JP6420972B2 - Train control system design simulator - Google Patents

Description

The present invention relates to a simulator for designing the train control system.

  In train control, safety is required for traveling a plurality of trains traveling on the same traveling path without causing a rear-end collision, and availability for increasing the operation rate. Therefore, when designing a train control system, the greatest requirement is to ensure safety and availability at the lowest possible cost.

  Train control systems can be broadly classified into track circuit systems and wireless train control systems, depending on the control system. In addition, there are ground position detection methods and on-vehicle position detection methods for on-line position detection methods of trains, a method using a track circuit as a representative example of the former, and a speed generator mounted on a train as a representative example of the latter Can be used.

  Wireless train control systems have recently been implemented in some areas. In the wireless train control system, moving blockage can be realized by adopting the on-vehicle position detection method. In the on-vehicle position detection type wireless train control system, position information calculated on the vehicle is wirelessly transmitted to the ground device. The ground device calculates the route and stop target position of each train based on the position information transmitted from all trains, and transmits the information wirelessly to each train. The speed control device on the vehicle creates an operation speed pattern (speed check pattern) based on the information transmitted from the ground device, and performs brake control so as not to exceed this pattern.

  The wireless train control system wirelessly exchanges information such as train position and speed check pattern, so the ground facilities can be greatly simplified and maintenance can be reduced, and detailed train control is possible. .

  In the on-vehicle position detection method using a speed generator, the existing line position is calculated (estimated) based on the number of pulses generated by the speed generator and the wheel diameter. However, position detection errors accumulate due to wheel diameter reduction or wheel idling, gliding, and other environmental conditions.For example, error accumulation can be reset by installing position correction ground elements at predetermined intervals. It corresponds.

  In addition to speed generators, position detection devices such as GPS and acceleration sensors are in practical use or in the stage of study for practical use. However, these position detection devices are not necessarily established with techniques and knowledge relating to error quantification, which contributes to an excessive safety margin.

  The design of the conventional train control system will be described by taking the design of the position of the position correction ground element as an example. In general, the installation position of the position correcting ground element is designed so that the accumulated error of the train position by the position sensor such as a speed generator is below a certain allowable value. The reason why the allowable value is provided for the accumulated error of the train position is that when an accumulated error exceeding the allowable value occurs, dangerous events such as derailment and rear-end collision may occur with an unacceptable probability. In practice, in order to ensure safety, measures such as emergency braking of the train are taken when the accumulated error estimated by the train exceeds a certain value. Therefore, when a cumulative error exceeding the allowable value occurs, stable transportation is hindered even if safety is ensured, and availability is reduced.

FIG. 9 is a diagram schematically showing the transition of position detection error in the on-vehicle detection method using a speed generator. As the travel distance from the position correction ground element PD of the train T increases, the standing line position estimated values are distributed over a wide position range, such as X1, X2, X3. Dispersion increases). Therefore, in the design of the conventional train control system, the installation position of the position correction ground element PD shown in FIG. However, the shape of the distribution of the on-line positions varies in a complicated manner depending on the operation status (acceleration / deceleration) of the train and the track environment. Therefore, the prior art does not directly evaluate the spread of the distribution of existing line positions, and generally, a cumulative error occurs at a certain rate (for example, ± 1%) that allows for a safety margin with respect to the travel distance. It is designed as a thing. As a result, the optimum safety margin cannot be quantitatively estimated, which may lead to an increase in equipment cost due to an excessive safety margin and a decrease in availability due to an excessive safety margin. Relocation and expansion of ground elements can improve the problem of reduced availability, but there is a problem that relocation and expansion of ground elements require construction cost and construction period.

  Non-Patent Document 1 discusses a technique using a Kalman filter and a GA (genetic algorithm) as a technique for realizing optimization of the ground element (varis) arrangement. However, Non-Patent Document 1 does not model idling / sliding, which is the largest error factor in position detection, and has a great influence on the spread of the distribution of existing line positions. On the other hand, in Non-Patent Document 1, since the error between the true value and the estimated value by the model becomes large due to idling / sliding in a place where the acceleration / deceleration is large, it is appropriate to place the ground element in a place where the acceleration / deceleration is large. It is concluded that.

  For idling / sliding, re-adhesion control that suppresses them by electric motor control has been put into practical use. In addition, a dynamic model of vehicle motion that can be applied to the study of re-adhesion control has been proposed, and simulation is possible (for example, see Non-Patent Document 2).

  Patent Document 1 discloses a moving body control system that calculates an optimal stop target by evaluating position estimation accuracy in addition to position estimation. In this system, an optimal stop target is calculated following the changing estimation accuracy. When this method is applied to a train control system using a speed generator and a position correcting ground element, it can be expected to realize an optimal train operation under a given condition of the ground element arrangement. On the other hand, Patent Document 1 does not mention an arrangement method of the ground unit for realizing the expected train operation.

  In addition, as another example of the design of the conventional train control system, a station radio station station design will be described. Some conventional wireless train control systems make an emergency stop of a train when a communication failure continues for a certain time or more in order to ensure safety. Therefore, in order to realize high availability, it is necessary to design the station placement and channel setting so that the dead zone can be eliminated and the handover can be performed reliably.

  As a technique for supporting station design, Patent Document 2 uses a reception strength characteristic, a delay characteristic, and a noise / interference characteristic as evaluation parameters, and a base for realizing an optimal base station arrangement design for an arbitrary communication environment. A station coverage area estimation method is shown. Further, Patent Document 3 discloses a technique for calculating reception power of all area evaluation points for each base station candidate point by simulation and calculating an optimal base station arrangement pattern by applying GA.

  Further, in general systems including uncertainty, there are examples in which probabilistic reasoning using a Bayesian network or the like and probabilistic simulation (Monte Carlo simulation) are used for design and analysis. For example, in Patent Document 4, as a method for handling uncertainty regarding train operation, a probability that a delay state at a station can be obtained is obtained by statistical processing to create a Bayesian network. Techniques for estimating or analyzing how it spreads are shown. Moreover, in patent document 5, in order to implement | achieve the optimal arrangement | positioning of the measuring device in the measurement of the retaining wall in excavation construction, the Monte Carlo simulation is utilized.

  In the mathematical optimization problem, metaheuristic techniques such as simulated annealing and particle swarm optimization have been developed and utilized, for example, GA used in Patent Document 3, Non-Patent Document 1, and the like.

JP 2012-144068 A Japanese Patent Laid-Open No. 2001-160982 JP 2001-285923 A Japanese Patent No. 5080422 JP2013-242175A

Mohammad Ali Sandidzadeh, Ali Khodadadi, “Optimization of balise placement in a railway track using a vehicle, an odometer and genetic algorithm” (Vehicle, odometer, and optimization of ballistic placement on railroad tracks using genetic logic) , Journal of scientific and industrial research, (India), March 2011, Vol. 70, No. 3, p. 210-214 Masatake Sato and five others "A study on the calculation method of torque reduction in idling re-adhesion control", Proceedings of the Symposium of the Railway Technology Association, 2011, Vol. 18, p. 301-304

  In designing a train control system, the greatest requirement is to ensure safety and availability at the lowest possible cost. Adoption of on-vehicle position detection and wireless train control meet the above requirements, but in many cases there is a trade-off relationship between cost, safety and availability. For example, in wireless train control, there is a problem that the cost of equipment increases when the number of wireless base stations is increased or the required performance of equipment is increased in order to achieve high safety and availability.

  Therefore, a design method for a train control system that realizes low cost, high safety, and high availability, that is, an optimum design is desired.

  In order for the train control system to fulfill the purpose of safe and reliable control of train travel, it is important to properly design each component of the train control system. Factors that should be considered in the design include the train status, ground equipment status, and track-side status. The train state includes, for example, drive device performance, brake performance, degree of aging, speed generator performance, and wheel wear. The ground equipment status includes the blockage boundary, the position of the position correction ground element, the position of the ground traffic signal, the position of the level crossing controller, the level crossing device, and the like. In addition, there are track grades, curve shapes, railroad crossing forms, rockfall hazard zones, tunnels, types of bridges, environment, etc.

  When designing a train control system, it is necessary to consider the various factors described above to ensure safety.

  However, for example, an on-vehicle position detection method using a speed generator, a GPS, an acceleration sensor, or the like is difficult to verify practical safety because accuracy and accuracy change dynamically. Therefore, an excessive safety margin distance tends to be set. In addition, the uncertainty of the position detection method makes it difficult to optimally design the position correction ground element installation position, blockage boundary, speed pattern, etc., which is inseparable from the position detection method.

  The method based on the error between the true value and the estimated value by the model in Non-Patent Document 1 is effective when the true value is given deterministically. However, the effect is limited in light of the fact that idling / sliding occurs probabilistically indeterminately.

  In addition, in the conventional radio base station placement design support technology such as Patent Document 2, the examination range is closed by the communication system, and only the evaluation of whether communication is possible or the communication quality satisfies the specified value is performed. . In other words, the specific influence of communication availability and communication quality on the application of a communication system called radio train control is not considered. This is functionally insufficient when designing a system that changes train control parameters and control methods according to communication quality.

  Probability estimation and stochastic simulation are effective in dealing with the uncertainty of train control systems. However, the application range of the probabilistic simulation related to the conventional train control is limited to the level of the operation schedule and the operation curve, and the stochastic simulation regarding the operation of the train control system and its constituent devices has not been sufficiently studied.

  In designing, not only verification based on simulation but also verification using actual equipment is effective. However, in the verification with an actual machine, the worst condition that is the most dangerous side is difficult to reproduce. Even if it can be technically reproduced, it may be difficult to implement because the equipment used for verification is expensive and the test involves danger. In addition, if a problem is discovered at the actual machine verification stage, the redesign (development) of design / development is large, which may lead to an increase in design / development period and cost.

The present invention is optimally designed to the realization, and for using the designing method of the train control system that can be made shortening design and development time at an early stage of design and development, also the train control and to provide a design for a simulator for checking the reasonableness of the design of the system.

The present invention is a simulator for designing a train control system,
The design simulator expresses the train control system with a statistical model, inputs design parameters, performs simulation using the statistical model, analyzes the error of the simulation result with a time series filter, acquires the probability distribution, the resulting probability distribution verify that satisfies the requirements, by repeating simulation and verification while changing the design parameters, the train control that determine the design parameters when a probability distribution satisfies the requirements as appropriate design parameters To use the design method of the system ,
The design parameter input device for inputting design parameters for design object, and a data output device for outputting read fixed data performance and running pattern of the train from the database, the design parameters inputted from the design parameter input device and fixed data input from the data output device, using the time series filter and sensor observation device to get likened to the observed value of the sensor used for estimating a state according to the design target, the acquired sensor observations analysis a state estimating device for estimating the probability distribution of the state of the design object and is characterized by having a monitor device for displaying the estimation result of the state estimation device.
The state relating to the design object is, for example, a train state or an environmental state.
The design parameter input device includes a design target designating unit for designating a design target and a design parameter input unit for inputting a design parameter.

That is, the design method of the train control system is characterized in that a substantial safety margin is clarified by simulation by expressing the train control system with a statistical model. Then, design parameters (such as ground position and radio base station position) are input and simulation is performed using a statistical model of the train control system. Variations in the results due to various uncertainties at this time are considered as probability distributions that are satisfied by conditions such as the train state by a time series filter. Then, it is verified whether or not the simulation result satisfies a requirement (probability of reaching a dangerous state is equal to or less than a threshold value). Furthermore, simulation and verification are repeated while changing the design parameter, and the design parameter that satisfies the requirements is determined as an appropriate design parameter. Furthermore, in order to obtain appropriate design parameters in a short time and with little effort, it is desirable to optimize the design parameters by combining metaheuristic techniques (such as GA).

  Since the sensor observation apparatus and the state estimation apparatus are used for computer simulation, they are not actual machines but simulation machines (numerical models). However, for the sake of clarity, simulated articles are omitted below. However, if it is more beneficial to use a real machine than to design and use a mock machine, HILS (Hardware-In-the-Loop Simulation) technology that incorporates the real machine into the simulation may be applied.

  When the design object is not related to train interval control, the sensor observation device and the state estimation device are each provided only with the device for the own train.

When the design target is related to train interval control, the sensor observation device is provided with a preceding train sensor observation device and a subsequent train sensor observation device. The preceding train sensor observation device receives the preceding train fixed data from the data output device and the preceding train design parameter from the design parameter input device. The subsequent train sensor observation device receives the subsequent train fixed data from the data output device and the subsequent train design parameters from the design parameter input device. Then, fixed data and design parameters are input to each sensor observation device is obtained likened to the observed value of the sensor used for estimating a state according to the design target (e.g. estimate of the actual train status or environmental conditions), Input to the state estimation device.

If the design target is related to the spacing control of the train, even in a state estimating apparatus, preceding train train state (state according to the design object including environmental state. Hereinafter, the same.) And estimates the preceding train state estimation apparatus And a subsequent train state estimating device for estimating the train state of the subsequent train. The preceding train state estimation device analyzes the sensor observation value acquired from the preceding train sensor observation device using a time series filter, and estimates the train state of the preceding train as a probability distribution. The subsequent train state estimation device analyzes the sensor observation value acquired from the sensor observation device for subsequent trains using a time series filter, and estimates the train state of the subsequent train as a probability distribution.

  The simulation result, that is, the train state of the own train or the preceding train and the succeeding train estimated by the state estimation device is input to the monitor device, and converted to a curve graph or contour diagram representing the probability distribution of the train state by the monitor device. And displayed for evaluation by simulation personnel. There is a certain error in the sensor observation value, and the error fluctuates according to the train position. Therefore, curve graphs and contour diagrams representing the probability distribution of train states also vary. The simulation staff can determine whether or not the input design parameters are appropriate from the display state of the curve graph and the contour diagram representing the probability distribution of the preceding train and the succeeding train displayed on the monitor device. The design parameter when it is determined to be appropriate can be determined as the formal design parameter.

  The simulation result may also be input to a speed control device that controls the speed of the subsequent train based on the estimated train state of the preceding train and the train state of the subsequent train. This speed control device may also be a simulator. Then, the control content may be input to another monitor device and displayed for evaluation by the simulation staff.

  The monitoring device for evaluation by the simulation staff may be replaced with an evaluation device for evaluation by a computer. An evaluation function corresponding to the evaluation of the simulation clerk is given in advance to the evaluation device in a form that can be calculated by a computer. The evaluation device applies an evaluation function to the simulation result. By using the meta-heuristic method and feeding back the result of the evaluation function to the design parameter, it is not necessary to spend the labor of the simulation staff in order to obtain an appropriate design parameter.

The simulator for designing a train control system according to the present invention expresses a train control system as a statistical model, inputs design parameters, performs simulation using the statistical model, and analyzes the error of the simulation result using a time series filter. To obtain a probability distribution, verify whether the obtained probability distribution meets the requirements, repeat the simulation and verification while changing the design parameters, and design the design parameters appropriately when the probability distribution meets the requirements Since the train control system design method determined as a parameter is used, by expressing the train control system with a statistical model, the substantial safety margin is clarified by simulation, and is optimal in the early stages of design and development. Realizing design and shortening the design and development period Kill.
In addition, it becomes possible to discuss safety and availability more quantitatively. And the reduction in rework of design and development can be expected to shorten the design and development period. Furthermore, abstracting train control systems with statistical models can improve the reusability of statistical models and elemental technologies and contribute to shortening the design and development period.
Furthermore, since the design parameter input device includes a design target designating unit for designating a design target and a design parameter input unit for inputting a design parameter, the suitability of the design parameter can be confirmed for each design target. A unique effect is obtained.

It is a flowchart which shows the basic concept of the design method of the train control system which concerns on this invention. It is a block diagram which shows an example in case the design object is not related to space | interval control among the structures of the simulator for design of the train control system which concerns on this invention. It is a block diagram which shows an example in case the design object is related to space | interval control among the structures of the simulator for design of the train control system which concerns on this invention. It is a figure explaining the design thought in the case of applying the probability theory by a time series filter to the installation design of the ground element for position correction. It is a flowchart explaining operation | movement of the simulator for design of FIG. FIG. 5 is a view similar to FIG. 4 for explaining a method for determining a design parameter when the design target is the installation position of the position correction ground element. It is a figure explaining the two-dimensional table expression of the relationship of idling / sliding duration with respect to speed and acceleration. It is a figure which shows the relationship between the speed at the time of assuming idling / sliding and the noise resulting from idling / sliding. It is a figure which shows typically transition of the error of the position detection in the on-vehicle detection system using a speed generator.

  Next, embodiments of the present invention will be described with reference to the drawings.

The design method of the train control system according to the present embodiment is performed by executing a design process based on a predetermined program by a computer. As shown in FIG. 1, first, the train control system is expressed by a statistical model (S1. S means a step. The same applies hereinafter). Next, the simulation staff inputs design parameters for the train control system (S2). Design parameters vary depending on the design object. Installation position of the design target, for example, the position correcting ground coil, if installed position location of the traffic signal, and the kilometrage from each line start point is entered for example by m units. Further, if the design object is, for example, the traveling speed of a train, it is the maximum traveling speed, and for example, speed data in units of m / second is input. If there is a restriction on the value of the design parameter, only the design parameter that satisfies the restriction is input. Alternatively, design parameter candidates may be limited in advance by a method different from the present design method. By not simulating design parameters that are not promising as solutions, computational resources can be used effectively.

Next, simulation of train control is performed using the input design parameters and statistical model (S3). That is, the predetermined train operation program read from the database is executed, and the virtual train is caused to travel at a set speed along the virtual track on the data imitating an actual track pattern. Subsequently, to obtain a probability distribution analyzed using the time-series filters observations obtained by the simulation, for example, the velocity data and position detection data rate generator is calculated based on the number of pulses and the wheel diameter occurring (S4). For example, a particle filter or a Kalman filter can be used as the time series filter. Of these, the particle filter is most preferable. The particle filter will be described in detail later.

  Next, it is determined whether or not the probability distribution obtained by the analysis by the time series filter after inputting the design parameters satisfies the requirement for guaranteeing safety (S5). If it is determined in S5 that the requirement is not satisfied (No in S5), the process returns to S2, and the design parameters are changed and input. Repeat the input value change and simulation until the requirements are met. Whether or not the requirement is satisfied can be determined by, for example, displaying the probability distribution graphically with a probability distribution diagram, or summary statistics of the probability distribution (eg, mean, variance, kurtosis, quantile, mode) It is preferable that the simulation staff can judge by expressing the number by a numeral. Alternatively, display means may be provided such that the obtained probability distribution is compared with a threshold value, and when the threshold value is exceeded, a flap is raised and displayed. If the requirement is satisfied in S5, the input value at that time is determined as an appropriate design parameter (S6). If one design parameter is determined, the design process for the design object is completed.

  When designing another design object, the design process for the design object is performed in the same manner as described above from S1.

The following will describe a train control system design for a simulator that is used to implement the above design method.

Figure 2 shows an example of a configuration of a design for a simulator when designed is not related to the distance control. This design simulator includes a design parameter input device 1, a data output device 2, and the train sensor observation apparatus 3B, and the train of train state estimation device 4B, and a monitor device 5.

Figure 3 shows an example of the configuration of a design for a simulator when designed is related to distance control. The design for simulator, the design parameter input device 1, a data output unit 2, a preceding train sensor observation device 3A, and the subsequent train sensor observation apparatus 3B, the preceding train of the train state estimation device 4A, the subsequent train A train state estimation device 4B and a monitor device 5 are provided. Design simulator of FIG. 3, the simulator design 2, is obtained by more generalized. Therefore, the following description will be given according to the configuration example of FIG.

The design parameter input device 1 includes a design target specifying unit 11 for specifying a design target and a design parameter input unit 12 for inputting a design parameter. Design target specification unit 11, for example, the installation position of the position correcting ground coil is designated as the design object by selecting one of the planting設位location or trains traveling speed of the ground traffic. The design parameters input from the design parameter input unit 12, as already mentioned, the installation position of the design target, for example, the position correcting ground coil, if planted設位location of terrestrial signals, kilometrage from each line origin For example, it is input in m units.

  Further, the design parameter input unit 12 inputs, as a design parameter, a safety margin distance, that is, a margin of a train interval that can travel without collision between the preceding train and the following train traveling on the same traveling route, for example, in m units. be able to.

  The data output device 2 sequentially reads out the running pattern data from a database (not shown) for the designated train operation program. And it outputs at the predetermined | prescribed timing corresponding to the time resolution of simulation, and inputs into sensor observation apparatus 3A for preceding trains, and sensor observation apparatus 3B for succeeding trains. The traveling pattern data is a representation of a set of time and train position as an array that determines the train state. Although it can be calculated from the time and the train position, speed and acceleration may be included in the travel pattern data. Moreover, you may include environmental conditions, such as a train state, such as notch information and a train electric current, and a signal display and the linear of a driving | running | working location as needed. Then, the data output device 2 inputs the design parameters input from the design parameter input unit 12 to the preceding train sensor observation device 3A and the subsequent train sensor observation device 3B. Furthermore, fixed data such as the performance of each train is input to the preceding train sensor observation device 3A and the subsequent train sensor observation device 3B.

  The preceding train sensor observation device 3 </ b> A and the subsequent train sensor observation device 3 </ b> B are simulated versions of the sensor observation device that observes the observation values of the sensors necessary for estimating the train state in the actual train control system. Then, the traveling pattern data input from the data output device 2 is acquired in the sense of sensor observation values necessary for estimating the train state. The observed value of the sensor is, for example, the presence or absence of detection of a pulse from a speed generator necessary for estimating the position of the existing line or a position correction ground element.

  The train state estimation device 4A for the preceding train calculates the train status of the preceding train using the time series filter from the sensor observation value acquired by the sensor observation device 3A for the preceding train, and here, as an example, the probability distribution of the standing line position Then, the current position of the preceding train is estimated. The probability distribution obtained by the train state estimation device 4 </ b> A of the preceding train is output to the monitor device 5. The train state estimation device 4B of the following train analyzes the probability distribution of the train state of the subsequent train from the sensor observation value acquired by the sensor observation device 3B for the following train, using a time series filter, here as an example, Obtain the probability distribution of the position and estimate the position of the following train. Then, the probability distribution obtained by the train state estimation device 4 </ b> A of the following train is also output to the monitor device 5.

  When the monitor device 5 inputs the probability distribution of the train state of the preceding train from the train state estimation device 4A of the preceding train (in this case, the presence probability distribution of the preceding train), the monitor device 5 converts the presence probability distribution into data (image) Process) to create a probability distribution graph. Then, the probability distribution graph is displayed on the screen. Even when the monitoring device 5 inputs the probability distribution of the train state of the following train from the train state estimation device 4B of the following train (in this example, the presence probability distribution of the following train), the monitoring device 5 converts the presence probability distribution into data (image). Process) to create a probability distribution graph. The monitor device 5 displays the probability distribution graph on the screen as shown in FIG.

  FIG. 4 shows the installation position of the position correction ground element from the starting point of the travel path as the design parameter after the design target specifying unit 11 of the design parameter input device 1 specifies the installation position of the position correction ground element as the design target. An example of a probability distribution graph displayed on the screen of the monitor device 5 when a distance of up to about km is input is shown.

  In FIG. 4, the vertical axis represents the standing line probability, and the horizontal axis represents the distance from the starting point of the train travel path. The shape of the on-track probability distribution X of one train is steep immediately after the train passes one position correction ground element, but has already become slow as the distance away from the position correction ground element increases. Stated.

Subsequently, the operation of the simulator for designing the train control system will be described with reference to the flowchart of FIG. In the present embodiment, the design target specifying unit 11 specifies the installation position of the position correction ground element as the design target (S21). Next, in the design parameter input unit 12 , the installation position of each position correction ground element is input in kilometer as one of the design parameters, and thereafter, the safety margin distance D1 is input as another design parameter. Then, a safety margin distance is temporarily set (S22). This safety margin distance D1 is determined so as to be the minimum train interval that can prevent a collision in consideration of the maximum train speed, braking performance, track gradient, and the like.

After inputting the design parameters, the simulation is started (S23). In this simulation, the running pattern data is read from the database to run the virtual train, and the design parameters input from the design parameter input unit 12 are observed as the observation values in the preceding train sensor observation device 3A and the subsequent train sensor observation device 3B. The train state estimation device 4A for the preceding train and the train state estimation device 4B for the succeeding train use the particle filter to analyze the observation values and obtain the presence probability distribution of the preceding train and the succeeding train. Find the integrated value. Finally, the obtained integrated value is compared with a threshold value.

When the simulation (S23) is completed, the estimation results (probability distribution and integrated value data) output by the train state estimation device 4A of the preceding train and the train state estimation device 4B of the subsequent train are output to the monitor device 5. Then, the probability distribution and its integrated value data are subjected to image processing in the image processing unit of the monitor device 5 (S24). That is, the probability distribution and the integrated value thereof is converted to a probability distribution graph is displayed on the display unit of the probability distribution graph ducks Nita apparatus 5 (S25).

FIG. 6 shows an example of the display content of the display unit of the monitor device 5. Xa and Xb are the positions of the two trains immediately after passing through one position correcting ground element PD1 when the preceding train and the following train travel according to a specific traveling pattern while maintaining the temporarily determined safety margin distance D1. Probability distribution. Similarly, Xa ′ and Xb ′ are standing line probability distributions of the two trains immediately before reaching the next position correcting ground element PD2. Xc ′ is an integrated value of the standing line probability distributions Xa ′ and Xb ′, that is, the collision probability distribution of two trains. In FIG. 6, the collision probability Xc is a threshold value TL indicating the necessity of danger avoidance. Is over.

  That is, when the safety margin distance is set to D1, when the distance between the position correction ground elements PD1 and PD2 is set to L1 as in the example of FIG. It can be seen that the distance between the position correction ground elements PD1 and PD2 needs to be smaller than L1. That is, a position where the collision probability Xc does not occur or a position L2 where the collision probability Xc does not exceed the threshold value TL is appropriate as the installation position of the correction ground element PD2.

  As another embodiment, as shown in FIGS. 2 and 3, in addition to the above configuration, a speed control device 6 and a monitor device 7 may be provided. In the speed control device 6 of FIG. 3, as shown in FIG. 6, the probability distribution of the train state of the following train from the train state estimating device 4A of the preceding train and / or the train state estimating device 4B of the following train, in this example Is input the probability distribution Xc ′ corresponding to the standing line probability distributions Xa ′ and Xb ′ and the collision probability that the preceding train and the succeeding train exist at the same position (S26). As a result, the speed control device 6 can limit the speed of the following train so that the mode of the collision probability Xc ′ does not exceed the threshold TL (S27). Speed limits include deceleration and emergency stop.

  The monitor device 7 displays the simulation result, that is, the control content (speed limit) by the simulated speed control device 6 for evaluation by the simulation staff (S28). The simulation clerk can also determine from the displayed control contents whether or not the input design parameters are appropriate.

  In the sensor observation devices 3A and 3B and the train state estimation devices 4A and 4B, a statistical model and its simulation are applied. By applying the statistical model and its simulation, it is possible to reflect more detailed characteristics of the sensor that have not been considered in the past for simplification. By designing the position of ground units with the minimum necessary safety margin, it is possible to reduce the number of ground units and the frequency of emergency stop of trains due to poor position correction.

  For example, if the position sensor is modeled according to a normal distribution N (0, 1) [m] with an average of 0 and variance of 1 when traveling 100 m, the error when traveling 1 km is N (0, 10 ) [M] can be estimated, and the probability that the error falls within a specific range can be calculated. In practice, a normal distribution is not always applicable to an error model. In order to easily handle more complex nonlinear models and models based on experimental statistics, a particle filter may be used.

  Below, the particle filter as an example of the time series filter used in sensor observation apparatus 3A, 3B and train state estimation apparatus 4A, 4B is demonstrated. Here, a simple model of only the speed generator and the position correction ground element is shown by taking the layout design of the position correction ground element as an example. In this example, for simplicity, the interval control is not considered and the configuration of FIG. 2 is adopted.

また、状態ベクトルｘtに基づく拡大状態ベクトルＸｔを以下のように定義する。

ここで、ｋは、速度発電機による速度算出（一定時間パルス計数方式）における平滑化の時間窓に相当する係数である。 The state vector xt at time t is defined as a vector composed of a train position (vehicle position) pt and a speed vt as follows.

Further, an expanded state vector Xt based on the state vector xt is defined as follows.

Here, k is a coefficient corresponding to a smoothing time window in the speed calculation (fixed time pulse counting method) by the speed generator.

The observation vector yt at time t is defined as a vector consisting of the pulse count ct within the smoothing time window of the speed generator and the ground element detection presence / absence bt as follows.

と定義する。ここで、ΔＴはシミュレーションの時間刻み、Randpは積分誤差相当のノイズ、Randｖは加速度および空転・滑走相当のノイズである。 The system model is

It is defined as Here, ΔT is a simulation time step, Randp is a noise equivalent to an integration error, and Randv is a noise equivalent to acceleration and idling / sliding.

と定義する。ここで、ｎは速度発電機の歯数、Ｄは車輪径、Randｃはサンプリングタイミングによる誤差に相当するノイズ、Randｓは空転・滑走に起因するノイズである。また、Baliseはパラメータで指定した位置を、パラメータで指定した速度で通過した際に、地上子を検知するかを真偽値で返す関数である。地上子の検知は、地上子の個体差や環境条件に左右されるため、一定のノイズRandｂを含む。 The observation model is

It is defined as Here, n is the number of teeth of the speed generator, D is the wheel diameter, Randc is noise corresponding to an error due to sampling timing, and Rands is noise caused by idling / sliding. Balise is a function that returns a true / false value indicating whether to detect a ground element when passing through the position specified by the parameter at the speed specified by the parameter. Since the detection of the ground element depends on individual differences of the ground element and environmental conditions, it includes a certain noise Randb.

  In order to determine Rands, the relationship between the speed and acceleration (deceleration) of the idling / sliding occurrence probability and the relationship of the idling / sliding duration when the idling / sliding occurs are defined in a form that can be handled by a computer. For example, the relationship between the idling / sliding duration is created as a database as a two-dimensional table of speed and acceleration as shown in FIG. As a method for acquiring the relationship between the idling / sliding occurrence probability and the relationship between the idling / sliding duration when the idling / sliding occurs, there is a method of obtaining statistics based on past traveling results. Further, a theoretical formula or simulation result of a mechanical model as shown in Non-Patent Document 2 may be used. Here, modeling is performed by adopting acceleration (difference between two speeds) as general information, but notch information may be used instead of acceleration if, for example, notch information can be observed or estimated.

  However, each time Rands is referenced, it is not practical from the viewpoint of calculation load to calculate the likelihood by probabilistically assuming various idling / sliding states each time. Therefore, various types of idling / sliding states are probabilistically assumed in advance, and an approximate distribution of Rands is obtained. FIG. 8 schematically shows the relationship between the speed and Rands when a specific idling / sliding is assumed.

  The train state estimation device 4B generates a predetermined number of particles represented by the expanded state vector Xt as an initialization process. The set of these particles represents a Monte Carlo approximation of the distribution of the expanded state vector Xt. The initial value of the particles is, for example, on the premise that the train is stopped, the train position is a value within a specific range, and the speed is zero.

  The train state estimation device 4B predicts the state vector xt at the time t from the state vector xt-1 at the previous time t-1 according to the system model expressed by Equation 4 for each simulation period, and further expands the state vector. Xt is obtained from Xt-1. This operation is applied to all particles.

  The sensor observation device 3B generates an observation vector yt from the travel pattern (time, position) and the position correction ground element installation position list for each simulation period, and inputs the observation vector yt to the train state estimation device 4B.

  The train state estimation device 4B performs particle filtering using the observation vector yt input from the sensor observation device 3B. In the filtering, the probability that each particle state can be obtained under the condition that the likelihood of each particle, that is, yt, is given based on the observation model. This is essentially synonymous with obtaining the probability that yt is observed under the given condition if the weight of each particle is uniform.

  The train state estimation device 4B performs particle resampling based on the calculated likelihood (likelihood ratio). Thereby, the Monte Carlo approximation of the distribution of the train position pt at time t is obtained. That is, a standing line probability distribution is obtained. This distribution is input to the monitor device 5.

  The monitor device 5 graphs and displays the standing line probability distribution input from the train state estimation device 4B. At this time, for example, by performing kernel density estimation, a continuous probability distribution can be obtained from the Monte Carlo approximation.

  The simulation clerk determines whether the spread of the standing line probability distribution is within a desired range, thereby determining the quality of the design parameter, that is, the ground element arrangement.

  Although it is not included in the above model for simplicity, some actual train control systems have a function of correcting the speed according to the maximum acceleration or maximum deceleration when idling / sliding is detected. The particle filter has an advantage that it can handle a non-linear model in which the calculation formula changes according to such conditions.

  By applying a statistical model and its simulation, the concept of an allowable value of cumulative error is ultimately eliminated. The reason why the allowable value is provided for the accumulated error is that, when an error exceeding the allowable value occurs, a dangerous event such as derailment or rear-end collision may occur with a possibility / frequency that cannot be permitted. For example, if the train derailment and rear-end collision is expressed by a statistical model and can be quantitatively evaluated as shown in Xc ′ in FIG. 6, a direct discussion on the possibility and frequency of dangerous events in the safety study Is possible. In this case, how many accumulated errors are only one factor leading to a dangerous event is not the essence of safety.

DESCRIPTION OF SYMBOLS 1 Design parameter input device 11 Design object designation | designated part 12 Design parameter input part 2 Data output device 3A Previous train sensor observation device 3B Own train or subsequent train sensor observation device 4A Previous train state estimation device 4B Own train or subsequent train state estimation Device 5 Monitor device 6 Speed control device 7 Monitor device for speed control result display In FIG. 6, D1 safety margin distance Xa, Xa ′ presence probability distribution Xb, Xb ′ of preceding train (last tail) The following train stopped due to emergency braking subsequent train rail probability distribution Xc 'preceding train and trailing train probability distribution TL threshold PD1 which is present in the same position, PD2 position installation interval of the correction ground coil L1, L2 land upper child of (foremost) when

Claims (2)

A simulator for designing a train control system,
The design simulator expresses the train control system as a statistical model, inputs design parameters, performs simulation using the statistical model, analyzes the error of the simulation result using a time series filter, and obtains a probability distribution, the resulting probability distribution verify that satisfies the requirements, by repeating simulation and verification while changing the design parameters, the train control that determine the design parameters when a probability distribution satisfies the requirements as appropriate design parameters To use the design method of the system ,
Design parameter input device and design parameter input and a data output device for outputting read fixed data performance and running pattern of the train from the database, from the design parameter input device for inputting design parameters for design target and fixed data input from the data output device, using a sensor monitoring device to get likened to the observed value of the sensor used for estimating a state according to the design target, the time series filter acquired sensor observations a state estimating device for estimating the probability distribution of the state according to the design target was analyzed, and a monitor device for displaying the estimation result of the state estimation device possess Te,
The design parameter input device, designed simulator, which comprises a design target designating section for designating a design object, a design parameter input unit for inputting the design parameters. Design simulator smell of claim 1 wherein Te,
Wherein the sensor monitoring apparatus, Ri and preceding train sensor observation device and subsequent trains sensor observation device there,
Preceding train for the design parameters from the preceding train fixed data and said design parameter input device from the data output device is input to the preceding train sensor observation device,
Subsequent train for the design parameters from the subsequent train fixed data and said design parameter input device from the data output device is input to the subsequent train sensor observation device,
The fixed data and the design parameters are input to the sensor monitoring unit is acquired to resemble the observed value of the sensor used for estimating a state according to the design target, the together is input to the state estimation device,
Also the state estimation device, a preceding train state estimation device for estimating a state according to the design target of preceding train, Ri subsequent train state estimation device and have to estimate the train state of the subsequent train,
The preceding train state estimation device estimates the state of the are a sensor observations obtained from preceding train sensor monitoring apparatus design target preceding train analyzed using time series filter as a probability distribution,
The subsequent train state estimation apparatus estimating means estimates the state of the are a sensor observations obtained from the subsequent train sensor monitoring apparatus design subject of a subsequent train analyzed using time series filter as a probability distribution design for the simulator to.

Images