JP4681513B2 - Real-time parallel distributed simulation system - Google Patents

Description

  The present invention relates to a real-time parallel distributed simulation system that performs parallel distributed simulation using a plurality of nodes while performing real-time information communication between a control device and a plurality of nodes.

  Conventionally, as disclosed in, for example, Japanese Patent Application Laid-Open No. 2000-163392 (Patent Document 1), parallel distributed simulation is performed using a plurality of nodes (simulators).

  By the way, in recent years, a so-called HIL simulation (Hardware In the Loop Simulation), which is performed while communicating information with a control device in real time, has attracted attention for the development and evaluation of the control device. That is, the HIL simulation is a so-called real-time simulation in which pseudo state information of the control object is output to the control device while inputting a control signal for controlling the actual control object from the control device. Such a real-time simulation enables the control device to be evaluated before the control object is manufactured. For example, Non-Patent Documents 1 and 2 disclose that this real-time simulation is performed by parallel distributed processing.

  Here, for example, a plurality of control devices and a plurality of control objects that can be controlled by the respective control devices are mounted on an automobile. When performing such a vehicle simulation, a real-time simulation is performed in which pseudo state information of a plurality of control objects is output to a corresponding control device while inputting control signals from the plurality of control devices. .

In this case, specifically, each node inputs a control signal from each control device, performs a process of generating pseudo state information of a control object of the control device based on the control signal, and The pseudo state information is output to the control device. The plurality of nodes communicate with each other and collect simulation results generated by the plurality of nodes, thereby performing a simulation of the entire vehicle.
JP 2000-163392 A Hisao Taoka, Yasushi Fujimoto, “Power System Simulation Technology”, “System / Control / Information Vol. 44, No. 2, 2000”, System Control Information Society, p. 73-77 Tatsuya Yamamoto, Kenichi Kuroda, Hisao Taoka, Hiroshi Enomoto, Yoshiyuki Kono, “Real-time power system simulator using PC clusters”, IEEJ Transactions B Vol. 124, No. 5 (2004), p. 733-740

  However, a plurality of control devices mounted on an automobile may perform operations related to each other. That is, as the state of the control object of a predetermined control device changes, the state of the control object of another control device may change. Furthermore, the state of the control object of another control device may change according to the control signal of a predetermined control device. Therefore, the above-described simulation has a problem that the real-time property of the mutual operation is inferior to the entire system including a plurality of control devices.

  Therefore, in order to improve the real-time nature of the interaction, it is considered that each node outputs the pseudo state information of the generated control object to another node and transmits the input control signal to the other node. It is done. That is, each node inputs a control signal of another control device via another node in addition to inputting a control signal of a directly connected control device, and is generated by another node. It is assumed that pseudo state information of a control object of another control device is input. Thus, each node improves the real-time nature of the interaction based on the control signal of the control device directly input, the control signal of the other control device, and the pseudo state information of the control object of the other control device. can do.

  However, in particular, in order to perform communication between the control device and the node in real time, a very long time is required for input / output of information. For this reason, it has been difficult to perform a simulation with improved real-time characteristics.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a real-time parallel distributed simulation system that can further improve real-time performance.

The real-time parallel distributed simulation system of the present invention is a real-time parallel distributed simulation system that targets a predetermined control object composed of a plurality of partial control objects, and includes a plurality of control devices, one I / O An O node, a plurality of operation nodes, and communication means are provided. The plurality of control devices are devices capable of controlling each partial control object by outputting a control signal corresponding to the state information of each of the partial control objects to the partial control object. One I / O node is a node that inputs respective control signals from a plurality of control devices and outputs pseudo state information, which is pseudo state information of the partial control object, to the plurality of control devices. . The plurality of operation nodes are nodes that generate respective pseudo state information based on the respective control signals input from the I / O nodes and each include an operation processing device. The communication means is means for communicating between the I / O node and the plurality of operation nodes and between each of the operation nodes every predetermined time.
Further, at least one of the computation nodes performs communication time management between the plurality of computation nodes and communication time management between the I / O node and the computation node. In addition, following the communication between the I / O node and the plurality of operation nodes and between the operation nodes at each predetermined time, the plurality of operation nodes transmit the pseudo state information. The I / O node performs input / output processing with the plurality of control devices in parallel with the generation of the pseudo state information by the arithmetic node.

  Here, the node includes a single arithmetic processing unit (such as a CPU) and a dual CPU including two CPUs. Of course, the node may include three or more arithmetic processing devices.

  The processing operation of the real-time parallel distributed simulation system of the present invention will be described. First, a control signal output by the control device is input to the I / O node. The control signal input by this I / O node is transmitted to the computation node at predetermined intervals via the communication means. Then, pseudo state information of the controlled object is generated based on the transmitted control signal at the operation node. Subsequently, the generated pseudo state information is transmitted from the operation node to the I / O node via the communication unit at predetermined time intervals. Then, the I / O node outputs the transmitted pseudo state information to the control device.

  Thus, according to the real-time parallel distributed simulation system of the present invention, the control device can be simulated in real time even when there is no control target that is the control target of the control device.

  Further, according to the present invention, the input of the control signal from the control device and the output of the pseudo state information to the control device are always performed by the I / O node. That is, the arithmetic node does not input a control signal from the control device and output pseudo state information to the control device. Therefore, the calculation node mainly generates pseudo state information of the control object, that is, performs a simulation calculation. On the other hand, the I / O node mainly inputs and outputs information with the control device. Further, the I / O node and the computation node perform information communication with each other via the communication means at predetermined time intervals.

  Here, in order to perform real-time simulation, input / output of information with the control device and simulation calculation require a considerable time. According to the present invention, input / output of information with respect to the control device requiring time and simulation calculation are processed by separate nodes. Therefore, since processing can be distributed very efficiently, real-time performance can be improved.

In the real-time parallel distributed simulation of the present invention , there are a plurality of operation nodes. That is , the communication means communicates between the I / O node and the plurality of operation nodes and between the operation nodes at every predetermined time. As described above, even when there are a plurality of operation nodes, only the I / O node inputs / outputs information to / from the control device. The plurality of computation nodes perform simulation computation without inputting / outputting information with the control device. In addition, by using a plurality of operation nodes, the operation processing can be distributed, so that the processing speed can be increased. It can also be used for large-scale systems.

  Here, in order to perform parallel and distributed simulation in real time, it is necessary for the communication means to communicate between nodes every predetermined time. In order to perform communication every predetermined time in this way, it is necessary to manage the communication time. Therefore, when there are a plurality of operation nodes, at least one of the plurality of operation nodes is connected to the time management of communication between the plurality of operation nodes and between the I / O node and the operation node. The communication time is managed. Specifically, when an arithmetic node that performs time management determines that a predetermined time has been reached, it instructs the nodes to communicate with each other. Further, when the I / O node and other computing nodes have finished the predetermined processing, the instruction is issued to wait for synchronization until it is determined that the computing node that performs time management has reached the predetermined time. In this manner, the parallel distributed simulation can be reliably performed in real time by instructing the arithmetic node to perform synchronous communication between the I / O node and the plurality of arithmetic nodes. Note that the computation node that performs time management may perform only time management, or may perform simulation computation in addition to time management.

  In addition, the real-time parallel distributed simulation system of the present invention includes a host computer that stores pseudo basic information related to an object to be controlled, inputs pseudo basic information from the host computer, and outputs pseudo basic information to an operation node. An information management node that inputs information from the operation node and outputs pseudo state information to the host computer may be further provided. In this case, the computing node generates pseudo state information based on the pseudo basic information and the control signal, and the communication means communicates between the information management node, the I / O node, and the computing node at predetermined time intervals. To.

  Here, the host computer only needs to store information necessary for the user, such as the simulation status and results, and pseudo basic information (basic model) of the controlled object. That is, in many cases, a device that cannot communicate with a computing node or the like in real time, that is, a device that cannot communicate with a computing node or the like every predetermined time is used for the host computer. On the other hand, in order for the computation node to perform the simulation computation in real time, the computation node needs to acquire the pseudo basic information of the controlled object in real time. Further, the pseudo state information generated by the operation node is sequentially updated in the operation node. Therefore, the operation node may not store pseudo state information at a predetermined time in the past.

  Thus, as described above, by providing the information management node, the information management node can input the pseudo basic information of the controlled object from the host computer and then output it to the computation node. Thereby, the operation node can acquire the pseudo basic information. Furthermore, the information management node can temporarily store the pseudo state information generated by the operation node and output the necessary pseudo state information to the host computer. Thereby, the host computer can acquire the pseudo state information at the past predetermined time.

In the real-time parallel distributed simulation according to the present invention, the control object includes a plurality of partial control objects, and the control device includes a plurality of control signals corresponding to the state information of the respective partial control objects. Each partial control object can be controlled by outputting to the partial control object . For example, an automobile has a plurality of control objects (partial control objects) such as an engine, a steering, and an air conditioner, and a plurality of control devices that respectively control the plurality of partial control objects. As described above, even when there are a plurality of control devices, each of the plurality of control devices inputs / outputs information to / from the I / O node and directly inputs / outputs information to / from the operation node. There is nothing. That is, even if there are a plurality of control devices, the real-time parallel distributed simulation of the present invention can be applied.

  The control object includes a first partial control object having a predetermined control cycle and a second partial control object having a control cycle shorter than that of the first partial control object, and the control device includes the first partial control object. A first control device capable of controlling the first partial control object by outputting a first control signal corresponding to the state information of the object to the first partial control object, and state information of the second partial control object When a second control device that can control the second partial control object by outputting a second control signal corresponding to the second partial control object to the second partial control object may be provided.

  That is, the I / O node inputs the first control signal from the first control device and outputs the first pseudo state information, which is pseudo state information of the first partial control object, to the first control device. Then, the operation node generates the first pseudo state information based on the first control signal. Furthermore, the real-time parallel distributed simulation system of the present invention receives the second control signal from the second control device, and supplies the second pseudo state information, which is the pseudo state information of the second partial control object, to the second control device. The second pseudo-state information is generated, and an ultrahigh-speed node having a processing speed higher than that of the operation node is provided. Then, the communication means communicates between the I / O node, the computation node, and the ultra high speed node every predetermined time.

  That is, by setting the real-time parallel distributed simulation system of the present invention as described above, input / output of information related to the first control device and simulation calculation are handled by the I / O node and the calculation node, and information related to the second control device. The input / output and simulation operations are handled by the ultra high-speed node. Thereby, the simulation calculation according to each control period can be performed. Note that since the processing speed of the ultra high-speed node is high, even if both input / output of information with the second control device and simulation calculation thereof are performed, it is possible to cope with it sufficiently.

  Here, the ultra high-speed node may include, for example, an FPGA (Field Programmable Gate Array), a CPLD (Complex Programmable Logic Device), or the like. In addition, the ultra-high speed node may have a CPU in addition to the FPGA and CPLD. In this case, the FPGA and CPLD may input the second control signal and generate the second pseudo state information, and the CPU may communicate with the I / O node and the operation node. Note that the ultra-high speed node may be composed of a plurality of CPUs and the like. For example, the ultra high-speed node may be composed of a dual CPU or a dual-core arithmetic device.

  In addition, the real-time parallel distributed simulation system of the present invention further includes a pipeline operation processing unit configured by a plurality of nodes and performing a pipeline operation, and the communication unit performs the I / O node and the pipeline operation processing every predetermined time. You may make it communicate between units. Here, the pipeline operation is an operation that is sequentially performed by each node so that other nodes perform an operation using an operation result of a predetermined node.

  In this case, in particular, the pipeline calculation processing unit may analyze the pseudo state information by pipeline calculation. The analysis of the pseudo state information is, for example, body collision analysis or engine combustion analysis in the case of an automobile HIL simulation. Such analysis of pseudo state information requires a great number of operations. By making such an analysis performed by the pipeline arithmetic processing unit, the real-time property of the entire system can be ensured.

  According to the real-time parallel distributed simulation system of the present invention, the input / output of information with the control device and the simulation calculation are performed from different nodes, thereby improving the real-time property in a large-scale simulation system or the like. it can.

  Next, the present invention will be described in more detail with reference to embodiments. Here, in this embodiment, a case where the real-time parallel distributed simulation system of the present invention is applied to an HIL simulation of a hybrid automobile system will be described as an example. The real-time parallel distributed simulation system of this embodiment will be described with reference to FIGS. FIG. 1 is a block diagram of the real-time parallel distributed simulation system of this embodiment. FIG. 2 is an explanatory diagram relating to the pipeline arithmetic processing unit. FIG. 3 is a diagram illustrating processing of each node.

  As shown in FIG. 1, the real-time parallel distributed simulation system includes an ECU 1a to 1c, an ECU 2, a host PC 3, an information management node 4, an I / O node 5, a first calculation node 6, and a second calculation node. 7, a third operation node 8, an ultrafast node 9, a pipeline operation processing unit 10, and a communication network unit 11.

  The ECUs 1a to 1c and the ECU 2 are control devices (hereinafter referred to as “ECUs”) for a hybrid vehicle. That is, the ECUs 1a to 1c and the ECU 2 as a whole control the hybrid vehicle. In particular, the ECUs 1a to 1c are those having a relatively long control cycle among ECUs of hybrid type vehicles. For example, the ECUs 1a to 1c are an engine ECU, a transmission ECU, a brake ECU, and the like. That is, the control objects (corresponding to the first partial control object in the present invention) of the ECUs 1a to 1c are engines, transmissions, brakes, and the like (hereinafter simply referred to as “engines”). And when this ECU1a-1c is connected to each control target object, by outputting the control signal according to state information, such as an engine which is a control target object, to a control target object, You can control things.

  Moreover, ECU2 is shorter than the control period of ECU1a-1c among ECU of a hybrid type vehicle. For example, the ECU 2 is an ECU or the like of a hybrid drive motor. That is, the control object of the ECU 2 (corresponding to the second partial control object in the present invention) is a drive motor or the like. Then, if the ECU 2 is connected to each control object, the ECU 2 outputs a control signal corresponding to the state information of the drive motor or the like that is the control object to the control object. You can control things.

  The host PC 3 (corresponding to a host computer in the present invention) is a computer that enables an operator to perform monitoring, evaluation, and the like. This host PC 3 stores a basic model (corresponding to the pseudo basic information in the present invention) of a hybrid vehicle to be controlled. Further, the host PC 3 can store and display simulation results.

  The information management node 4 includes an interface 41 communicatively connected to the host PC 3, a CPU 42, a main memory 43, and an interface 44 connected to the communication network unit 11. Then, the CPU 42 of the information management node 4 inputs a basic model from the host PC 3 via the interface 41 and temporarily stores it in the main memory 43. Further, the CPU 42 of the information management node 4 transmits the basic model stored in the main memory 43 to the communication network unit 11 via the interface 44. In addition, the CPU 42 of the information management node 4 sequentially inputs simulation results including pseudo state information generated by first to third operation nodes 6 to 8, an ultrahigh-speed node 9, and a pipeline operation processing unit 10 described later. To do. Further, the CPU 42 of the information management node 4 outputs the input simulation result to the host PC 3.

  The I / O node 5 is communicatively connected to the ECUs 1a to 1c and inputs / outputs information to / from the ECUs 1a to 1c. The I / O node 5 includes an interface 51 communicatively connected to the ECUs 1 a to 1 c, an FPGA 52, a CPU 53, a main memory 54, and an interface 55 connected to the communication network unit 11. The FPGA 52 of the I / O node 5 inputs control signals from the ECUs 1 a to 1 c via the interface 51 and temporarily stores them in the main memory 54. The control signals of the ECUs 1 a to 1 c stored in the main memory 54 are transmitted to the communication network unit 11 via the interface 55 by the CPU 53 of the I / O node 5.

  Further, the CPU 53 of the I / O node 5 inputs the respective pseudo state information from the first to third operation nodes 6 to 8 through the interface 55 and temporarily stores them in the main memory 54. The pseudo state information stored in the main memory 54 is output to the corresponding ECUs 1 a to 1 c by the FPGA 52 of the I / O node 5.

  The first to third operation nodes 6, 7, and 8 are mainly based on the control signals of the ECUs 1 a to 1 c input to the I / O node 5 and the basic model input to the information management node 4. A simulation on the engine or the like that is the control object is performed. And the 1st-3rd operation nodes 6, 7, and 8 produce | generate pseudo state information, such as an engine.

  Specifically, the first computation node 6 includes an interface 61, a CPU 62, and a main memory 63 that are communicatively connected to the communication network unit 11. In addition to the basic model and control signals of the ECUs 1 a to 1 c, the CPU 62 of the first computation node 6 inputs pseudo state information such as a driving motor generated by the ultra high speed node 9 and temporarily enters the main memory 63. Remember me. Then, based on this input information, the CPU 62 of the first calculation node 6 performs a simulation calculation related to the engine or the like, and generates pseudo state information such as the engine. Further, the CPU 62 transmits the generated pseudo state information such as the engine to the communication network unit 11. In addition, the CPU 62 of the first computation node 6 performs time management of each of the nodes 4, 5, 6, 7, 8, 9, and 10 that are communicatively connected to the communication network unit 11. Here, in order to perform the simulation in real time, it is necessary for the respective nodes to communicate with each other every predetermined time. That is, the CPU 62 of the first computation node 6 is responsible for time management for performing in real time.

  Each of the second computation node 7 and the third computation node 8 includes interfaces 71, 82, CPUs 72, 82, and main memories 73, 83 that are communicatively connected to the communication network unit 11. In addition to the basic model and the control signals of the ECUs 1a to 1c, the CPUs 72 and 82 of the second calculation node 7 and the third calculation node 8 input pseudo state information such as a driving motor generated by the ultra high speed node 9. Thus, it is temporarily stored in the main memories 73 and 83. Then, the CPUs 72 and 82 of the second calculation node 7 and the third calculation node 8 perform a simulation calculation related to the engine or the like based on the input information, and generate pseudo state information such as the engine. Further, the CPUs 72 and 82 transmit the generated pseudo state information such as the engine to the communication network unit 11.

  The first calculation node 6, the second calculation node 7, and the third calculation node 8 perform simulation calculation by parallel distributed processing related to the engine or the like. That is, pseudo state information relating to the engine or the like is generated in the entire first to third operation nodes 6 to 8. Further, the processing ratio of the simulation calculation by the second calculation node 7 and the third calculation node 8 is larger than the processing ratio of the simulation calculation by the first calculation node 6. This is because the first computation node 6 performs time management, whereas the second computation node 7 and the third computation node 8 do not perform time management.

  The ultra high-speed node 9 particularly performs a simulation calculation related to the drive motor. Since this drive motor has a very short control cycle, it is necessary to perform input / output of information with the ECU 2 and generation of pseudo state information of the drive motor at high speed. Therefore, the super high speed node 9 has a higher processing speed than the first to third operation nodes 6 to 8.

  The ultra high speed node 9 includes an interface 91 communicatively connected to the ECU 2, an FPGA 92, a CPU 93, a main memory 94, and an interface 95 connected to the communication network unit 11. The FPGA 92 of the ultra high speed node 9 inputs a control signal from the ECU 2 via the interface 91. Further, the CPU 93 of the super high speed node 9 inputs pseudo state information such as the engine from the first to third arithmetic nodes 6 to 8 through the interface 95 and temporarily stores them in the main memory 94.

  Then, the FPGA 92 performs a simulation calculation related to the drive motor based on the control signal of the ECU 2 and the pseudo state information of the engine or the like, and generates the pseudo state information of the drive motor. Further, the FPGA 92 outputs the generated pseudo state information of the drive motor to the ECU 2 and temporarily stores the pseudo state information in the main memory 94. Note that the FPGA 92 of the ultrahigh-speed node 9 inputs and outputs information with the ECU 2 at high speed in order to match the high-speed processing of the ECU 2. Then, the pseudo state information of the driving motor stored in the main memory 94 is transmitted to the communication network unit 11 via the interface 95 by the CPU 93 of the ultra high speed node 9.

  The pipeline arithmetic processing unit 10 includes a plurality of nodes (including a CPU). Then, pipeline operation is performed by each node. Here, the pipeline operation will be described in detail with reference to FIG. In FIG. 2, alphabets A to F indicate respective nodes, and arrows indicate information transmission paths.

  As shown in FIG. 2, the pipeline operation is an operation in which a predetermined node performs an operation and the result is used by another node. Specifically, after node A and node D perform calculations, node B and node E perform new calculations using the calculation results of nodes A and D, respectively. Then, the node C and the node F perform a new calculation using the calculation results of the node B and the node C.

  Inputs from other than the pipeline arithmetic processing unit 10 are performed only from A and D, and outputs from other than the pipeline arithmetic processing unit 10 are performed from C and F only. That is, communication congestion can be reduced and real-time performance is improved.

  The pipeline arithmetic processing unit 10 includes a basic model input by the information management node 4, a control signal of the ECUs 1a to 1c input by the I / O node 5, a control signal of the ECU 2 input by the ultra high speed node 9, Based on the pseudo state information of the engine and the like generated by the third operation nodes 6 to 8 and the pseudo state information of the driving motor generated by the ultra high speed node 9, the body collision analysis and the engine combustion analysis are performed. Here, the body collision analysis is an analysis of the deformation state of each part, the impact force transmitted to the occupant, etc., assuming that the simulation target automobile collides with an object.

  The communication network unit 11 uses Ethernet or the like. However, the communication network unit 11 may be a shared memory for each connected node.

  In the real-time parallel distributed simulation system configured as described above, the processing of the I / O node 5, the first to third operation nodes 6 to 8, and the ultrafast node 9 will be described with reference to FIG. In FIG. 3, two steps, step A and step B, are shown.

  As shown in FIG. 3, the time of communication performed between each of the nodes 4 to 9 and the pipeline arithmetic processing unit 10 is determined by the time management by the CPU 62 of the first arithmetic node 6. When the CPU 62 of the first calculation node 6 determines that it is time to communicate between the nodes 4 to 9 and the pipeline calculation processing unit 10, the CPU 62 of the first calculation node 6 determines that the nodes 4 to 10 are connected. Command to communicate.

  That is, in the first stage of the step time, the CPU 42 of the information management node 4, the FPGA 52 and CPU 53 of the I / O node 5, the CPU 62 of the first computation node 6, the CPU 72 of the second computation node 7, and the third computation node 8. The CPU 82 and the CPU 93 of the super high-speed node 9 communicate via the communication network unit 11. Subsequently, the CPU 62 of the first calculation node 6 performs a simulation calculation of the engine or the like, and thereafter performs time management until the step end time comes.

  The CPUs 72 and 82 of the second calculation node 7 and the third calculation node 8 perform the simulation calculation of the engine and the like after performing the above communication, and until the next communication command comes from the CPU 62 of the first calculation node 6. Wait for synchronization.

  After performing the above communication, the CPU 53 of the I / O node 5 waits for synchronization until the next communication command is received from the CPU 62 of the first computation node 6. Further, the FPGA 5252 of the I / O node 5 performs input / output of information (shown as “I / O processing” in FIG. 3) with the ECUs 1 a to 1 c after performing the above communication. That is, the FPGA 52 inputs control signals from the ECUs 1a to 1c and outputs the pseudo state information generated by the first to third calculation nodes 6 to 8 to the ECUs 1a to 1c.

  The CPU 93 of the super high speed node 9 waits for synchronization until the next communication command comes from the CPU 62 of the first computation node 6 after performing the above communication. Further, the FPGA 92 of the ultra high speed node 9 inputs / outputs information to / from the ECU 2 and repeatedly performs a simulation calculation regarding the driving motor, that is, generation of pseudo state information of the driving motor.

  Subsequently, the CPU 62 of the first computation node 6 outputs the next communication command, and the processing of the next step is performed.

  As described above, the input of control signals from the ECUs 1 a to 1 c and the output of pseudo state information to the ECUs 1 a to 1 c are always performed by the I / O node 5. That is, the first to third operation nodes 6 to 8 do not input control signals from the ECUs 1a to 1c and do not output pseudo state information to the ECUs 1a to 1c. Accordingly, the first to third calculation nodes 6 to 8 mainly perform generation of pseudo state information of the engine or the like, that is, simulation calculation related to the engine or the like. On the other hand, the I / O node 5 mainly inputs and outputs information with the ECUs 1a to 1c. Therefore, different nodes process information input / output and simulation computation with the ECUs 1a to 1c, which require time to perform simulation in real time. Therefore, since processing can be distributed very efficiently, real-time performance can be improved.

  In the above embodiment, the CPU 62 of the first computation node 6 performs time management, but the present invention is not limited to this. For example, a counter may be provided separately, and each node may communicate when the counter reaches a predetermined time.

It is a block diagram of the real-time parallel distributed simulation system of this embodiment. It is explanatory drawing regarding a pipeline arithmetic processing part. It is a figure which shows the process of each node.

Explanation of symbols

1a to 1c: ECU, 2: ECU, 3: Host PC, 4: Information management node,
5: I / O node 6: First operation node 7: Second operation node
8: 3rd operation node, 9: Super high speed node, 10: Pipeline operation processing part,
11: Communication network

Claims (4)

A real-time parallel distributed simulation system that targets a predetermined control object composed of a plurality of partial control objects,
A plurality of control devices capable of controlling each of the partial control objects by outputting a control signal corresponding to the state information of each of the partial control objects to the partial control object;
One I / O node that inputs the control signals from the plurality of control devices and outputs pseudo state information that is pseudo state information of the partial control object to the plurality of control devices. When,
A plurality of operation nodes that generate each of the pseudo-state information based on each of the control signals input from the I / O node, and each include an operation processing device;
Communication means for communicating between the I / O node and the plurality of operation nodes and between the operation nodes at predetermined time intervals;
With
At least one of the operational node, time management of communications between a plurality of the operational node, and have rows time management of communication between the I / O node and the calculation nodes,
Following the communication between the I / O node and the plurality of operation nodes at each predetermined time and between the operation nodes, the plurality of operation nodes generate the pseudo state information, and I / O node, real-time parallel distributed simulation system, wherein line Ukoto output processing of generating the parallel plurality of said control device of the pseudo status information by the operation node. A host computer for storing pseudo-basic information about the control object;
Information for inputting the pseudo-basic information from the host computer and outputting the pseudo-basic information to the computation node, and inputting the pseudo-state information from the computation node and outputting the pseudo-state information to the host computer A management node;
Further comprising
The operation node generates the pseudo state information based on the pseudo basic information and the control signal,
The real-time parallel distributed simulation system according to claim 1, wherein the communication unit communicates between the information management node, the I / O node, and the computing node at predetermined time intervals. It further includes a pipeline operation processing unit configured by a plurality of nodes and performing pipeline operations,
The real-time parallel distributed simulation system according to claim 1 , wherein the communication unit communicates between the I / O node and the pipeline arithmetic processing unit at predetermined time intervals. The real-time parallel distributed simulation system according to claim 3 , wherein the pipeline arithmetic processing unit analyzes the pseudo state information by the pipeline arithmetic.

Images