which it absolutely requires data for computing the event-driven task, whereupon these network nodes acquire these data (measurement, computation), and then transmit the acquired data back to the reporting first network node. Thus, the computation of the event-driven task on the first network node is then made possible. The result that only involved network nodes participate is achieved by the directly addressed back-and-forth transmission of detection signals, requests for information, and the information itself. On the other hand, this procedure is quite time-intensive. Other methods, in which separate physical signal lines are provided for transmission of triggering events or of the related detection signals, entail a considerable added effort (cabling, complicated connectors, etc.).
The object of the present invention is to specify the method set forth above for operating a real-time-capable simulation network having multiple network nodes with which the above-described disadvantages are at least partially remedied.
The object derived above is initially and substantially attained according to the invention for the above-presented method by the means that the detection signal is sent from the second network node in the form of a multicast data bus message or a broadcast data bus message to multiple network nodes of the simulation network or to all network nodes of the simulation network over the serial data bus. The detection signal here—unlike in the prior art—is thus sent not only to the first network node with the event-driven task that is to be executed, but instead is sent simultaneously to multiple network nodes (multicast data bus message) or to all network nodes (broadcast data bus message).
When the detection signal is sent by means of a broadcast data bus message, all network nodes are notified practically simultaneously of the detection of the triggering event, creating the possibility that all network nodes participating in the computation of the event-driven task can also trigger the necessary measures practically simultaneously. When the detection signal is sent by means of a multicast data bus message, the data bus message is sent, for example, to all network nodes that relate to the event-driven task initiated by the nondeterministic triggering event. A function of the event-driven task can be the sending of signals over I/O interfaces, for example. The I/O interfaces here can be connected directly to one of the network nodes participating in the event-driven task. Optionally, context information can be sent in addition to the detection signal in the same message or in an additional broadcast or multicast message. Context information includes information that relates to the triggering event. Context information can be, e.g., the state of the second network node at the time of the detection of the triggering event or the current measurement value of an I/O interface of the second network node. Context information can be taken into account by network nodes when processing the event-driven task.
In a preferred embodiment of the method, provision is made that the network nodes that have a data item necessary for computing the event-driven task are equipped with a correspondence information item, wherein the correspondence information item includes the triggering event, the source of the data item, and—directly or indirectly—the first network node on which the event-driven task is computed. The correspondence information item makes it possible for the network nodes that have a data item necessary for computing the event-driven task to determine, with the aid of the received detection signal, which data item within their sphere of influence should be prepared and where the data item of interest should be sent.
In this case, the necessary data item can be sent directly to the first network node, where the event-driven task is in fact being computed. However, the data item can also be sent indirectly, which is to say by way of the detour through at least one other network node. In any case, the correspondence information item includes a description of the “path to the destination” that should be taken through network nodes of the simulation network until the necessary data item—in its original form or modified along the way—reaches its destination in the form of the first network node. In other words, the method is designed accordingly such that, upon receiving the multicast data bus message or the broadcast data bus message, the network nodes that have a data item necessary for computing the event-driven task prepare this necessary data item and send it with a data bus message directed to another network node. This other network node need not necessarily be the first network node; rather, the necessary data item may also take a detour through additional network nodes to the first network node.
An improvement of the described method is distinguished in that a local event of the second network node and/or an external event of the second network node is detected as a nondeterministic triggering event. A local event can be, e.g., an error or a state change that is taking place. The detection of an external event preferably occurs by means of an I/O interface of the second network node, and in this case it may be, for example, an event from the connected technical process that is detected through the I/O interface. The I/O interface can also be an—additional—communication interface, which is to say the event can also consist in the reception of a specific data bus message.
It has previously been stated that the network nodes that have a data item necessary for computing the event-driven task prepare this necessary data item. In one embodiment of the method, the preparation of the necessary data item comprises the readout of the aforementioned I/O interface of the second network node, or even of the I/O interface of another network node through which the first network node is indirectly informed. Alternatively or in addition, provision can be made that the preparation of the necessary data item comprises the determination of the necessary data item through a preprocessing of other data. Thus, for example, a multidimensional family of characteristic curves that places multiple quantities in relationship to one another can be stored in a network node, wherein all quantities except for one quantity are determined through, e.g., readout of measured values or even internal calculation variables, and the remaining quantity is determined from the family of characteristic curves.
A preferred embodiment of the method is distinguished in that the preparation of the necessary data item and/or the sending of the data bus message directed to a network node—in particular the first network node—takes place with higher priority than the computation of other processes or the sending of other data bus messages. In order to incur the lowest possible latencies, provision is made in particular for the preparation and/or the sending to take place with the highest priority. This ensures that possible delays in the response to the nondeterministic triggering event are minimized. Of course, for this purpose it is necessary for both the operating system on the network nodes and the protocol on which the exchange of data bus messages is based to permit a corresponding prioritization.
Furthermore, provision is made in a particular embodiment of the method that the correspondence information item includes an information item about the event-driven task, in particular wherein the data bus message directed to the first network node performing the computation also includes an information item on the basis of which the first network node can associate the data item with the event-driven task. Thus, in a certain sense the information item is an identifier for unique identification. This information item can be a memory address in which the data are stored in the first network node, for example. The applicable event-driven task can read the applicable current data from this address out of the local memory of the first network node and use it for the task computation.
Another embodiment of the method is characterized in that the correspondence information item is determined at least partially automatically from a simulation model and is stored in the network nodes before the start of computation of the simulation model. In addition or alternatively, the correspondence information item can also be dynamically changed at simulation run time.
In an especially preferred improvement of the method, the event-driven task on the first network node starts with the processing of another associated data item as early as following receipt of the multicast data bus message or the broadcast data bus message. This means, in particular, a data item that is already present on the first network node and is sufficiently up-to-date. All such additional associated data items that are already currently available can be incorporated into the computation of the event-driven task this early without resulting in inaccuracies. Moreover, this mode of procedure reduces the overall response time or processing time of the event-driven task. To ensure that only a current necessary data item that originates from a different network node than the first network node is used in every case, all such received necessary data items or the memory locations in which they are stored are marked with a flag that at first indicates that the data item has not yet been received in updated form since the calling of the event-driven task. Only after the necessary data item or necessary data items have been sent by network nodes different from the first network node is each corresponding flag switched such that it indicates valid data. Each time, before the event-driven task on the first network node reads memory locations corresponding to necessary data items sent from outside, it checks whether the data are in fact current—which is to say valid. After computation of the event-driven task is complete, the corresponding flags are switched again so that the memory contents stored there are marked as invalid—which is to say not current.
An especially preferred embodiment of the method makes provision that the time is determined that a necessary data item requires to reach the first network node from the second network node after receipt of the multicast data bus message or the broadcast data bus message, that furthermore the time is calculated that the first network node requires after receipt of the multicast data bus message or the broadcast data bus message to start computation of the event-driven task and reach the execution point in the event-driven task that accesses the necessary data item, and that the computation of the event-driven task is started on the first network node, taking the ascertained times into account, such that the event-driven task can access a necessary data item without additional waiting for a necessary data item. As a result of this mode of procedure, it is ensured through the use of advance knowledge of certain execution times that the computation of the event-driven task automatically uses only current and externally obtained data items. In any case, this could make a validation system for ensuring the freshness of received data unnecessary.
Another variant for ensuring a current data basis in the calculation of the event-driven task consists in that the first network node determines whether all data necessary for computing an event-driven task have arrived, and computation of the event-driven task on the first network node does not start until all necessary data have arrived. Although this method does not realize the shortest response time, the method can be implemented with very simple means.
Another measure for reducing latencies consists in that the multicast data bus message or the broadcast data bus message is sent and/or forwarded with a higher priority than all other data bus messages. For this purpose the serial data bus used must also have appropriate instruments for prioritizing messages.
The invention additionally includes a computer program product comprising instructions that, when the programs are executed by multiple network nodes of a real-time capable simulation network for computing a simulation model, cause the network nodes to carry out the steps of the above-described method. In addition, the invention likewise includes a computer-readable storage medium on which the above-described computer program product is stored.
Further scope of applicability of the present invention will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.
The present invention will become more fully understood from the detailed description given hereinbelow and the accompanying drawings which are given by way of illustration only, and thus, are not limitive of the present invention, and wherein:
Methods 1 for operating a real-time-capable simulation network 2 having multiple network nodes 3, 3.1, 3.2, 3.3, 3.4 for computing a simulation model 4, 4.1, 4.2 are shown in each of the various figures. Shown in each case are simulation networks 2 with multiple network nodes 3, wherein process sequences from the diagrams logically result from the signal flow and the sequences of data communication that are shown and described.
Shown here is that the simulation model 4 is distributed over two network nodes 3.1, 3.4, namely in the form of the parts 4.1, 4.2 of the simulation model 4. An event-driven task 7 is implemented on the network node 4.1 [sic; should be 3.1] within the framework of the part 4.1 of the simulation model 4 on the network node 3.1.
A nondeterministic triggering event 8 is detected by the second network node 3.2, with this second network node 3.2 communicating the detected triggering event 8 to the first network node 3.1 by transmission of a detection signal 9. This detection signal 9 is thus transmitted by means of a data bus message 6. It is known to the first network node 3.1 that it must compute the event-driven task 7 that is connected with the triggering event 8 after receiving the detection signal 9. However, in order to compute the event-driven task 7, the first network node 3.1 needs additional information that is not stored on the first network node 3.1 itself but instead must first be obtained from other network nodes 3, in this case from the additional network node 3.3. To this end, the first network node 3.1 requests a necessary information item from the additional network node 3.3 (directed arrow with a data bus message 6 from the first network node 3.1 to the additional network node 3.3). In response to this request for information, the additional network node 3.3 then supplies the required information in the form of another data bus message 6 (directed arrow from the additional network node 3.3 to the first network node 3.1). As a result of the above-described procedure, only the network nodes 3 that are absolutely required in order to compute the event-driven task 7 are addressed, but a relatively long response time to the triggering event 8 must be accepted.
In the method illustrated in
It is shown in
In the examples shown, an external event of the second network node 3.2 is detected as a nondeterministic triggering event 8, namely in the present case a nondeterministic event in the physical process P is detected with an I/O interface 10 of the second network node 3.2.
It is shown in
The behavior in the exemplary embodiments from
The sending of the necessary data item need not always take place directly with a data bus message 6 that is directed immediately to the first network node 3.1. Shown in
In the embodiments shown, the preparation of the necessary data item takes place with the highest priority, which is to say that the additional network node 3.3 immediately stops processing other tasks upon receipt of the detection signal 9 and, in the present case, reads out the third channel of its I/O interface 10. Provision is likewise made in the exemplary embodiments that the sending of the data bus messages 6 directed to the additional network nodes 3.1, 3.3, 3.4, and in particular to the first network node 3.1, takes place with higher priority than the sending of other data bus messages 6, so that a short response time is guaranteed in this way.
It is shown schematically in
In the method from
The invention being thus described, it will be obvious that the same may be varied in many ways. Such variations are not to be regarded as a departure from the spirit and scope of the invention, and all such modifications as would be obvious to one skilled in the art are to be included within the scope of the following claims.
