JP2015189361A - radio train control method and radio train control system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a radio train control method and system which can realize low cost and high availability by allowing restricted communication failures and seamless degradation operation.SOLUTION: On-vehicle systems 1a, 1b are each equipped with a sensor observation device 11 that acquires a sensor observed value necessary to estimate an own train's train condition, an own train condition estimation device 12 that calculates an own train's train condition on the basis of an acquired sensor observed value, another train condition estimation device 15 that calculates an other train's train condition on the basis of another train's sensor observed value received from a ground device or an other on-vehicle device, and a speed control device 16 that controls an own train on the basis of a calculated own train's train condition and a calculated another train's train condition, thereby making the safety side control of the own train condition estimation device and the other train condition estimation device by estimating train conditions such as position and speed of an own train or another train both with a static model and a time series filter from sensor observed values.

Description

  The present invention transfers a plurality of trains traveling on the same track by exchanging information by wireless communication between a device provided on the ground (ground device) and a device provided on the vehicle (vehicle device). The present invention relates to a radio train control method and a radio train control system that perform control for traveling safely without causing a collision.

  The train control system controls the running speed and stops so that a plurality of trains traveling on the same traveling route can safely travel without colliding. In order to realize this, The position and speed of the track must be detected.

  There are ground detection methods and on-vehicle detection methods for on-line detection methods of trains. The former uses a track circuit as a representative example, and the latter uses a speed generator mounted on a train as a representative example. Can be mentioned.

  In the ground detection method using the track circuit, the track circuit having a certain length is the minimum unit of the closed section. In order to prevent a collision with the preceding train, the following train cannot enter the track circuit where the preceding train exists. Therefore, the operation time interval cannot be shortened compared to the time for the preceding train to pass through the track circuit. Therefore, in order to meet the demand for improvement in operation density, it is necessary to shorten the track circuit length, which increases the equipment cost.

  In the on-vehicle detection method using a speed generator, the position and speed of the track are estimated based on the number of pulses generated by the speed generator and the wheel diameter. Therefore, since higher detection accuracy than the ground detection method using the track circuit can be obtained, the movement blockage can be realized, and the operation interval can be shortened. However, errors may accumulate in the position estimation value due to deviation from the set value of the wheel diameter or wheel idling, sliding and other environmental conditions, so as an example, position correction ground elements should be installed at predetermined intervals. Yes. Therefore, there is a trade-off relationship between the improvement in position detection accuracy, and hence the operation density, and the equipment cost.

  Moreover, in the railway system, a train control system using wireless communication has been developed and partly put into practical use from the viewpoint of cost reduction and higher functionality of train control. In the radio train control system, the train position is managed by radio. In this management method, communication data that summarizes information necessary for control is transmitted and received between the on-board device and the ground device. That is, the on-board device transmits its own train position information to the ground device, and the ground device calculates the position and speed patterns that each train can travel based on the received position information of the multiple trains. Sent to train onboard equipment. Or in the on-board apparatus which received the positional information on the preceding train from the ground apparatus, train control may be performed based on the positional information on the preceding train. In any case, in radio train control, in order to perform safe train control, it is most important to accurately detect the position of the train.

  In the conventional radio train control system, there is a system that makes an emergency stop when a communication failure continues for a certain period of time or when a communication failure continues for a certain number of times or more in order to ensure safety.

  Patent Document 1 discloses a system that enables safe train control by calculating the range of trains using GPS data. However, since the train track probability is generally not uniform even within the calculated track range, there is a possibility that an excessive safety margin distance is set. Patent Document 2 discloses a technique for calculating an optimal stop target by evaluating position estimation accuracy in addition to position estimation. However, the optimum stop target is not always obtained only by the estimated position and the position estimation accuracy. Actually, Patent Document 2 refers to the case where the stop target is lowered from the previous stop target point when the position estimation accuracy of the preceding train or the own train deteriorates, and shows a countermeasure.

  As a technique for improving the availability (operating rate) of the radio train control system, Patent Document 3 shifts to normal operation by accumulating the train ID and train position information when the radio function is down. Techniques for reducing time are shown. However, even if this technology is used, train stops due to communication failures cannot be avoided, and ground facilities are also required. Patent Document 4 discloses a radio train control technique that realizes fail-safe and safe train operation by checking rationality based on continuity of position information or speed information. However, if a communication failure occurs and information cannot be exchanged, a rationality check cannot be performed.

  As a position estimation technique, a position estimation method using a time series filter such as a particle filter (particle filter) or a Kalman filter has been proposed. For example, Patent Document 5 discloses a technique capable of performing highly accurate position estimation of a moving body by using a particle filter. However, in many applications of known position estimation methods using time series filters, the emphasis is on how to obtain an estimate close to the true value. For this reason, the average value, median value, or mode value of the probability distribution calculated by the time series filter has an important meaning. However, when considering train interval control, a safety margin distance must be provided because a rear-end collision risk occurs when the true train position is closer to the estimated position of the preceding train. Therefore, even if the position is estimated with high accuracy using a time series filter, the safety margin distance to be secured is more than the actual required distance in the method of obtaining the estimated position by the average value, median value or mode value. There is still the possibility of becoming excessive.

  A technique called predictive control that controls a train in accordance with the operation status of a preceding train has been proposed as a means to realize early recovery of diamond disturbance in a high-density line section, shortening of operation interval, and energy consumption reduction (for example, Non-Patent Document 1). In this prediction control, the prediction information is given deterministically, and sufficient consideration has not been given to the uncertainty and error of the prediction information. In order to exert the effect of predictive control, it is indispensable to generate accurate predictive information, and practical use involves technical difficulties and increased equipment costs.

Japanese Patent No. 5373761 JP 2012-144068 A Japanese Patent No. 5373761 JP 2013-23104 A Japanese Patent No. 5370122

Hiraguri, Tomii, Hase, “Train Control Method by Predictive Control”, Railway Research Institute Report, June 2003, Vol. 17, No. 6, p. 29-34

  Even when a communication failure occurs in the train, depending on the position and speed of the preceding train and the own train, there is a case where the danger such as a rear-end collision can be avoided without an emergency stop immediately. Therefore, in a conventional wireless train control system that stops a train when a communication failure continues for a certain time or more, availability (operation rate) may be unnecessarily lowered. Also, in order to eliminate dead zones where wireless communication becomes impossible and to realize continuous handover between wireless base stations, increasing the number of wireless base stations or increasing the required performance of equipment increases the cost of equipment. become.

  In view of the above problems, the present invention provides a radio train control method and a radio train control system that can realize a low cost and high availability by allowing limited communication failures and performing seamless degenerate operation. The purpose is to do.

  In order to achieve the above object, a wireless train control method of the present invention includes a position / speed sensor and an on-board radio station on a vehicle, and the on-board radio station is connected to the ground device or the ground radio station and the ground device. , Give and receive information on train status. Based on the information on the train state, a statistical model and a time series filter are used to estimate the train state of the own train and the train state of other trains while maintaining the overall shape of the probability distribution. Based on the above, safe side control is performed.

  In the safety side control, it is possible to know not only the possibility of collision between two front and rear trains but also the probability of collision. Then, the deceleration rate or the stop position can be controlled after ensuring that the collision probability does not exceed a threshold value (for example, 10 −9 / h). Therefore, the safety margin distance can be shortened. In addition, when some or all of the information of some trains is missing due to communication failure, a conditional posterior peripheral distribution is obtained under the observation condition that the information is missing. When information is missing, in many cases, the width and skirt of the probability density distribution at the existing line position are widened and become flat. For this reason, for example, it is possible to seamlessly shift to safe-side control with a larger train interval.

  Further, in order to achieve the above object, the wireless train control system of the present invention has a sensor observation device for acquiring sensor observation values necessary for estimating the train state of the own train in the on-board device, and the acquired sensor observation. The own train state estimation device that calculates the train state of the own train based on the value and the other train state that calculates the train state of the other train based on the sensor observation value of the other train received from the ground device or other on-board device It has an estimation device and a train control device that controls the own train based on the calculated train state of the own train and the train state of the other train, and both the own train state estimation device and the other train state estimation device The probability distribution of the train state such as the current train and other trains from time to time and the future position and speed is analyzed by analysis using a time series filter (for example, particle filter) using sensor observation values. Using the overall shape of the rate distribution, the train speed is controlled by the product of the probability density distribution at the position of the preceding train and the probability density distribution at the position of the own train, or whether the integrated value is larger or smaller than the threshold. It is characterized by doing.

  In the conventional method that does not use the probability distribution, the train traveling speed is controlled on the assumption that the train is not “absolute” in a certain range, so an excessive safety margin distance is required to ensure “absolute”. It was necessary to ensure. On the other hand, in the present invention, the operation can be continued without reducing the speed under the condition that the product of the probability density distribution at the position of the preceding train and the probability density distribution at the position of the own train is small in the future. .

There are a centralized type and a distributed type of radio train control systems that can achieve the above object.
The centralized radio train control system concentrates the observation values of the sensors of each train on the ground device and estimates the train state of all trains in the ground device. In this case, the on-board device is acquired with a sensor observation device that acquires sensor observation values necessary for estimating the train state of the own train, and a transmission device that transmits the acquired sensor observation values to the ground device. The own train state estimation device that calculates the train state of the own train based on the sensor observation value, the receiving device that receives the sensor observation value of the other train from the ground device, and the other based on the received sensor observation value of the other train The other train state estimation device for calculating the train state of the train, and the train control device for controlling the own train based on the calculated train state of the own train and the train state of the other train. In addition, the ground device includes a receiving device that receives sensor observation values from the on-board device, a train state estimation device that calculates the train state of each train based on the received sensor observation values, and a train for each calculated train A transmission device that transmits the state to the on-board device. The own train state estimation device, the other train state estimation device, and the train state estimation device are all characterized in that the train state is calculated using a time series filter based on the sensor observation value.

  The distributed wireless train control system directly transfers sensor observation values or train status information between the preceding train and the succeeding train without centrally estimating the train status of all trains with the ground equipment. Then, the train state of other trains is estimated by being distributed to each train. The following four forms are possible for this distributed type.

  In the distributed wireless train control system according to the first embodiment, the on-board device includes a sensor observation device that acquires sensor observation values necessary for estimating the train state of the own train, and the acquired sensor observation values as ground devices. Transmitting device for transmitting, own train state estimating device for calculating train state of own train based on acquired sensor observation values, receiving device for receiving sensor observation values of other trains via ground device, and other received Another train state estimation device that calculates the train state of the other train based on the sensor observation value of the train, and a train control device that controls the own train based on the calculated train state of the own train and the train state of the other train, and Is provided. Further, the ground device includes a receiving device that receives the sensor observation value from the on-board device of the preceding train or the following train, and a transmission device that transmits the received sensor observation value to the on-board device of the following train or the preceding train. . Each of the own train state estimation device and the other train state estimation device calculates a train state using a time series filter.

  In the distributed wireless train control system according to the second embodiment, the on-board device includes a sensor observation device that acquires sensor observation values necessary for estimating the train state of the own train, and an autonomous device based on the acquired sensor observation values. The own train state estimating device for calculating the train state of the train, the transmitting device for transmitting the calculated train state of the own train to the ground device, the receiving device for receiving the train state of the other train via the ground device, Other train state estimation device that calculates the train state of the other train based on the train state of the other train, and a train control device that controls the own train based on the calculated train state of the own train and the train state of the other train, Is provided. Further, the ground device includes a receiving device that receives the train state from the on-board device of the preceding train or the following train, and a transmission device that transmits the received train state to the on-board device of the following train or the preceding train. Each of the own train state estimation device and the other train state estimation device calculates a train state using a time series filter.

  The distributed radio train control system according to the third embodiment does not use the ground device according to the first embodiment, and is between the preceding train transmitting device and the following train receiving device and between the preceding train receiving device and the following train transmitting device. It is characterized in that each communicates directly with each other.

  The distributed radio train control system according to the fourth embodiment does not use the ground device according to the second embodiment, and is between the preceding train transmitting device and the succeeding train receiving device and between the preceding train receiving device and the succeeding train transmitting device. It is characterized in that each communicates directly with each other.

  According to the present invention, even if a communication failure occurs, an emergency stop does not occur immediately, so availability is improved. In addition, since the frequency of sudden deceleration decreases, the risk of a passenger or cargo falling in the train is reduced. In addition, the trade-off between function and cost can be incorporated as design flexibility. By setting restrictions on functions such as allowing the occurrence of wireless dead zones and allowing large delays due to handover processing, it is possible to reduce the equipment cost to a level that was not possible before. . Furthermore, the likelihood calculation in the time series filter has a rationality check function. In a reasonably designed system, the irrational state is less likely. As a result, an irrational state can be avoided from acting on the control.

It is a figure which shows the structure of a centralized radio train control system among the radio train control systems which concern on embodiment of this invention. It is a figure which shows the control flow of the on-board apparatus of the centralized radio train control system of FIG. It is a figure which shows the control flow of the ground apparatus of the centralized radio train control system of FIG. It is a figure which shows conditional posterior periphery distribution calculation by the recurrence formula in the graphical model of the lock state structure used in a particle filter. It is a figure which shows the example of the output screen in the initial state of the position and speed estimation simulator by a particle filter. It is a figure which similarly shows the example of the output screen at the time of driving | running | working start. It is a figure which similarly shows the example of the output screen before a ground child detection. It is a figure which similarly shows the example of the output screen after a ground child detection. It is a figure explaining the function of the speed control using the probability distribution of the train position detection by the speed control apparatus of FIG. It is a figure which shows the structure of an example of a distributed radio train control system among the radio train control systems which concern on embodiment of this invention. It is a figure which shows the structure of the other example of a distributed radio train control system. It is a figure which shows the structure of the further another example of a distributed radio train control system. It is a figure which shows the structure of the further another example of a distributed radio train control system.

  Next, embodiments of the present invention will be described with reference to the drawings. The radio train control system of the present invention can be realized by either a centralized type or a distributed type. In addition, the distributed type can have four forms. Hereinafter, these will be sequentially described.

[Centralized wireless train control system]
FIG. 1 shows an example of the configuration of a centralized radio train control system. The centralized radio train control system A is a form in which estimation of the train state (for example, position and travel speed) of each train is performed centrally in the ground device 2.

  The centralized radio train control system A has on-board devices 1a and 1b provided on the respective trains Ta and Tb and a ground device 2 installed on the ground. The on-board devices 1a and 1b have the same configuration. That is, the sensor observation device 11, the own train state estimation device 12, the transmission device 13, the reception device 14, the speed control device 15, and the other train state estimation device 16 are included. The ground device 2 includes a receiving device 21, a train state estimating device 22, and a transmitting device 23.

  The transmitting device 13 and the receiving device 14 of the onboard devices 1a and 1b exchange predetermined information by wireless communication with the receiving device 21 and the transmitting device 23 of the ground device 2 via the onboard antenna 17 and the ground antenna 24. I do.

  In the on-board devices 1a and 1b, as shown in FIG. 2, the sensor observation device 11 requires sensor observation values (for example, the number of wheel rotation pulses of the speed generator, ground element reception required for estimating the train state of the own train). Information, notch state, etc.) (step 11; hereinafter, step is referred to as S). On the other hand, the acquired observation value of the sensor is transmitted from the transmission device 13 to the reception device 21 of the ground device 2 (S12). On the other hand, the acquired observation value of the sensor is input to the own train state estimating device 12, and the own train state estimating device 12 calculates the train state of the own train (S13).

  When the receiving device 14 inputs the train state of the other train from the ground device 2 (S14), the data is given to the other train state estimating device 15. The other train state estimation device 15 holds the data as the train state of the other train when the data is valid (Yes in S15). When the data is invalid (No in S15), the train state of another train is calculated (predicted) based on the train state being held (S16). The train state of the own train calculated in S12 and the train state of the other train received in S14 or calculated in S16 are input to the speed control device 16. Therefore, the speed control device 16 generates a verification speed pattern that ensures the prescribed safety from the train state of the input own train and the train state of the other train (S17), and optimal speed control for the own train. (S18). After the process of S18 is completed, the process returns to S11.

  In the ground device 2, when the receiving device 21 receives the sensor observation value from the transmission device 13 of the on-board device 1a or 1b (Yes in S21), the sensor observation value is input to the train state estimation device 22. Therefore, the train state estimation device 22 calculates the train state of each train based on the sensor observation value (S22). The calculated train state is transmitted to the receiving device 14 of each train by the transmitting device 23 (S23). After the process of S23 is completed, the process returns to S21.

  For the calculation of the train state by the own train state estimation device 12, the other train state estimation device 15 and the train state estimation device 22, a statistical model and a time series filter, among which the most preferable particle filter is used. For example, a Kalman filter or the like can be used instead of the particle filter.

FIG. 4 shows a process of conditional posterior peripheral distribution calculation by a recurrence formula in a graphical model of a locked state structure used for calculation of a train state (position and speed distribution) by a particle filter. The vertical axis in FIG. 4 indicates the number of observation data (observation values), and the horizontal axis indicates the state vector time. In FIG. 4, x is a state vector that is an aggregate of the position, speed, wheel diameter, and the like of each train, and y is an observation vector that is an aggregate of the axle rotation pulse count of each train, presence / absence of ground detection, and the like. . Also, p (x _j | y _{1: i} ) means the posterior marginal distribution (y ₁ , y ₂ ,..., Y _i is a given x _i conditional distribution).

  The solid arrows in FIG. 4 are an example of a calculation route by the own train state estimation device 12 when the observation data is normally obtained. The specific route to be taken depends on the number of input observation data and the state vector time at that time. Also, the white arrow in FIG. 4 is an example of a calculation path when observation data cannot be obtained due to a communication failure or the like (when No in S15 in FIG. 2). Of course, future observation data cannot be obtained from the current time. When estimating the future state, the calculation path is the same as that of the white arrow. Here, for the sake of simplicity, the explanation has been made on the assumption that the observation data can be obtained completely or not at all. However, in reality, there are situations where only specific observation data cannot be obtained. sell. In that case, a part of the observation vector may be treated as empty (invalid).

  Here, the particle filter used for the calculation of the train state (estimation of the state vector) will be described using a position / speed estimation simulation model using the particle filter.

[Simulation model]
(1) Premise Here, a simple model is assumed on the basis of only the wheel rotation pulse of the speed generator and the position correcting ground element. Assuming a periodic process, the time variation ΔT is assumed to be constant. The observed values for the time being are the pulse count of the speed generator (the amount of change from the previous trigger timing) and the presence or absence of ground unit reception.

(2) On-vehicle equipment simulation model Speed generator model The speed generator model, which is one of the on-vehicle equipment simulation models, uses wheel diameter, number of teeth, eccentricity, and tooth position accuracy as parameters. Of the true vehicle information given as input, the wheel angle is obtained from the change (displacement) of the position, and the number of teeth included between the previous wheel angle is counted and output. In the speed generator, idling / sliding can be an error factor, but for simplicity, idling / sliding is not considered. In addition, since the wheel diameter is included in the system model, the wheel diameter error is not included in the observation noise.

Ground unit model The ground unit model (the on-vehicle unit model as the on-vehicle device) uses a list of true positions of the ground unit as a parameter. When the position of the true vehicle information given as an input is in the vicinity of one of the ground child positions in the list, “detected” is output, and otherwise “not detected” is output. Actually, the range of “detected” varies depending on individual differences of the ground child, environment, vehicle speed, etc., but it is not considered for simplicity.

は十分小さいと考えられるため、全てステップ変化とした場合と大きな差異は生じないはずである。 (3) System model and system noise The acceleration is a step change between each time point. Therefore, the speed is linear interpolation and the position is quadratic spline interpolation. However, if the time variation ΔT is small,

Is considered to be sufficiently small, so there should be no significant difference from the case of all step changes.

For a state variable that is an element of the state vector, assuming that the position is x _t , the speed is V _t , the acceleration is a _t , and the wheel diameter is D _t , the system model is expressed by the following equation.

The system noise Rand _x and Rand _v are theoretically unnecessary (zero), but an appropriate random number is applied in consideration of the inaccuracy of the model itself. This also has a convenient meaning of preventing particle degeneration. Acceleration _{a t} varies in the range of maximum deceleration _{a min} and a maximum acceleration _{a max} Considering the slope or the like. The main point of the acceleration a _t is changed, when the notch operation, a time of slope change. Here, it is assumed that notch / gradient information is unknown. As the simplest model, a uniform distribution between the maximum deceleration a _min and the maximum acceleration a _max is considered. However, for example, the model can be enhanced by giving constraints on jerk. Similarly, the wheel diameter _{D t} is varied in the range of minimum wheel diameter _{D min} and a maximum wheel diameter _{D max.}

この車輪径のモデルは、前回車輪径にノイズＲａｎｄ_Ｄを加え、最小車輪径Ｄ_ｍｉｎから最大車輪径Ｄ_ｍａｘの間にクリッピングしたものである。Ｒａｎｄ_Ｄは一様分布を適用する。もっとも、分岐箇所等を除けば、車輪径が急激に時間変化することはないと考えられるので、Ｒａｎｄ_Ｄはゼロにピークを持つ分布にすることで、より高い精度で推定が行えると考える。

This wheel diameter model is obtained by adding noise Rand _D to the previous wheel diameter and clipping between the minimum wheel diameter D _min and the maximum wheel diameter D _max . Rand _D applies a uniform distribution. However, since it is considered that the wheel diameter does not change rapidly with the exception of branch points, it is considered that Rand _D can be estimated with higher accuracy by making the distribution to have a peak at zero.

ｋは、速度発電機の歯の数で決まる係数である。Ｒａｎｄ_ｃは、サンプリングタイミングによるカウントの揺らぎに相当するノイズで、［０、１］の一様分布である。厳密には、過去のサンプリングタイミングとの相関があるため、一様分布ではないが、ここではマルコフ過程で近似する。ｆ_ｂ（ｘ_ｔ）は、地上子モデルと同様に、地上子位置近傍となった際に「検知あり」とし、地上子位置近傍以外では「検知なし」とする。地上子位置は車上データベースに記録されていると想定して、既知（地上子モデルの真の地上子位置一覧と同一）としている。 (4) observation model observation vector includes a wheel rotation pulse count c _t rate generator consists of two elements of the ground coils detect the presence or absence b _t, is expressed by the following equation.

k is a coefficient determined by the number of teeth of the speed generator. Rand _c is a noise corresponding to the fluctuation of the count due to the sampling timing, and has a uniform distribution of [0, 1]. Strictly speaking, since there is a correlation with the past sampling timing, the distribution is not uniform, but here it is approximated by a Markov process. Similarly to the ground child model, f _b (x _t ) is “detected” when near the ground child position, and “not detected” except near the ground child position. Assuming that the ground child position is recorded in the on-board database, it is known (same as the true ground child position list of the ground child model).

としたとき、観測された車輪回転パルスカウントＣ_ｔの尤度Ｌ_ｃは

となる。 (5) Likelihood In the process of filtering by the particle filter, each particle (particle) is weighted. Weighting is performed based on likelihood. The value corresponding to the wheel rotation pulse count expected from the past train state

The likelihood L _c of the observed wheel rotation pulse count C _t is

It becomes.

For example, when c ′ _t = 2.3 is predicted, if the observed value c _t = 2, the likelihood is 0.7, and if c _t = 3, the likelihood is 0.3. On the other hand, the likelihood L _b regarding the ground element is obtained from Table 1 below.

Here, 0 ≦ l _b1 ≦ 1 and 0 ≦ l _b2 ≦ 10. The values of l _b1 and l _{b2 depend on} the actual characteristics of the sensor / determination unit. In addition, the characteristics (such as misdetection is likely to occur or detection is likely to be lost) depend on the fail-safe design. The likelihood of the entire model is obtained by L _c × L _b . This is the likelihood under the condition that the system operates as expected. In practice, the actual likelihood density function is flatter because events that do not satisfy the model assumptions may occur.

  A specific example of applying the above particle filter estimation theory to a position / velocity estimation simulation will be described. FIG. 5A is a position / speed graph shown on the output screen of the simulator under development, and shows an initial state before the train travels. The horizontal axis is position, and the vertical axis is speed. Although it is actually stopping at a position of 50 m (speed 0), the train state estimation device gives an initial condition that it is stopped between 25 and 75 m as shown by particles in FIG. 5A. . In FIG. 5A, the denser the particles, the higher the probability.

  FIG. 5B is an output screen when the train starts to travel. The train state estimation device estimates the speed based on the observation value (wheel rotation pulse count output from the speed generator model) obtained from the sensor observation device. The position is still ambiguous and the mode value is also different from the true value. The solid curve in FIG. 5B is the true position / velocity.

  FIG. 5C shows a state before the sensor observation device detects the ground element. The ground unit exists at a position of 100 m, but the ground unit model has not yet output “with reception”. The probability that the train exists in front of the 100 m position is increased by the estimation by the particle filter (likelihood evaluation regarding the detection of the ground element).

  FIG. 5D shows a state after the ground child model outputs “with reception” and the sensor observation device detects the ground child. By detecting the ground element at a position of 100 m, the position / speed estimation by the train state estimation device converges near the true value.

  As described above, in the particle filter executed by the present invention, the train position / speed can be estimated with high accuracy by referring to the mode value, median value, or average value of the probability distribution. At the same time, information on the probability distribution shape that is indispensable for safe-side control can be obtained.

  As described above, the observation value acquired by the sensor observation device 11 on the vehicle is analyzed by the own train state estimation device 12 using the time series filter, and the train state of the own train is calculated. On the other hand, the sensor observation value of each train received by the receiving device 21 on the vehicle is input to the train state estimating device 22, and the train state of each train is calculated using the time series filter here as well.

  In Patent Document 2, the standing line probability is calculated after performing position estimation and position estimation accuracy evaluation in the preceding train and the subsequent train, respectively. However, what is obtained from the position estimation result and the position accuracy evaluation result is the standing line probability at the time of calculation. In particular, there is insufficient information to predict the standing line probability ahead of the time of calculation with high accuracy, and prediction is difficult. In the present invention, not only the position but also the state of each train is calculated and held, so not only the position / speed estimation approaches the true value but also the future train state can be predicted with higher accuracy. is there.

  The train state calculation models of the own train state estimation device 12 and the train state estimation device 22 are basically the same. However, the former targets a single train and does not consider information loss or alteration due to wireless communication failure, but the latter targets a plurality of trains and is different in considering wireless communication failure.

  Information on the train state calculated by the train state estimation device 22 is transmitted from the transmission device 23 to the reception device 14 of all trains or a specific train that can be related.

  In each train, based on the train state information received by the receiving device 14, the train state of the train that can be related by the other train state estimating device 15 is estimated again. The estimation by the other train state estimation device 15 is performed for a train state at a time earlier than the current time as well as considering a radio communication failure. However, the estimation of the train state at a time earlier than the current time may be performed by the train state estimation device 22 instead of the other train state estimation device 15 and transmitted to each train.

  The speed control device 16 of each train determines a safe verification speed pattern based on the current train state obtained from the own train state estimation device 12 and the other train state estimation device 15 and the train state at a time earlier than the current time. Then, control is performed so that the speed of the own train is less than the calculated safe speed.

  The speed control device 16 will be further described. FIG. 6 is a diagram for explaining the function of speed control using the probability distribution of train position detection by the speed control device 16. The vertical axis is the probability, the horizontal axis is the kilometer (position), and C1 is a curve indicating the probability distribution of the own train position detection at a specific time inputted from the own train state estimation device 12 to the speed control device 16, C2. These are curves which show the probability distribution of other train position detection of the same time inputted from the other train train state estimation device 15 to the speed control device 16. The shapes of the curves C1 and C2 differ depending on the sensor observation values so far and change with time. For example, a steep curve can be obtained immediately after detecting the position correction ground element. Alternatively, if the status of another train cannot be received due to a communication failure, it can be a slow curve. A curve C3 where the curves C1 and C2 overlap each other indicates a probability distribution in which two trains are located at the same position, that is, a probability distribution of collision (collision). The probability distribution C3 is obtained by the product of two probability distributions C1 and C2. The probability of collision (collision) in a certain position range is the C3 integral value in that position range.

  There are several methods for determining the verification speed pattern, but an example is shown here. A predetermined check speed pattern is applied to the on-board device, and the own train state estimation device 12 has a probability distribution C1 of the future position of the current train on the train under the condition that the brake acts when the check speed pattern is exceeded. Ask for. The other train state estimation device 15 calculates the probability distribution C2 of the future on-line position for the preceding train. At that time, the verification speed pattern applied to the preceding train is assumed to be unknown. However, when information on the check speed pattern is included in the train state, the information may be applied as a condition. In the speed control device 16, a threshold is set for the probability distribution C3. If the mode of the probability distribution C3 at any time exceeds the threshold in a certain time range from the current time, the threshold is exceeded. The check speed pattern is updated according to the degree of deceleration and decelerated. On the other hand, when the mode value of the probability distribution C3 is less than the threshold value, the verification speed pattern is updated to delay the brake action start point or reduce the deceleration.

  In the above method, the check speed pattern can be determined even if information is temporarily lost from each train or ground device due to communication failure or the like. If the communication failure time is short, there is little deviation in the probability distribution before and after information acquisition, and the change in the verification speed pattern, that is, the influence on the train operation is small. Even if the communication failure time is long, as long as the verification speed pattern is followed, a predetermined safety is ensured under the assumption that an unexpected event, that is, an event not considered in the statistical model does not occur.

  The major difference between the radio train control system of Patent Document 2 and the radio train control system according to the present invention is that the content of information to be transmitted / received is [stop target] in the former, whereas [use time series filter] The train state is calculated]. By transmitting and receiving such a train state, even if a communication failure occurs, it is possible to estimate the state of surrounding trains for the time being. The estimated value increases the uncertainty due to the missing information (distribution approaches flatness), but in order to calculate a safe speed including that uncertainty, an emergency stop is always performed based on the occurrence of a communication failure. It is possible to continue operation safely without any problems.

[Distributed wireless train control system 1st form]
As shown in FIG. 7, the distributed radio train control system B1 of the first form has almost the same configuration as the case where the onboard devices 1a and 1b are centralized, whereas the ground device 2 and the receiving device 21 transmit. It consists only of the device 23. Therefore, the ground device 2 has only a function as a relay for exchanging sensor observation values between trains. And the function of the train state estimation apparatus 22 of the ground apparatus in a centralized structure is implement | achieved by the other train state estimation apparatus 15 of each on-board apparatus 1a, 1b. This distributed radio train control system B1 is also resistant to the occurrence of communication failures and can be operated safely.

[Distributed radio train control system second form]
As shown in FIG. 8, the distributed radio train control system B2 of the second form is obtained by changing the exchange of the observation value of the sensor in the first form to the exchange of the train state. In other words, the sensor observation device 11 is connected to the transmission device 13 in the first form, but the own train state estimation device 12 is connected to the transmission device 13 in the second form.

[Distributed Wireless Train Control System Third Form]
As shown in FIG. 9, the distributed radio train control system B3 of the third form replaces the transmission device 13 and the reception device 14 of the on-board device 1a in the first form with the reception device 14 and the transmission device 13 of the on-board device 1b. Directly connected, the observation value of the sensor in the first form is realized by inter-vehicle communication, and the ground device is eliminated.

[Distributed radio train control system 4th form]
As shown in FIG. 10, the distributed radio train control system B4 according to the fourth embodiment uses the transmission device 13 and the reception device 14 of the on-board device 1a in the second embodiment as the reception device 14 and the transmission device 13 of the on-board device 1b. Directly connected, the transfer of the train state in the second form is realized by inter-vehicle communication, and the ground device is eliminated.

  In the configuration described above, at least a function for collecting sensor data in each train, a wireless communication function, and a speed control function are necessary, but there are no restrictions on the method of mounting each function on a device. For example, the functions of the transmission device 13 and the reception device 14 of the on-board device may be implemented as one device. Moreover, you may implement the function of the own train state estimation apparatus 12 and the other train state estimation apparatus 15 as one apparatus. Furthermore, since a statistical model including a communication system can be created, there is no restriction on the communication path. For example, there may be a communication relay device using wired or wireless between the transmission / reception antenna 24 and the reception device 21 and the transmission device 23.

  Also, it is not necessary to send and receive all information with the same quality and frequency. For example, since the train length usually does not change between stations, it is not necessary to transmit train length information while traveling between stations. As described above, the information with less dynamic change can improve the train control capability with respect to the communication line capability by, for example, a method of exchanging information when the station stops.

A Radio train control system (centralized)
B1-B4 Wireless train control system (distributed)
DESCRIPTION OF SYMBOLS 1a, 1b On-board apparatus 11 Sensor observation apparatus 12 Own train state estimation apparatus 13 Transmission apparatus 14 Reception apparatus 15 Other train state estimation apparatus 16 Speed control apparatus 17 Transmission / reception antenna 2 Ground apparatus 21 Reception apparatus 22 Train state estimation apparatus 23 Transmission apparatus 24 Transmit / receive antenna

Claims (7)

  A position / speed sensor and an on-board radio station are provided on the vehicle, and the on-board radio station transmits and receives information on the train state of the train via the ground device or the ground radio station and the ground device. A radio train control method characterized in that the own train state and other train states are estimated based on a statistical model and a time series filter, and each train performs safety-side control based on the estimated information.   On-board device, sensor observation device that acquires sensor observation values necessary for estimating the train state of the own train, and own train state estimation that calculates the train state of the own train based on the acquired sensor observation values Apparatus, an other train state estimation device that calculates a train state of the other train based on a sensor observation value of the other train received from the ground device or another on-board device, the calculated train state of the own train, and the A wireless train control system having a speed control device that controls the own train based on a train state of another train, wherein both the own train state estimation device and the other train state estimation device use the sensor observation values. A radio train control system characterized by estimating a train state such as a position and a speed of the own train or the other train from time to time using a statistical model and a time series filter.   The on-board device includes a sensor observation device that acquires an observation value of a sensor necessary for estimating a train state of the own train, a transmission device that transmits the acquired sensor observation value to a ground device, and the acquired sensor The own train state estimating device that calculates the train state of the own train based on the observation value, the receiving device that receives the train state of the other train from the ground device, and the other based on the received train state of the other train An other train state estimating device for calculating the train state of the train, and a speed control device for controlling the own train based on the calculated train state of the own train and the train state of the other train. Is a receiving device that receives the sensor observation values from the on-board device, a train state estimating device that calculates a train state of each train based on the received sensor observation values, and a calculated train state of each train. Previous A radio train control system including a transmission device that transmits to an on-board device, wherein the own train state estimation device, the other train state estimation device, and the train state estimation device are all based on the sensor observation values A radio train control system that estimates the train state using a statistical model and a time series filter.   The on-board device includes a sensor observation device that acquires an observation value of a sensor necessary for estimating a train state of the own train, a transmission device that transmits the acquired sensor observation value to a ground device, and the acquired sensor Based on the own train state estimation device that calculates the train state of the own train based on the observation value, the receiving device that receives the sensor observation value of the other train via the ground device, and the received sensor observation value of the other train An other train state estimating device for calculating the train state of the other train, and a speed control device for controlling the own train based on the calculated train state of the own train and the train state of the other train, The ground device includes a receiving device that receives sensor observation values from the on-board device of the preceding train or the following train, and a transmitting device that transmits the received sensor observation values to the on-board device of the following train or the preceding train. The own train state estimating device and the other train state estimating device are both for estimating the train state using a statistical model and a time series filter. Wireless train control system.   The on-board device includes a sensor observation device that acquires sensor observation values necessary for estimating the train state of the own train, and the own train state that calculates the train state of the own train based on the acquired sensor observation values. Based on the estimation device, the transmission device that transmits the calculated train state of the own train to the ground device, the reception device that receives the train state of the other train via the ground device, and the received train state of the other train An other train state estimating device for calculating the train state of the other train, and a speed control device for controlling the own train based on the calculated train state of the own train and the train state of the other train. A wireless train control system comprising: a receiving device that receives a train state from the on-board device of a preceding train or a subsequent train; and a transmitting device that transmits the received train state to the on-board device of the following train or the preceding train. A is, the self-train state estimation device and other train state estimation apparatus, a radio train control system, characterized in that both is to estimate the train state using the statistical model and a time series filter.   5. The radio train control system according to claim 4, wherein the ground device is not provided, the transmitter between the preceding train and the receiver of the subsequent train, and the receiver of the preceding train and the transmitter of the subsequent train. A radio train control system characterized by direct communication between the two.   6. The radio train control system according to claim 5, wherein the ground device is not provided, the transmitter between the preceding train and the receiver of the subsequent train, and the receiver of the preceding train and the transmitter of the subsequent train. Distributed radio train control system characterized by direct communication between each other.

Images