JP4708136B2 - Simulation execution method - Google Patents

Description

  This invention aims to shorten the total calculation time by thinning out unnecessary simulations using the results of the first execution when the simulation is executed for the second and subsequent times in a Monte Carlo simulation mainly for moving objects. The present invention relates to a simulation execution method.

  In general, in a computer simulation including a stochastic element, it is necessary to obtain a statistical result by trying the simulation many times. This type of simulation is called Monte Carlo simulation. In recent years, as the cost performance of computers has improved, simulations have become larger and more detailed. When the required performance cannot be satisfied with one computer, a method using a parallel distributed simulation technique using a number of computers is widely used if parallelism is inherent in the processing. When a Monte Carlo simulation is executed in a simulation that requires a large amount of calculation, the total execution time becomes serious.

  On the other hand, in a simulation in which a large number of objects (mainly moving objects) influence each other, as the number of the objects increases, each other affects each other in a complex manner, so even in a deterministic case where there is no stochastic element, It becomes difficult to calculate which event occurs at which time. In such a type of simulation, it is common to use a time step method in which a minute time interval (Δt) for advancing the simulation time is set and the time is advanced at the timing of Δt (for example, non-stepping). Patent Document 1).

Japanese Patent Laid-Open No. 7-71895 A. Ozaki, M. Furuichi, K. Takahashi, and H. Matsukawa, "Design and Implementation of Parallel and Distributed Wargame Simulation System and Its Evaluation" IEICE Trans.IEICE / IEEE Joint Special Issue Vol.E84-D No.10, pp.1376-1384, October 2001.

  In the parallel distributed simulation based on the conventional time step method as described above, contrivances such as changing Δt according to the moving speed of the object are taken, but as a result, unnecessary simulations are included depending on the situation. When the number of objects is very large or when it is necessary to simulate in more detail, there is a problem that a large overhead occurs.

  The present invention has been made to solve the above-described problems. The purpose of the present invention is to use the result executed for the first time when executing the Monte Carlo simulation, so that the second and subsequent times are unnecessary as a result. To obtain a simulation execution method that can reduce the total execution time.

In the Monte Carlo simulation based on the time step method, the simulation execution method according to the present invention includes a first step of executing a simulation of a scenario relating to a moving object by setting a first time step size for the first time, and the second and subsequent times. And a second step of executing a simulation of a scenario related to the moving object by omitting the simulation based on the first time interval between events whose occurrence times are known in the first simulation . Step 2 is between events whose occurrence times are known in the first simulation, and the moving object is not affected by other objects, or the moving object does not affect other objects. , And the period during which the specification value of the moving object is unchanged or constantly changing It is intended to eliminate the simulation based on definitive the first time step size.

  When executing the Monte Carlo simulation, the simulation execution method according to the present invention uses the result executed for the first time, and deletes simulations that become unnecessary as a result after the second time, thereby reducing the total execution time. There is an effect that can be done.

Embodiment 1 FIG.
A simulation execution method according to Embodiment 1 of the present invention will be described with reference to FIGS.

  The simulation execution method according to the first embodiment of the present invention is a computer simulation based on a time step method in which a large number of moving objects (MOs) such as aircraft, vehicles, and people appear, and when executing a Monte Carlo simulation, By using the results executed for the first time, simulations that are unnecessary as a result after the second time are deleted, and the total execution time is shortened. According to the present invention, even when a parallel distributed simulation technique is required, the same can be realized with a smaller number of computers. Here, “simulation” refers to a process of updating a state of a moving object at a timing of a minute time step width Δt. Therefore, deleting the above simulation is equivalent to increasing the time interval Δt.

  It should be noted that the simulation that can be deleted is limited to a case where a stochastic element is not included in the event group that affects the simulation by the time when the simulation is executed. This is because the probability value may change due to the Monte Carlo method, resulting in a simulation that cannot be deleted.

  Next, the operation of the simulation execution method according to the first embodiment will be described with reference to the drawings. FIG. 1 is a diagram showing an example scenario to be evaluated by the simulation execution method according to the first embodiment of the present invention. FIG. 2 is a timing chart showing a simulation execution method using a time step method with respect to the example scenario of FIG. Further, FIG. 3 is a timing chart showing a simulation execution method according to the first embodiment of the present invention with respect to the scenario example of FIG. In addition, in each figure, the same code | symbol shows the same or equivalent part.

  FIG. 1 shows an example scenario in which the tank 1 (moving object) starts from the starting point and travels through the A point, the obstacle 11, the B point, the C point, the obstacle 12, and the D point. Is. In this case, since there are no obstacles from the start point to the A point, the arrival time to the A point can be calculated (predictable) from the specification information such as the speed and acceleration of the tank 1. However, there is an obstacle 11 from the point A to the point B, and the tank 1 travels while accelerating and decelerating according to the state of the obstacle 11, so that the arrival time at the point B is difficult to calculate (difficult to predict). Therefore, since the arrival time at the point B is not known, the arrival times at the subsequent points C and D are also unknown.

  In such a case, as shown in FIG. 2, a time step method in which a fine time step Δt1 (one arrow) is set, a simulation is performed based on the time step Δt1, and the time is advanced is used. It is common. In this time step method, the simulation result is rounded with the accuracy of the time interval Δt1, but it is not necessary to calculate the time when the next event occurs (for example, the arrival time at the point B). Further, since calculation from the B point to the C point can be performed in the same manner as from the start point to the A point, the simulation based on the fine time increment Δt1 during that period (from the B point to the C point) can be deleted. .

  Therefore, when executing the Monte Carlo simulation based on the scenario example of FIG. 1, in the first execution, as shown in FIG. 2, the tank 1 departs from the start point based on the fine time increment Δt1. It is necessary to execute a simulation that travels via points A, B, C, and D. However, from the second time onwards, as shown in FIG. 3 (a) or (b), the arrival times at point B, point C, point D as well as point A are known. The simulation can be executed by deleting the step width Δt1.

  Even if the arrival time at each point is known, if you want to more accurately simulate the fuel consumption state from point A to point B or from point C to point D, for example, As shown in FIG. 3A, it is also conceivable that simulation is executed with a small time step width Δt1 between the above points, and the fine time step width Δt1 is deleted between other points. Further, for example, if it is not necessary to simulate the fuel consumption state, as shown in FIG. 3 (b), it is conceivable that the detailed time interval Δt1 between all points is deleted and simulated.

  That is, in the simulation execution method according to the first embodiment, it is difficult to calculate the occurrence time of an event that occurs during the simulation in a computer simulation mainly for a plurality of moving objects in the real world. When a large number of trials are performed based on the Monte Carlo method in a type of simulation that is executed at time steps by setting a step size Δt1, the time of occurrence of the event that is known by executing the simulation for the first time is Even if the simulation is retried, if there is no stochastic element up to that time and it remains unchanged, the same event occurrence time information will be used for the second and subsequent similar simulations. As a result, the simulation based on the fine time interval Δt1 between the events that becomes unnecessary is performed. It is something to omit.

  Furthermore, when the object (moving object and non-moving object) is not affected by other objects, or when the object does not affect other objects, various specification values of the object are unchanged or constantly changed. For example, during the period of time, for example, from the start point to the point A in FIG. 3A or from the point B to the point C, the simulation based on the fine time interval Δt1 is omitted.

Embodiment 2. FIG.
A simulation execution method according to Embodiment 2 of the present invention will be described with reference to FIGS.

  FIG. 4 is a diagram showing an example scenario to be evaluated by the simulation execution method according to the second embodiment of the present invention. FIG. 5 is a timing chart showing an example of a simulation execution method according to the second embodiment of the present invention with respect to the scenario example of FIG. FIG. 6 is a timing chart showing another example of the simulation execution method according to the second embodiment of the present invention with respect to the scenario example of FIG.

  In FIG. 4, the tank 1 departs from the start point and travels through the points A, 11, 11, B, C, 12 and D, and the tank 2 departs from the start point. An example scenario is shown in which the vehicle travels via point E, obstacle 13, and point F, and then two tanks 1 and 2 engage.

  For example, as shown in FIG. 4, when a scenario in which two tanks 1 and 2 are engaged is evaluated, the simulation is based on a time interval Δt2 that is finer than the time interval Δt1. Sometimes it is necessary. In such a case, if the meeting time of the tank 1 and the tank 2 (time for starting the battle) becomes known in the first simulation, in the second and subsequent simulations, the time increment Δt2 is further reduced from the battle start time. It is possible to perform a simulation based on this.

  Also, when the engagement start time is determined based on a probabilistic element, and the engagement start point can be almost specified, from the point where the closest arrival time until reaching that point is known, that is, the tank As shown in FIG. 5, it is conceivable to perform simulation based on a finer time step width Δt2 from point D in the case of 1 and point F in the case of tank 2 as shown in FIG.

  FIG. 6 is an example of a time chart when the engagement start time is invariable (when it can be specified), and the simulation is performed based on a further smaller time interval Δt2 from the engagement start time known in the first simulation. The case where it performs is shown.

  In other words, the simulation execution method according to the second embodiment not only thins out simulations that are unnecessary as a result of using the results of the first execution after the second time, but conversely reduces the time increment Δt2. With respect to the period for which it is determined that the simulation is necessary, the simulation is performed based on a finer time step width Δt2.

  In other words, the period affected by or affecting other objects is simulated based on the finer time increment Δt2.

Embodiment 3 FIG.
A simulation execution method according to Embodiment 3 of the present invention will be described with reference to FIGS.

  FIG. 7 is a diagram showing an example scenario to be evaluated by the simulation execution method according to the third embodiment of the present invention. FIG. 8 is a timing chart showing a simulation execution method according to Embodiment 3 of the present invention with respect to the scenario example of FIG.

  FIG. 7 shows a scenario example in which only the initial state and scenario of the tank 2 are changed from the scenario example shown in FIG. When the time when the tank 1 scenario is affected by this change is known, the period until the tank 1 is affected can be executed based on the time increment Δt1 used in the scenario of FIG. Conceivable.

  Further, the tank 2 needs to perform a simulation based on the fine time step Δt1 in the first execution, but the tank 1 and the tank 2 are unnecessary as a result based on the first simulation result after the second time. It is possible to delete the simulation based on the time interval Δt1.

  FIG. 8 shows an example of a timing chart when the simulation is executed for the second and subsequent times in the scenario example of FIG. Since the engagement start time is uncertain because it includes a stochastic element, the time closest to the engagement start point, that is, the point D for tank 1 and the time H for tank 2 is a finer time. This shows a case where simulation based on the step size Δt2 is performed.

  That is, in the simulation execution method according to the third embodiment, when the specification value and the scenario of the moving object appearing in the simulation are changed, up to the point in time that does not affect the change in other unchanged objects. The simulation is thinned out in the same manner as in the first embodiment using the information (occurrence time of various events) obtained from the simulation result already (first time).

Embodiment 4 FIG.
A simulation execution method according to Embodiment 4 of the present invention will be described with reference to FIG.

  FIG. 9 is a diagram showing an example scenario to be evaluated by the simulation execution method according to the fourth embodiment of the present invention.

  In FIG. 9, three tanks 3, 4, and 5 start from the starting point and travel through the obstacle 14, the obstacle 15, and the obstacle 16, and the three tanks 6, 7, and 8 start. A scenario example is shown in the case where the vehicle departs from the point and travels through the obstacle 18 and the obstacle 17 and then the six tanks 3 to 8 engage.

  FIG. 9 shows an example scenario when the number of moving objects increases. Thus, as the number of moving objects to be simulated increases, the effect of shortening the total execution time of the Monte Carlo simulation based on this idea increases.

It is a figure which shows the example scenario of the evaluation object of the simulation execution method which concerns on Embodiment 1 of this invention. It is a timing chart which shows the simulation execution method using the time step method with respect to the example scenario of FIG. 5 is a timing chart showing a simulation execution method according to Embodiment 1 of the present invention with respect to the scenario example of FIG. It is a figure which shows the example of a scenario of the evaluation object of the simulation execution method which concerns on Embodiment 2 of this invention. 5 is a timing chart showing an example of a simulation execution method according to Embodiment 2 of the present invention with respect to the scenario example of FIG. 7 is a timing chart showing another example of the simulation execution method according to the second embodiment of the present invention with respect to the scenario example of FIG. It is a figure which shows the example scenario of the evaluation object of the simulation execution method which concerns on Embodiment 3 of this invention. FIG. 9 is a timing chart showing a simulation execution method according to Embodiment 3 of the present invention with respect to the scenario example of FIG. 7. FIG. It is a figure which shows the example of a scenario of the evaluation object of the simulation execution method which concerns on Embodiment 4 of this invention.

Explanation of symbols

  1, 2, 3, 4, 5, 6, 7, 8 Tanks, 11, 12, 13, 14, 15, 16, 17, 18 Obstacles.

Claims (4)

In Monte Carlo simulation based on the time step method,
A first step of setting a first time step and executing a simulation of a scenario relating to a moving object for the first time;
A second step of executing a simulation of a scenario relating to the moving object, omitting a simulation based on the first time interval between events whose occurrence times are known in the first simulation after the second time;
Including
The second step is between events whose occurrence times are known in the first simulation, and the moving object is not affected by other objects, or the moving object does not affect other objects. In this case, the simulation execution method is characterized in that the simulation based on the first time step size in a period in which the specification value of the moving object is unchanged or constantly changed is omitted. In the second step, the simulation is executed based on the second time step size for a period in which the simulation based on the second time step size smaller than the first time step size is necessary. The simulation execution method according to claim 1. In the second step, a simulation is executed based on the second time step size with respect to a period in which the moving object is influenced by another object or the moving object affects another object. The simulation execution method according to claim 2, wherein: The second step is characterized in that, when a part of the scenario is changed, a simulation based on the first time step size is omitted for a moving object that is not involved in the change until a time point that does not affect the change. Item 2. The simulation execution method according to Item 1.

Images