JP2006107121A - Vehicle dispatching plan preparation device, vehicle dispatching plan preparation method, computer program and computer readable recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically assign an AGV suitable for carrying a carrying object to a destination. <P>SOLUTION: The travel of the AGV is constituted of a Petri net model, a target state amount M<SB>r</SB>set corresponding to an AGV state and a carriage instruction is defined as a target, and an operation vector u(k) is determined from the state equation of a Petri net. A control means 213 obtains the solution of a vehicle dispatching problem using an optimum control theory by performing the moving operation of respective coils and the respective AGVs on the basis of the operation vector. In the meantime, a function to be the linear sum of the total empty vehicle traveling time of the AGV and the total carriage standby time of the coil is set in a cost function setting means 208, and the solution for which the cost function becomes minimum is obtained in a vehicle dispatching combination calculation means 209. A simulation means 203 performs respective simulations using the solution of the vehicle dispatching problem, the solution for which the cost function becomes minimum is obtained in a cost function comparison means 211, and the solution is returned to a high-order process computer 101. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a vehicle allocation plan creation device, a vehicle allocation plan creation method, a computer program, and a computer-readable recording medium, and is particularly suitable for use in an apparatus for performing a vehicle allocation plan (assignment of conveyance instructions) for an automatic guided vehicle. Is.

  In manufacturing processes in many industries including steel, a plurality of products are processed by a plurality of different processes. Therefore, it is necessary to transport materials and products between the production lines. Conventionally, for example, a system using a plurality of automatic guided vehicles (AGV = Automatic Guided Vehicle) is used as a physical distribution system for transporting an object.

  In this system using AGV, while communicating with AGV and a higher-level process computer (hereinafter referred to as “procone”), the traveling route data is set for each AGV by allocating a transfer command given from the upper-level control computer to AGV. An object is transported between a plurality of AGV skids (such as a buffer for storing an object) by unmanned traveling on a predetermined traveling path.

  At this time, a plurality of optical space transmission devices (OCD) are installed along the traveling route of the AGV, and communication is performed between the AGV and the host computer main body via the OCD. Thus, as described above, a transport command from the host process computer can be assigned to each AGV, and what sort of object is being transported by which AGV can be grasped by the host process computer.

  In the transport system using AGV as described above, when a material or product is transported between the AGV skids, which AGV is used among a plurality of AGVs, that is, output from a host computer. It is necessary to determine to which AGV one or more transfer instructions are assigned. Conventionally, with respect to this determination method, among AGVs that have become empty due to the end of transportation, etc., allocation is performed in ascending order of the distance from the current position to the receiving position indicated in the transportation instruction. (See Patent Document 1).

JP-A-11-327644

  By the way, in the conveyance system as described above, for example, depending on the state of the traveling path of the AGV, the traveling time to the cargo receiving position may be long even if the distance to the cargo receiving position is short. However, in the above-described conventional method, even in such a case, a conveyance command is assigned to an AGV that is close to the receiving position. That is, the above-described conventional technique has a problem in that the transport command is not always assigned to the AGV that has a short distance to the load receiving position but does not necessarily travel to the load receiving position.

  In general, a travel path on which AGV travels is designed with a single wire due to problems such as equipment cost and physical space in the factory, and AGV travels in both directions in many cases. For this reason, a plurality of AGVs may cause a collision (interference) during traveling. When the AGV causes interference, it is necessary to stop, evacuate or reversely travel, resulting in a long time.

  Even if such interference does not occur, the distance is shorter than other AGVs due to the condition of the traveling path (for example, the traveling speed varies depending on whether the traveling path is a straight line or a curve) and other factors. However, the running time may not be shorter than the other AGVs.

The present invention has been made to solve such a problem, and a first object is to automatically assign a transport vehicle such as an AGV suitable for transporting a transported object to a destination. .
In addition, by assigning such a transport command, the travel time of the transport vehicle can be shortened, the transport wait time until the transported object is carried out can be shortened, and the empty vehicle travel time and transport wait time can be traded off. The second object is to be able to optimally adjust.

The vehicle allocation plan creation device of the present invention represents a transportation process with a Petri net model, and is a vehicle allocation plan creation device that determines the allocation of the transportation command to the transportation vehicle from the Petri net model, as a Petri net model of the transportation process, A model construction means for constructing a Petri net model in which a traveling path of the transport vehicle is represented by a place and a moving operation end, and a transport object and a transport vehicle are represented by tokens, and a Petri net model constructed by the model construction means. In addition to obtaining a state equation, a gain matrix calculation means for obtaining a feedback gain matrix from the obtained state equation and an evaluation function represented by the set evaluation matrix and target state quantity, and represents an in-process state of a token in the transport process A state vector input means for obtaining a state vector; A feedback gain matrix obtained by the matrix calculating means, a state vector representing the in-process state obtained by the state vector input means, and a target state quantity set in accordance with the state of the transport vehicle and a given transport command. Using the operation vector calculating means for obtaining the operation vector for the moving operation end of the transported object and the transport vehicle, and the operation vector for each moving operation end based on the operation vector obtained by the operation vector calculating means. A state vector after the operation is obtained by performing an operation of moving each transported object and each transported vehicle by the movable number in the order in which a large operation amount is obtained with a positive value in Control means for finding a solution to a vehicle allocation problem for allocating each transport object to each transport vehicle using a vector, and the control means includes The state vector after the operation was characterized by feeding back to the state vector input means as a new state vector representing the progress state of the token in the transport process.
Another feature of the present invention is that the model building means includes a Petri net model module for a skid that is a place where the transport vehicle performs work, and a relay that is a place where the transport vehicle communicates. The Petri net model module is combined with the Petri net model module.
Another feature of the present invention is that the model construction means constructs the Petri net model for each transport vehicle, and uses the Petri net model individually for transport vehicles to which the transport instruction has been assigned, This is because a single Petri net model is used collectively for a transport vehicle to which the transport command is not assigned, and the transport command is assigned to the unassigned transport vehicle.
According to another feature of the present invention, the control means is configured to apply to each moving operation end based on an operation vector obtained by integrating the operation vectors calculated for each of the transport vehicles by the operation vector calculation means. In the integrated operation vector, an operation is performed to move each conveyed product and the transported vehicle by a movable number in the order in which a large positive operation amount is obtained.
Another feature of the present invention is that it includes state management means for controlling and managing the state transition of the transport vehicle according to the firing of the moving operation end.
Another feature of the present invention is that a cost function defined by the position and work state of each transport vehicle and the transport source and transport destination of each transport object assigned to each transport vehicle is set. Cost function setting means, target value changing means for changing the target state quantity as a function of the coefficient ratio of the cost function, and each of the transport vehicles so that the value of the cost function is as small as possible. A vehicle allocation combination calculation means for obtaining a plurality of solutions for a vehicle allocation problem for allocating a transported object using a combination optimization method, a vehicle allocation problem solution obtained from the control means, and a vehicle allocation obtained by the vehicle allocation combination calculation means Each of the cost solutions calculated by the simulation means for simulating the problem solution and calculating the value of the cost function is performed. And a cost function comparison unit that outputs a solution of a vehicle allocation problem corresponding to the smallest value, and the operation vector calculation unit includes a feedback gain matrix obtained by the gain matrix calculation unit, and the state vector input An operation vector for the transported object and the moving operation end of the transport vehicle is obtained using the state vector representing the in-process state acquired by the means and the target state quantity changed by the target value changing means.
According to another feature of the present invention, the cost function setting means includes a cost function that calculates a linear sum of a monotonically increasing function of an empty vehicle traveling time of each transport vehicle and a monotonically increasing function of a transport waiting time of each transport object. There is to use as.
Further, another feature of the present invention is that the cost function setting means includes a monotonically increasing function of an empty travel time of each transport vehicle, a monotonically increasing function of a transport waiting time of each transport object, and a specific zone. A linear sum with a function that increases in value due to the concentration of transport vehicles is used as a cost function.
In addition, another feature of the present invention is that the vehicle allocation combination calculating means is configured so that each transport vehicle is placed in the nearest battery field so that the remaining battery level of each transport vehicle does not fall below a specified value. The solution to the vehicle allocation problem is to be determined by taking into account the time it takes to travel and the charging time of each transport vehicle in the nearest battery field.
In addition, another feature of the present invention is that the vehicle combination calculation means divides the traveling path of the transport vehicle into a plurality of areas, and an empty vehicle and an unloading transport vehicle in the area, When the number exceeds the designated number, the solution is to find a solution to the vehicle allocation problem so that the transport vehicle is evacuated to another area so that the number is less than the designated number.

  The vehicle allocation plan creation method of the present invention is a vehicle allocation plan creation method that represents a transfer process by a Petri net model, and determines the allocation of the transfer command to the transfer vehicle from the Petri net model, as a Petri net model of the transfer process, A model construction step for constructing a Petri net model in which a traveling path of the transport vehicle is represented by a place and a moving operation end, and a transport object and a transport vehicle are represented by tokens, and a Petri net model constructed by the model construction step. In addition to obtaining a state equation, a gain matrix calculation step for obtaining a feedback gain matrix from the obtained state equation and an evaluation function represented by the set evaluation matrix and target state quantity, and represents a token in-process state in the transfer process State vector input step to obtain the state vector , The feedback gain matrix obtained in the gain matrix calculation step, the state vector representing the in-process state acquired in the state vector input step, the state of the transport vehicle and the transport command to be given. Using the target state quantity, an operation vector calculating step for obtaining an operation vector for the moving operation end of the conveyed product and the transport vehicle, and for each moving operation end based on the operation vector obtained by the operation vector calculating step. Then, by performing an operation of moving each transported object and each transported vehicle by a movable number in the order in which a large operation amount with a positive value in the operation vector is obtained, a state vector after the operation is obtained and obtained. A control step for finding a solution to the vehicle allocation problem of assigning each transported object to each transport vehicle using the state vector after operation. And a flop, the state vector input step, the state vector after the operation determined by the control step, and acquires a new state vector representing the progress state of the token in the transport process.

A computer program according to the present invention is a computer program for causing a computer to execute a transfer process represented by a Petri net model and determining assignment of a transfer command to a transfer vehicle from the Petri net model. As a net model, the travel path of the transport vehicle is represented by a place and a moving operation end, and a model construction step for constructing a Petri net model in which a transport object and a transport vehicle are represented by tokens, and the model construction step is constructed. A state equation of the Petri net model is obtained, a gain matrix calculating step for obtaining a feedback gain matrix from the obtained state equation and an evaluation function represented by the set evaluation matrix and the target state quantity, and a token of the transfer process In-process status A state vector input step for obtaining a state vector to be represented, a feedback gain matrix obtained by the gain matrix calculation step, a state vector representing an in-process state obtained by the state vector input step, a state of the carrier, and giving An operation vector calculation step for obtaining an operation vector for the moving operation end of the transported object and the transport vehicle using a target state quantity set in accordance with a transport command to be performed, and an operation vector obtained by the operation vector calculation step. On the basis of each moving operation end, an operation is performed to move each transported object and each transported vehicle by a movable number in the order in which a large and positive operation amount in the operation vector is obtained. Obtain the state vector after the operation, and use the obtained state vector after the operation for each transport vehicle. A control step for finding a solution to a vehicle allocation problem for allocating each transported object, and the state vector input step uses the state vector after the operation obtained by the control step as a token in-process in the transport process. It is obtained as a new state vector representing a state.
A computer-readable recording medium according to the present invention records the above-described computer program.

According to the present invention, the transport process is represented by a Petri net model, and a transport command is sent to each transport vehicle according to the state equation obtained from the Petri net model and the evaluation function represented by the set evaluation matrix and target state quantity. Since it is assigned, compared to the conventional example in which a transfer command is assigned to a transport vehicle that is close in distance, the traveling time of the transport vehicle can be shortened, and interference between the transport vehicles can be remarkably reduced. Further, it is possible to prevent the traveling time of the transport vehicle from becoming long due to the interference.
According to another aspect of the present invention, the cost function value defined by the position and work state of the transport vehicle and the transport source and transport destination of the transport object assigned to the transport vehicle is calculated. Since a smaller solution is obtained and the transport vehicles are allocated based on the solution, it is possible to perform the allocation in consideration of the trade-off between the empty vehicle travel time of the transport vehicle and the transport waiting time of the transported object.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(Outline of vehicle allocation plan creation device)
FIG. 1 is a block diagram illustrating an example of a schematic configuration of a vehicle allocation plan creation device according to the present embodiment.
In FIG. 1, the vehicle allocation plan creation device includes a simulation control unit 1 including a vehicle allocation simulator 1a, an optimization calculation device 2, and the like.

  In the present embodiment, the vehicle allocation simulator 1a operates using a simulator that simulates a conveyance process that expresses the state of the automatic guided vehicle (AGV) and the restrictions when the automatic guided vehicle (AGV) is allocated, in other words, AGV. A large simulator simulating a factory. More specifically, for example, the vehicle dispatch simulator 1a is constructed using a Petri net model, and is configured as a discrete system that moves an object for each event (event of the simulator 1a).

In addition, the vehicle allocation plan creation device includes a mathematical model 3 that mathematically expresses the vehicle allocation simulator 1a. In the present embodiment, the state equation as shown in the following equation (A) is obtained by taking in the elements related to the allocation plan to be created from the AGV status in the transfer process and the constraints when the AGV is allocated. The mathematical model 3 is created using The mathematical model 3 may be held by a mathematical model holding means (not shown) constituted by a semiconductor memory device or the like.
M (k + 1) = M (k) + B · u (k) (B)
That is, a state vector indicating the number of tokens present in each place at a certain discrete time k, in other words, a state quantity is represented by M (k), and whether or not each transition is ignited is represented by “1” and “0”. When the manipulated vector is represented by u (k), the state vector M (k + 1) at the next time k + 1 can be represented by the above equation (A) using the connection matrix B. Here, the connection matrix B is the number of increase / decrease of tokens existing in the i-th place of M (k) when the j-th transition of u (k) is ignited (i, j). An integer matrix with elements.

The optimization calculation apparatus 2 performs an optimization calculation process on the mathematical model 3 to calculate the feedback gain K. The optimization calculation performed by the optimization calculation device 2 uses linear quadratic (LQ) control, and is performed using an evaluation function S as shown in the following equation (b).
S = Σ [M ′ (k) QM (k) + u ′ (k) Ru (k)] (b)
In the evaluation function S, Q and R are appropriate matrices set in accordance with the control purpose, and M ′ (k) and u ′ (k) are the state vector M (k) and the operation vector u (k), respectively. Is the transpose vector. Σ represents the sum of values in {} from k = 0 to ∞. And considering that the evaluation function S is controlled to be minimized,
u (k) = − K · M (k) (C)
The feedback gain K for performing the state feedback control can be obtained by the optimal control theory.

  The optimal control policy unit 1b of the simulation control unit 1 uses the feedback gain K calculated by the optimization calculation device 2 and the AGV state (state vector M) to calculate a vehicle allocation instruction (operation vector u). The process of giving the vehicle dispatch simulator 1 and proceeding with the simulation to obtain a new AGV state and calculating a new vehicle dispatch instruction based on the new AGV state is repeated. And the dispatch schedule in a conveyance process is created from the simulation result 4 obtained in this way.

(First embodiment)
FIG. 2 is a block diagram illustrating an example of the overall configuration of a scheduling system that implements a vehicle allocation plan creation device for an automatic guided vehicle (AGV) according to the first embodiment of this invention. In the following embodiments, a case where a coil is transferred between a plurality of AGV skids in an iron manufacturing process will be described as an example.

  As shown in FIG. 2, the scheduling system of the present embodiment is configured to include a host computer 100 and a personal computer 101 having a function as an AGV dispatch plan creation device. The host computer 100 supplies information representing a predetermined transport command, AGV state, coil state, and current position of the AGV to the vehicle allocation plan creation device (personal computer) 101. The transfer command includes information such as a coil number for identifying the coil to be transferred, and From / To indicating from where to where the coil is transferred.

  In addition, in the actual process, the AGV status and the current position are higher when the AGV and the host computer 100 communicate with each other via a plurality of optical space transmission devices (OCDs) installed along the AGV travel route. Although it is grasped by the process control 100, it is grasped from the result of the simulation for each discrete time when the dispatch plan is created in advance. In other words, the host computer 100 knows which AGV is currently performing what kind of work and where each AGV is currently located.

  The AGV state handled in the present embodiment indicates an AGV working state such as an empty vehicle, requested travel, cargo receiving, transport, unloading, charging travel, charging, and retreat travel. An “empty vehicle” refers to an AGV that is not loaded with a coil and is not assigned a transfer command. That is, it means a state in which no work is performed and no work to be performed is determined. For example, when the AGV becomes empty due to the completion of the coil conveyance, the AGV basically remains at the conveyance completion position.

  The “required travel” is a state in which the coil is not actually transporting, but the AGV, which has been empty, receives the transport command and travels toward the receiving position of the coil (From position in the transport command). Say. The “load receiving” refers to a state where the AGV in the requested traveling state has subsequently arrived at the load receiving position and the coil has been transferred from the AGV skid to the AGV. “Conveyance” means a state in which the receipt of the cargo is completed and the coils are actually loaded and conveyed toward the unloading position (To position in the conveyance command).

  “Unloading” means a state in which the AGV has arrived at the unloading position and the coil has been lowered from the AGV to the AGV skid. When this unloading is completed, the AGV returns to the “empty” state again. “Charge running” refers to a state in which the vehicle is running toward the charging station for AGV battery replacement, and “charging” refers to a state in which charging is actually performed. “Evacuation traveling” refers to a state in which the vehicle is traveling toward a retraction position determined in advance by a rule or the like in order to avoid interference between AGVs. This retreat travel command is issued to the AGV of the empty vehicle.

  As shown in FIG. 2, the host computer 100 transmits data representing the current state to the vehicle allocation plan creation device 101 during the state transition as described above, and this is managed by the state management unit 205. The state management unit 205 stores not only the working state such as the empty vehicle, the requested traveling, and the conveyance, but also the current position of each AGV. The vehicle allocation plan creation apparatus of this embodiment obtains information on which AGV has become empty in the state management means 205, and assigns a transport command to the empty vehicle AGV.

  Specifically, as shown in FIG. 2, the vehicle allocation plan creating apparatus 101 includes an input unit 301, a gain matrix calculation unit 302, a storage unit 201, an operation vector calculation unit 202, a state vector input unit 204, and a state. A management means 205, a transport command management means 206, a processing plan output means 207, a model construction means 212, and a control means 213 are provided, and an AGV dispatch plan creation method according to this embodiment as described below To implement.

The input means 301 includes AGV skids and OCDs constituting the AGV transport process, travel paths and distances between the AGV skids, evaluation matrices Q and R representing a predetermined evaluation function related to the Petri net model, and further, each AGV each information of the target state quantity M _r representing a target state (a state to be achieved for the initial state) is for inputting each AGV, for example constituted by a keyboard or the like. The evaluation matrices Q and R are
Q = Q ^T ≧ 0, R = R ^T > 0
(T in the shoulder of the matrices Q and R indicates a transposed matrix).

  The model construction unit 212 constructs, for each AGV, a Petri net model in which AGV travel paths are represented by places and transitions and AGVs and coils are represented by tokens in accordance with the information input by the input unit 301. At this time, a connection matrix B describing a state equation of the Petri net model is obtained. By this connection matrix B, the movement of the token by the firing of the transition that is the movement operation end of the token (AGV and coil) is expressed by a mathematical expression.

Here, a state vector indicating the number of tokens present in each place at a discretized time k, that is, a state quantity is represented by M (k), and the presence / absence of each transition firing is represented by “1” and “0”. If the vector is represented by u (k), the state vector M (k + 1) at the next time k + 1 can be represented by the state equation shown in the following equation (1).
M (k + 1) = M (k) + B · u (k) (1)

  FIG. 3 shows an example of a Petri net model. FIG. 3A shows a simple example of an actual transfer process, and FIG. 3B shows an example of a Petri net model created according to this transfer process. That is, a position where each AGV skid or OCD in the traveling path is a zone, and the AGV skid or OCD in each zone is represented by a place, and each zone is connected by an arrow to represent a traveling path. Each zone is a Petri net model in module units (details will be described later), and arrows indicate arcs of the Petri net model. Note that the OCD is a node that represents the connection of each AGV skid.

Gain matrix calculation means 302, the input evaluation matrix Q by the input means 301, and R and the target state quantity M _r, by using the connection matrix B obtained by modeling means 212, the value of the predetermined evaluation function A feedback gain matrix K of the Petri net model that minimizes is obtained. The calculation of the feedback gain matrix K will be described in detail below.

As described above, the state equation of the Petri net model is expressed by the above equation (1). Furthermore, in view of the target state quantity M _r that is introduced in this embodiment, the above formula (1) can be written as the following equation (2).
M (k + 1) −M _r = (M (k) −M _r ) + B · u (k) (2)
Here, M (k) −M _r can be regarded as a new state x (k), and Equation (2) can be written as the following Equation (3).
x (k + 1) = x (k) + B · u (k) (3)

The state equation of the above equation (1) corresponds to a case where the target amount of each place state is zero. In contrast, as the state equation of the formula (2) or the formula (3), by introducing the target state quantity M _r, you are possible to set a target amount of each place state freely. Although the details will be described later, in the present embodiment, except the target state quantity M _r with respect to the initial state to enter by the input means 301, the state of the AGV to ever-transitions in the course of the simulation (above unladen, requested travel, transport, ... The setting content of the target state quantity _Mr is changed according to the state).

  In the present embodiment, the moving operation end (transition) in the above formula (2) or the above formula (3) coincides with the moving operation end of the actual process. That is, the signal of the operation vector u (k) is an actual process operation signal. As a result, a state vector in which a value obtained by subtracting the target state quantity from the number of tokens present in each place is represented by x, and an operation vector in which the presence / absence of firing of each transition is represented by “1” and “0” is represented by u. Then, using the connection matrix B, a state change from the vector x (k) to the vector x (k + 1) can be expressed as in the above equation (3).

  In this way, by representing the AGV travel path with the connection matrix B using the Petri net model, it is possible to express that the token moves through a place that requires a plurality of discrete step times as the discrete time k advances by one. As a result, the optimum control theory of the discrete time system can be applied.

As described above, in the present embodiment, the Petri net model can be made into a form to which the optimal control theory of the discrete time system can be applied. Therefore, appropriate evaluation matrices Q and R are set according to the purpose of the control, and the following: Considering that the operation vector u (k) is controlled so that the evaluation function J shown in Expression (5) is minimized,
u (k) = − K (M (k) −M _r ) = − K · x (k) (4)
The feedback gain matrix K for performing the state feedback control can be obtained from the optimal control theory.

  The above-described evaluation matrices Q and R are square matrices each having a dimension of the number of places and a dimension of the number of transitions, respectively, for representing the control transient characteristics and the input energy of the transition operation as evaluation functions, respectively. What is necessary is just to set suitably according to. For example, when the transient characteristics are emphasized, an evaluation matrix Q having a large numerical value may be set. Moreover, when it is desired to keep the input energy required for the operation small, an evaluation matrix R having a large numerical value may be set.

  If the connection matrix B and the evaluation matrices Q and R that define the predetermined evaluation function J are determined in this way, the feedback gain matrix K can be obtained by the optimal control theory in the discrete time system. As a method for calculating the feedback gain matrix K, for example, (Kazuaki Ando, et al., “Control system design by numerical analysis method” published by the Society of Instrument and Control Engineers, pp.126-130, the first edition of the 1993 second edition) There are several methods described.

  The feedback gain matrix K obtained by the gain matrix calculation unit 302 in this way is used when the control is performed in the operation vector calculation unit 202 and the control unit 213 together with the connection matrix B obtained by the model construction unit 212. Therefore, it is stored and stored in the storage means 201.

  Next, the state management means 205 receives and stores the AGV and coil states represented as tokens of the Petri net model from the host computer 100. The AGV states include the above-mentioned empty (Empty), requested travel (Request), load receiving (Load), transport (Carry), unloading (Unload), charging travel (Battery), charging (Charge), evacuation travel (Escape) ) Each state. In addition, the coil state means each state of unassigned, assigned, loaded, loaded, unloaded, and completed. In the present embodiment, the attribute (color) is given to the AGV and the coil token according to these states.

The transport command management unit 206 manages one or more transport commands given from the host computer 100, and based on the transport command and the AGV state managed by the state management unit 205, each token of AGV and coil And set attributes on them. Further, the target state quantity _Mr is set according to the given AGV state. Method for setting the target state quantity M _r is determined in advance by, for example, predetermined rules.

  As an example, when the AGV state is the requested travel (when the attribute is Request), the target value is set to the requested travel destination, that is, the From position (load receiving position) in the transport command. When the AGV state is transport (when the attribute is Carry), the target value is set at the transport destination, that is, the To position (unloading position) in the transport command.

The operation vector calculation means 202 is a feedback gain matrix K obtained by the gain matrix calculation means 302 and stored in the storage means 201, and a virtual transfer process inputted by the state vector input means 204 in each zone at the present time. the state vector M representing a progress status of the token (k), the presence or absence of fire and a target state quantity M _r after initial state to be supplied to each AGV from transportation command management unit 206, the moving operation end of tokens for (transition) An operation vector u (k) representing is obtained.

  Based on the operation vector u (k) obtained by the operation vector calculation unit 202, the control unit 213 checks the possibility of firing in descending order of the positive value of the operation amount (element of the operation vector u (k)). Until all the transfer commands are assigned to some AGV, using the method that fires all the operation ends that can be ignited for each movement operation end in a simple transfer process and sequentially performs the move operation as many as possible. A transportation process simulation is executed to find a solution Sol (0) of the vehicle allocation problem.

  That is, on the basis of the calculated operation vector u (k), for each transition, the possibility of firing is checked in order of positive and large operation amount, and the operation amount is changed again for the moveable number of tokens. “1” is set, and the operation amount is set to “0” for the other tokens, and the final operation vector u (k) is obtained. As a result, if the transition with the manipulated variable “1” is ignited, the state vector M (k + 1) can be obtained as the value {M (k) + B · u (k)} shown in the equation (1). It becomes.

At this time, when the simulation of a one time unit, a connection matrix B is input from the time storage unit 201, a target state quantity M _r like inputted from the transport instruction managing unit 206, the state vector input unit 204 The state vector M (k + 1) after the moving operation is obtained from the state equation using the current state vector M (k) input from the state, and the state vector is used for the simulation in the next time unit. Feedback is provided to the input means 204. Then, the simulation is sequentially performed in such a manner that as soon as one moving operation is virtually performed, the next moving operation is performed.

  The processing plan output unit 207 feeds back the result of the simulation performed by the control unit 213 as described above to the host computer 100. At that time, the processing plan output means 207 analyzes the recording file indicating which token has moved from where to where, and the data indicating how the AGV is assigned with the transfer instruction and the route through which the transfer is performed. Generate and output according to the format of the process control 100. The host computer 100 outputs the simulation result to an output device such as a display or a printer (not shown).

  By the way, the Petri net model of FIG. 3B described above shows a case where there is one AGV. When there are a plurality of AGVs, the same module as the number of AGVs is created with the model shown in FIG. 3B as one module. At this time, the destination of the traveling destination is determined for each AGV that is being transported and whose transport command assignment has been confirmed. On the other hand, the destination of the AGV of an empty vehicle (including one that is in evacuation) varies depending on how a new transfer command is assigned.

  Therefore, AGVs that are being transported and whose transport command assignment has been confirmed are distinguished for each vehicle as shown in the upper part of FIG. 4 (AGV1, AGV2, AGV3,...). No distinction is made as in the lower row (AGV0). That is, only one AGV can enter each of the upper modules in FIG. 4, but a plurality of empty AGVs can enter the lower module. A new transfer command is assigned to the empty vehicle and the AGV that is being evacuated.

  In each module, the corresponding zone (the position where the AGV skid or OCD is located) physically represents the same place and generally has a limited capacity. For example, skid A of AGV1 and skid A of AGV2 represent the same physical location, and when AGV1 is in skid A, if the capacity of skid A is 1, no other AGV can enter skid A .

  In order to realize such a restriction on software, the state management unit 205 may manage the number of tokens in each zone as a state related to the capacity in addition to the AGV state and the coil state. That is, each time the simulation of one time unit proceeds, the capacity of each zone (upper limit value of the number of tokens allowed to enter) is compared with the current number of tokens of each zone so that the capacity of each zone is exceeded. Prohibit excessive ignition.

  Note that FIG. 4 shows a diagram in which the modules of each AGV are clearly divided, but it is not essential to divide each module, and it is only necessary to distinguish each AGV on a mathematical model (on software).

By considering This separation Petri net model for each AGV, the target state quantity M _r that is set according to the AGV state by the conveyance instruction management unit 206 is also set for each AGV. Therefore, the evaluation function J shown in the above equation (5) also differs for each AGV. For example, if there are four AGVs, AGV1 is being transported, AGV2 is in the required travel, and AGV3 and AGV4 are empty, the respective evaluation functions J ₁ , J ₂ , J ₃ , ₄ (empty and evacuation travel) The case is considered as a whole) as shown in the following equation (6).

In the formula (6), M _r1 is the target state quantity for AGV1, this case is obtained by setting the transport destination of the coil, i.e., the To position in the allocation already transportation command as a target value. Further, M _r2 is the target state quantity for AGV2, this case is obtained by setting the transport origin of the coil, i.e., the From position in the allocation already transportation command as a target value. In addition, _Mr 3 and 4 are target state quantities for AGV3 and AGV4. In this case, the transfer source (From position) of an unassigned transfer command is set as the target value.

For example, the value of the element corresponding to the destination is set to “1” and the values of the other elements are set to “0” in the target state quantity _Mr matrix. Further, the target value may be set to a value other than “1”, for example, an integer value or a decimal value greater than that, and the values of other elements may not necessarily be “0”. In short, the value of the destination element may be set to be larger than the values of the other elements.

The operation vector u (k) that minimizes the evaluation functions J ₁ , J ₂ , J ₃ and ₄ for each AGV is obtained by the above equation (4) for each AGV. That is, in this case, three operation vectors u (k) ₁ , u (k) ₂ , u (k) _{3, 4} are obtained. The operation vector calculation means 202 is a vector u (k) (= [u (k) ₁ u (k) ₂ obtained by integrating the operation vectors u (k) ₁ , u (k) ₂ , u (k) ₃ , _4. u (k) _3,4 ]) is further generated.

  Based on the integrated operation vector u (k) obtained by the operation vector calculation unit 202, the control unit 213 checks the possibility of firing in descending order of the positive value of the operation amount, and moves each movement in the virtual manufacturing process. A simulation of the conveyance process for a predetermined period is executed by using a method in which all the operation ends that can be ignited with respect to the operation end are fired, and the moving operation is sequentially performed as many as possible.

  It should be noted that the model construction means 212, gain matrix calculation means 302, storage means 201, operation vector calculation means 202, state vector input means 204, state management means 205, transfer command management means 206, processing plan output means 207, and control described above. The means 213 is constituted by a microcomputer including, for example, a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), and the CPU work program is stored in the ROM or RAM. Yes.

Further, the gain matrix calculation means 302 corresponds to the optimization calculation apparatus 2 shown in FIG. 1, and a mathematical model ((1)) obtained by substituting the connection matrix B output from the model construction means 212 into the above formula (1). For example, the equation (2)) corresponds to the mathematical model 3 shown in FIG. The evaluation evaluation matrix Q to be inputted based on the operation by the user, the evaluation function obtained based on the R and the target state quantity M _r (the evaluation function J of Equation (5)) is shown in Figure 1 of the input means 301 It corresponds to the function S.

  Furthermore, the simulation control unit 200 corresponds to the simulation control unit 1 shown in FIG. In particular, the control unit 213 and the state vector input unit 204 correspond to the vehicle allocation simulator 1a shown in FIG. 1, and the operation vector calculation unit 202 corresponds to the optimal control policy unit 1b shown in FIG.

As described above, a method for modeling an AGV allocation problem with a Petri net, a method for transforming a Petri net model into a mathematical formula, a method for achieving a transfer purpose by an evaluation function in which a target state amount _Mr is set according to an AGV state, and a minimum evaluation function I described the principle of the method to make it.
Next, specific Petri net model settings for actual implementation based on such a principle will be described.

  5 and 6 are diagrams showing in detail the Petri net model in each zone shown in FIG. 3B, FIG. 5 shows the Petri net model of each AGV skid zone, and FIG. The Petri net model of the zone for use is shown. In FIGS. 5 and 6, white circles represent places, bar marks represent transitions, black circles within white circles represent tokens, and tokens move between places due to the firing of transitions.

  The place where “AGV” is written in each module is a place for AGV, and one place is provided in each module for AGV skid and OCD. Further, the places where “From” and “To” are written are places for coils, and one place is provided in each module for AGV skid. These places are limited to one capacity. Furthermore, the places written as “Common”, “Complete”, and “Request” are common places for each module, and only one place is provided in total. A plurality of tokens can be set in this common place.

  Also, “Load”, “Unload”, “Allocate”, “Reallocate”, and “Move” are written as transitions. Here, the transition “Load” can be ignited when there is a token in the place “AGV” and the place “From”. The transition “Unload” can be ignited when a token exists in the place “AGV”. The transition “Move” can be ignited when there is a token in the place “AGV”, and controls the movement of the AGV. The transitions “Allocate” and “Reallocate” can be fired arbitrarily.

  FIG. 7 is a rewrite of the Petri net model shown in FIG. 3B according to the specific examples of FIGS. As shown in FIG. 7, serial numbers (numbers in parentheses) are assigned to places “AGV”, “From”, and “To”, and serial numbers (circles) are assigned to transitions “Load”, “Unload”, and “Move”. Each number is distinguished by affixing the number in it. In addition to the state, each token of AGV and coil has attributes such as a current position, a number corresponding to the assigned coil and AGV, and a number corresponding to each of them.

  Also, as shown in FIGS. 5 to 7, in this embodiment, the transition “Move” is provided only in the direction going out of the module, and not in the incoming direction. By configuring in this way for all AGV skid and OCD modules, the entire Petri net model can be created simply by changing the number of input and output legs as needed and connecting each module. it can. Therefore, the design can be easily changed.

FIG. 8 is a flowchart showing an example of an operation at the time of simulation of the vehicle allocation plan creating apparatus 101 according to the present embodiment configured as described above. The operation will be described below with reference to the flowchart of FIG.
In FIG. 8, first, in step S <b> 1, a transport command and AGV state information are received from the host computer 100. Next, in step S2, preprocessing is performed based on the received information. In this pre-processing, AGV and coil tokens are generated and attributes are set respectively. The AGV token is set in the place “Common” and the coil token is set in the place “Request”.

  Next, in step S3, AGV allocation processing is performed using the ignition control of each transition based on the above-described optimal control theory. For example, assume that there are AGV and coil tokens as shown in FIG. In this embodiment, this amount of information is managed by the state management unit 205 and the transport command management unit 206. Based on these pieces of information, each token placed in the coil place “Request” is set to the coil token when the transition connected to the carrier place indicated by the attribute is ignited. . In addition, each token placed in the AGV place “Common” is assigned to each AGV token at the current position of each AGV by firing a transition connected to each AGV. At this time, a target value is assigned to the place of the transfer source or transfer destination according to each AGV state.

  In the present embodiment, among the AGV states, except for “Empty” and “Escape”, since the destination of travel is determined, the original AGV number is given as the AGV color (attribute) ( In this example, AGV1 is color 1 and AGV2 is color 2). The states of “Empty” and “Escape” give a color (for example, color 0) indicating that the transfer command can be assigned. The coil state follows the same concept. When the assignment has been confirmed (other than “Unassigned”), the color of the assigned AGV number is given, and when the assignment is unassigned, an unassigned number (for example, color 0) is given.

  By assigning color numbers in this way and enabling the transition "Load" to be fired only when tokens with the same number are aligned, correct allocation of assigned coils and AGVs, and unassigned ones is realized. Can do.

Since the state vector M (k) and the target state quantity M _r can be calculated based on the token set in each place and the target value by the above method, the operation vector u (k) is calculated by the above equation (4). it can. The priority of transition firing is given by the operation vector u (k), and the transition to be fired is determined according to the above-described method based on the operation vector u (k). When the transition is actually fired, the state M (k + 1) of the next step is obtained, and the transitions that are fired sequentially are determined.

  FIG. 10 shows each transition firing and the state (attribute) transition caused thereby, FIG. 10 (a) shows the transition firing, and FIG. 10 (b) shows the state transition. In FIG. 10, when the AGV in the requested travel (Request) arrives at the transport source (From position), the transition “Load” is ignited, and the AGV state changes to “Load”.

  Also, the AGV for empty or evacuation (Escape) is given a color indicating unassigned, but by setting a target value to the transport source when setting a token for an unassigned coil, The AGV during the retreat travel can be assigned to the transport command (coil). That is, the AGV and the coils that are not assigned to each other change the AGV state to “Load” when the transition “Load” is ignited, and thereafter perform the same operation as “Request”.

  Next, the transition “Reallocate” is fired, and the AGV state changes to “Carry”. At this time, the token of the AGV having the attribute “Carry” moves to the place “Common”, and the target value is obtained by the firing of the transition “Allocate” in the same manner as when the AGV and the coil being transferred are set. Is set at the transport destination and transport is started. The transfer is performed by the ignition of the transition “Move”, but the AGV state does not change by this ignition.

  When the AGV being transported arrives at the transport destination (To position), the transition “Unload” is fired, and the AGV state changes to “Unload”. At the same time, a coil token with the attribute “Unload” is also generated. Next, the transition “Reallocate” is ignited, and the AGV state is set to “Request” when there is a coil to be conveyed next, and to “Empty” when there is no coil to be transferred.

  When the coil token moves to the transport destination by the traveling of the AGV, the coil token moves to the place “Complete”. Then, when all the coils have been transported, the AGV allocation process in step S3 of FIG. As described above, only one place “Complete” is provided in common for all modules, and a plurality of tokens can be set. Therefore, when you look inside this place “Complete”, you can see which coil's token has been transported in what order.

Then, post-processing is performed in step S4. In this post-processing, data (data such as an assigned AGV number of an unassigned coil and an AGV travel route) to be returned to the host computer 100 is extracted from a file representing token movement stored as a final result. At this time, when there is a transfer margin in the unassigned coil, and it is not particularly necessary to carry it urgently, in order to suppress interference between the AGVs, the transfer time of the coil when simulated with the Petri net is free from AGV interference. If it is greatly different from the transfer time, the unassigned coil may not be returned to the host computer 100.
Finally, in step S5, the extracted data is converted into the format of the host computer 100 and transmitted.

As described above in detail, according to the present embodiment, a Petri net model in which the traveling path of the AGV is represented by a place and a transition and the coil and the AGV are represented by a token is constructed. Then, the target state quantity M _r so as to variably set according to a transfer instruction and AGV state, the operation of the target state quantity M _r from the state equation in consideration for the transition of the coil and AGV vector u (k) Since the movement operation of each coil and AGV is performed based on this operation vector u (k), the optimal control theory of the discrete time system using the Petri net model is applied to the assignment problem for the AGV of the transfer command. It can be shaped.
In particular, when the scale of the transport process is not as large as the difference, since the number of elements of the evaluation matrices Q and R is small, recalculation of the feedback gain K does not take time, and the evaluation matrices Q and R can be adjusted online. It is possible to shorten the traveling time of the AGV and shorten the waiting time for transporting the transported object. Therefore, when the scale of the transfer process is not so large, the performance desired by the coordinator can be achieved while simplifying the entire configuration.

(Second Embodiment)
Next, a second embodiment of the present invention will be described. In the following description, detailed description of the same parts as those of the first embodiment described above will be omitted.
As described above, in the vehicle allocation plan creating apparatus of the first embodiment, in order to adjust the vehicle allocation performance, it is necessary to change the elements of the evaluation functions Q and R and recalculate the feedback gain. And if the scale of a conveyance process becomes large, calculation of a feedback gain will require time. For this reason, it is necessary to calculate the feedback gain offline. Thus, when the scale increases, it is difficult for the coordinator to easily adjust the vehicle allocation performance offline. In addition, the influence of the evaluation matrices Q and R on the vehicle allocation performance becomes difficult for the coordinator to understand.

  Therefore, in the present embodiment, by solving such a problem in the first embodiment and efficiently assigning the transport command to the AGV, the travel time of the AGV is shortened or the transported object is carried out. This makes it possible to shorten the transport waiting time and to optimally adjust the trade-off between the two.

FIG. 11 is a block diagram illustrating an example of an overall configuration of a scheduling system in which a vehicle allocation plan creating apparatus for an automatic guided vehicle (AGV) according to a second embodiment of the present invention is implemented.
As shown in FIG. 11, the scheduling system of the present embodiment is similar to the scheduling system (allocation plan creation apparatus 101) of the first embodiment shown in FIG. The combination calculation unit 209, the target value change unit 210, and the cost function comparison unit 211 are added.

  The cost function setting means 208 sets a cost function defined from the current position and current state of each AGV, and the transfer source and transfer destination of the coil assigned to each AGV. For example, the cost function is an equation obtained by linearly combining the total empty running time of each AGV and the total conveyance waiting time of each coil with coefficients Ka and Kb as in the following equation (7).

  The coefficients Ka and Kb in the above formula (7) are adjustment coefficients that influence the vehicle allocation performance of the AGV. In the vehicle allocation plan creating apparatus according to the present embodiment, the allocation result between the AGV and the coil in which the value of the cost C in the above formula (7) becomes small is obtained. At this time, if the adjustment count Ka is increased, the AGV idle time is reduced, and the energy cost of the AGV for transferring the coil is reduced, so that the number of conveyances per unit time is increased. However, a coil that is far away, which takes a long time to travel empty, is difficult to be transported, and there is a disadvantage that a transport delay occurs. On the other hand, if the adjustment count Kb is increased, the transfer waiting time of the coil is reduced, so that there is an advantage that transfer delay is less likely to occur. However, the energy cost of the AGV is relatively increased, and in a long time, There is also a drawback that the number of transported units per unit time is reduced.

  The vehicle allocation combination calculation means 209 inputs the current position of each AGV and the AGV state (empty vehicle, requested travel, cargo receiving, transport, unloading, charge travel, charge, evacuation travel) from the state management means 205 and a transport command. The transfer source and transfer destination of each coil are input from the management means 206, and the solution of the vehicle allocation problem in which the value of the cost function of the equation (7) set by the cost function setting means 208 becomes small (which coil is selected with which AGV) Calculate the order of transport). Here, when calculating the cost function of the above formula (7), the calculation is performed based on the standard travel time that does not include interference between AGVs in order to shorten the calculation time. In addition, when a strict method such as a branch and bound method is used, since there are many combinations, it cannot be solved in a practical time. Therefore, using combinatorial optimization methods such as tab search, GA, and simulated annealing, upper limits are set on the number of iterations and calculation time, and the top L solutions (Sol (1) to Sol (L )). A method for finding a solution to the vehicle allocation problem using tab search will be described later.

In order to make it easier to find a solution that reduces the value of the cost function in the above equation (7) in the vehicle allocation calculation by the optimal control, the adjustment coefficients Ka and Kb set by the cost function setting unit 208 by the target value changing unit 210 are obtained. It is used to change the target value M _r. Specifically, since there is a correlation between increasing the target value and the ratio (Kb / Ka) of the total AGV idle time and the total coil waiting time, the target value is expressed by the following formula (8 ).
M _r <-M _r (0.1 + Kb / Ka) (8)

  The simulation means 203 recalculates the value of the cost function of the above equation (7). At this time, since the simulation is performed including the interference between the AGVs, the value of the cost function of the above formula (7) is calculated as a strict value close to the actual traveling including the interference between the AGVs. Here, Cost (0) is the cost function value obtained when simulating the solution Sol (0) of the vehicle assignment problem (obtained by the control means 213) obtained by the vehicle assignment calculation simulation by optimal control, and tab search is performed. Cost (1) to Cost (L) are values of the cost function obtained when simulating Sol (1) to Sol (L) of the vehicle allocation problem obtained in (1).

  The cost function value comparison means 208 simply compares the cost function values Cost (0) to Cost (L) to obtain the smallest value, and the corresponding recording file output by the simulation (which token is from when and where Only the file that records where it moved) is left as the result of the final dispatch problem.

  The simulation unit 203, the cost function setting unit 208, the vehicle allocation combination calculation unit 209, the target value changing unit 210, and the cost function comparison unit 211 described above are, for example, a CPU (central processing unit), a RAM (random access memory). And a microcomputer composed of a ROM (Read Only Memory) or the like, and a CPU work program is stored in the ROM or RAM.

FIG. 12 is a flowchart showing an example of an operation at the time of simulation of the vehicle allocation plan creating apparatus 101 according to the present embodiment configured as described above. The operation will be described below with reference to the flowchart of FIG.
In FIG. 12, in step S <b> 11, similarly to step S <b> 1 in FIG. 8, a transport command and AGV state information are received from the host computer 100. Next, in step S12, as in step S2 of FIG. 8, preprocessing is performed based on the received information.

  Next, in step S13, similarly to step S3 of FIG. 8, AGV allocation processing is performed using the ignition control of each transition based on the above-described optimal control theory. However, in the present embodiment, the AGV combination allocation process is performed in step S14 in parallel with the AGV allocation process. In this AGV combination allocating process, a plurality of combinations that reduce the value of the cost function of the above equation (7) are calculated from the AGV token and coil token obtained in the preprocessing. This method includes various methods such as tab search, GA, and simulated annealing. Here, tab search is described as an example. An example of a solution to the vehicle allocation problem by tab search is shown in FIG.

  Here, a case where the coil and the AGV token are as shown in FIG. 9 will be described as an example. However, it is assumed that the layout of the travel path is as shown in FIG. 14 and that the travel time between the grids of AGVs 1201 to 1204 takes 60 seconds for empty travel and 90 seconds for loading travel. Further, the receiving time and unloading time of the coils 1211 to 1214 are both 10 seconds. For example, in FIG. 14, it takes 800 seconds (= 60 × 4 + 10 + 90 × 6 + 10) for the AGV presently present at the position A1 to move from the position C1 to the position D1 until it becomes empty at the position D1. It becomes. Further, the adjustment coefficients Ka and Kb of the cost function of the above formula (7) are set to 10.0 and 1.0, respectively.

  In FIG. 13, in step S21, an initial solution is obtained by a greedy method. There are various methods for the greedy method. Here, each unassigned coil is examined in turn, and the coil is assigned to the AGV that can start transporting earliest. That is, the unassigned coil No. 100 is assigned to the No. 3 car, and the unassigned coil No. 103 is assigned to the No. 4 car (see FIG. 9). This combination of AGV and coil is described as shown in Table 1. In Table 1, the coil numbers with parentheses indicate that the operation is already in progress and cannot be moved because it has been confirmed.

  Next, in step S22, any two unassigned coils are selected and exchanged. In the example shown in FIG. 9, there are only combinations of the coil numbers 100 and 103. Therefore, when the coil numbers 100 and 103 are exchanged, the combinations of AGVs and coils are as shown in Table 2.

  Then, among the combinations after replacing the unassigned coils, a combination A having the smallest cost function value is obtained. In this example, since there is only one combination to be exchanged, the combination A having the smallest cost function value is the case shown in Table 2.

  Next, in step S23, any two combinations of unassigned coils are selected, and one is inserted in front of the other. In the case of this example, there are two types of Tables 3 and 4.

  Of these combinations, a combination B having the smallest cost function value is obtained. In the case of this example, the combination B with the smallest cost function value is the case of Table 3.

  Next, in step S24, one arbitrary coil is selected and used as a coil to be conveyed at the end of any AGV. In this example, there are six patterns shown in Tables 5 to 10.

  Therefore, in this example, the combination C having the smallest cost function among the combinations after selecting any one coil and performing the operation to make the coil to be conveyed at the end of any AGV is the case of Table 5. .

  Next, in step S25, among the combinations A to C obtained in steps S22 to S24, the combination having the smallest cost function value is obtained. In this example, the case of Table 5 is obtained. This solution is set as a new solution, and this solution is added to the tab list. When the combinations A to C are obtained by executing the processing of steps S22 to S24, the smallest solution other than the solutions existing in the tab list is searched.

  Then, in step S26, it is determined whether or not such an operation has been repeated a specified number of times, and among the solutions obtained by the specified number of times of processing, the top L items from the smallest are stored.

  After performing the AGV combination allocation processing as described above, in the flowchart S15 of FIG. 12, the solution problem Sol (0) obtained by the optimal control and the L solutions Sol ( 1) to Sol (L) are respectively simulated, and cost function values Cost (0) to Cost (L) corresponding to the respective solutions are calculated. At this time, since all the coils are assigned to some AGV, there is no unassigned coil. The simulation is performed with all the coils that have been assigned. That is, all AGV tokens enter the module corresponding to the car number on the upper side in FIG.

  Specifically, when simulating in step S15, the value of the cost function of equation (7) is obtained by simulation. In the simulation, a file representing token movement is created.

  In the comparison process in step S16, the cost function values calculated from the simulation based on the optimal control and the simulation based on the L solutions obtained by the tab search are compared to obtain the smallest value, A file representing the movement of the corresponding token is stored as the final result.

  Then, in FIG. S17 and S18, post-processing and allocation result transmission processing similar to those in steps S4 and S5 in FIG. 8 are performed.

In this embodiment as described above, to build a traveling AGV Petri net model, the goal set target state quantity M _r in accordance with the conveyance instructions and AGV state, operation from a state equation of the Petri net vector u ( k) is determined. The control means 213 obtains a solution to the vehicle allocation problem using the optimal control theory by performing a moving operation of each coil and each AGV based on this operation vector. On the other hand, a cost function setting unit 208 sets a function that is a linear sum of the total AGV empty travel time and the total coil waiting time, and a vehicle allocation calculation unit 209 solves the vehicle allocation problem that minimizes the cost function. Are obtained by tab search. Then, the simulation unit 203 performs simulations using the solutions of the above-mentioned allocation problem, and the cost function comparison unit 211 obtains a solution with the smallest cost function, and returns the solution to the host computer 101. As a result, the AGV travel time is shortened, the transport waiting time until the transported object is carried out is shortened, and the optimal allocation considering the trade-off between the AGV empty travel time and the coil transport waiting time is performed. You will be able to

(Third embodiment)
Next, a third embodiment of the present invention will be described. In the following description, detailed description of the same parts as those of the first embodiment described above will be omitted.
In AGV operation control, there is a case where concentration of AGV becomes a problem. It is good if the coils are generated evenly in multiple skids, but if the coils are concentrated in one place, the AGV will be concentrated there and the interference will become very large, or it may cause deadlock. . Therefore, it is desirable that the AGV is dispersed as much as possible.

  In order to realize this, a traffic congestion function whose value increases as AGV concentrates is defined, a cost function is defined as in the following equation (9), and a solution with a value as small as possible is obtained. The method described above may be used.

  An example of the definition and calculation method of the congestion degree will be described. First, the skid on the runway is divided into a plurality of areas, and if the number of AGVs seated in each area exceeds the designated number, the degree of congestion is set to a large value. The seated AGV refers to an AGV that is empty or being received and unloaded, and does not include an AGV that is passing through the area.

  Next, an example of calculating the congestion degree will be described. For simplicity, it is assumed that the number of areas is 3, the number of AGVs is 3, and the number limit values of the areas are all 2. Then, consider a two-dimensional matrix (called a traffic jam matrix) in which rows are areas and columns are times. Here, the time is discretized every minute, and the number of columns is 10 (minutes). Each element of the traffic jam matrix represents the total number of AGVs present at that time in the area. For example, if an AGV in Area 1 is present, leaves the skid in Area 1 after 3 minutes, and arrives at the skid in Area 2 after 5 minutes, the congestion matrix is as shown in Table 11 below.

  One AGV is in Area 1 and two AGVs are in Area 2, and the AGV in Area 1 is assigned to a transfer command from Area 1 to Area 2, leaving the skid in Area 1 after 3 minutes, If it arrives at the skid of area 2 after 5 minutes, the traffic jam matrix is as shown in Table 12 below.

  Here, the degree of congestion is defined by the following equation (10).

The (2, 5) to (2, 10) elements of the previous matrix exceed the maximum number 2 of areas. Therefore, the degree of congestion is 6.
On the other hand, when the AGV in area 2 carries the same transport command, it leaves area 2 after 0 minutes, arrives at area 1 after 2 minutes, leaves area 1 after 4 minutes, and returns to area 2 after 6 minutes. Then, the traffic jam matrix is as shown in Table 13 below.

  In this case, since all elements do not exceed the maximum number of areas, the degree of congestion is zero. This congestion degree is multiplied by a coefficient Kc and added to the cost function. By setting the count Kc to a large value, it is possible to obtain a vehicle allocation with a low degree of traffic congestion (AGV is less likely to concentrate at one location).

  The above is a case where the amount of charge is not taken into account, but the actual AGV is often battery-driven, and the battery is consumed as the AGV works. In this case, it is necessary to operate the AGV so that the remaining battery level does not fall below a certain lower limit. For example, as shown in FIG. 15, when a charging station 1301 is provided at an appropriate place and the batteries of the AGVs 1311 to 1314 are close to the lower limit, the coil transportation is temporarily interrupted, traveled toward the charging station, and the charging station Then, the charging of the coils 1321 to 1324 is resumed after charging.

  The present embodiment can also be applied when considering such a charge amount. For example, if one charge is always charged when one transfer is completed, the empty vehicle travel time from the skid to the charging station after completion of the transfer of a certain coil and the empty vehicle from the charging station to the load receiving skid of the next coil to be transferred The travel time is considered as the all-empty travel time, and the transfer waiting time of the second coil to be transported may be increased by the increase in the empty travel time and the charge time.

(Fourth embodiment)
Next, a fourth embodiment of the present invention will be described. In the following description, detailed description of the same parts as those in the first to third embodiments described above will be omitted.
When the skid in which the coil is generated cannot be specified as in the case of coil conveyance at a steelworks, it is desirable that the AGV be dispersed appropriately without concentrating it at one place. Therefore, in the present embodiment, when the skid is divided into a plurality of areas and the number of empty vehicles and the number of AGVs being unloaded in the area exceeds the specified number, the empty vehicles AGV are separated so as to be equal to or less than the specified number. AGV can be dispersed by retreating to the area.

  For example, as shown in FIG. 16, the skid is divided into four areas. If each allowable number is set as shown in Table 14, the number of areas 1402 (area 2) is two and exceeds the allowable number, so the third car 1413 or 4 that is an empty AGV. Car No. 1414 is evacuated to another area 1401, 1403, 1404 (area 1, 3, 4). The evacuation destination is located near the area 1402, and an area having a margin for the allowable number is searched. In the example illustrated in FIG. 16, the number of AGVs in the area 1403 is 0, so the save destination is the area 1403. The AGV to be evacuated is the AGV closer to the area 1403, and the evacuation location of the area 1403 is the skid closest to the AGV. That is, the fourth car 1414 is retracted to the skid D1. Thereby, the number of AGVs in each of the areas 1401 to 1404 is all 1, and the AGVs 1411 to 1414 can be distributed.

In each of the embodiments described above, an example in which the coil is transported on the steel production line is given as an example of the transport system, but the present invention is not limited to this.
In each of the above embodiments, AGV is used to transport an object, but the present invention can be applied to other transport systems such as an automatic claim and a carrier pallet.
Moreover, although the case where this invention was applied to the simulation apparatus was demonstrated in the said embodiment, it can also be applied to an actual process and used as a physical distribution control apparatus.

(Other embodiments of the present invention)
In order to operate various devices in order to realize the functions of the above-described embodiments, a program code of software for realizing the functions of the above-described embodiments is provided to an apparatus connected to the various devices or a computer in the system. What is implemented by operating the various devices according to a program supplied and stored in a computer (CPU or MPU) of the system or apparatus is also included in the scope of the present invention.

  In this case, the program code of the software itself realizes the functions of the above-described embodiments, and the program code itself and means for supplying the program code to the computer, for example, the program code are stored. The recorded medium constitutes the present invention. As a recording medium for storing the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

  Further, by executing the program code supplied by the computer, not only the functions of the above-described embodiments are realized, but also the OS (operating system) or other application software in which the program code is running on the computer, etc. It goes without saying that the program code is also included in the embodiment of the present invention even when the functions of the above-described embodiment are realized in cooperation with the embodiment.

  Further, after the supplied program code is stored in the memory provided in the function expansion board of the computer or the function expansion unit connected to the computer, the CPU provided in the function expansion board or function expansion unit based on the instruction of the program code Needless to say, the present invention also includes a case where the functions of the above-described embodiment are realized by performing part or all of the actual processing.

It is a block diagram which shows embodiment of this invention and demonstrates an example of schematic structure of a vehicle allocation plan preparation apparatus. 1 is a block diagram illustrating an overall configuration of a scheduling system that implements a vehicle allocation plan creation device for an automated guided vehicle (AGV) according to a first embodiment of this invention. It is a figure which shows the 1st Embodiment of this invention and shows an example of a Petri net model. It is a figure which shows the 1st Embodiment of this invention and shows a mode that the module of a Petri net model is produced | generated for each AGV. FIG. 4 is a diagram showing in detail a Petri net model in each AGV skid zone shown in FIG. 3B according to the first embodiment of the present invention. FIG. 4 is a diagram showing in detail a Petri net model in the OCD zone shown in FIG. 3B according to the first embodiment of the present invention. FIG. 6 shows the first embodiment of the present invention and is a diagram in which the Petri net model shown in FIG. 3B is rewritten according to the specific examples of FIGS. 4 and 5. It is a flowchart which shows the 1st Embodiment of this invention and demonstrates an example of the operation | movement at the time of the simulation of the allocation plan preparation apparatus of AGV. It is a figure which shows the 1st Embodiment of this invention and shows an example of the token of AGV and a coil. It is the figure which showed the 1st Embodiment of this invention and showed the mode of each transition ignition and the state transition by it. It is a block diagram which shows the whole structure of the scheduling system which showed the 2nd Embodiment of this invention and implemented the dispatch plan preparation apparatus of an automatic guided vehicle (AGV). It is a flowchart which shows the 2nd Embodiment of this invention and demonstrates an example of the operation | movement at the time of simulation of the allocation plan preparation apparatus of AGV. It is a flowchart which shows the 2nd Embodiment of this invention and demonstrates an example of operation | movement of the AGV vehicle allocation plan preparation apparatus at the time of calculating | requiring the vehicle allocation combination by a tab search. It is a figure which shows the 2nd Embodiment of this invention and shows an example of the traveling path of AGV. It is a figure which shows the 3rd Embodiment of this invention and shows an example of the traveling path of AGV which added the charging field. It is a figure which shows the 4th Embodiment of this invention and shows an example of the traveling path of AGV divided | segmented into the several area.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1,200 Simulation control part 1a Vehicle dispatch simulator 1b Optimal control policy part 2 Optimization calculation apparatus 3 Numerical formula model 4 Simulation result 100 High-order computer 101 Vehicle dispatch plan preparation apparatus 201 Storage means 202 Operation vector calculation means 203 Simulation means 204 State vector input means 205 State management means 206 Transport command management means 207 Processing plan output means 208 Cost function setting means 209 Vehicle allocation combination calculation means 210 Target value changing means 211 Cost function value comparison means 212 Model construction means 213 Control means 301 Input means 302 Gain matrix calculation means

Claims (13)

A vehicle allocation plan creation device that represents a transportation process by a Petri net model, and determines allocation of a transportation command to a transportation vehicle from the Petri net model,
As a Petri net model of the transport process, model construction means for constructing a Petri net model in which a traveling path of the transport vehicle is represented by a place and a moving operation end, and a transport object and a transport vehicle are represented by tokens;
Gain matrix calculation means for obtaining a state equation of the Petri net model constructed by the model construction means and for obtaining a feedback gain matrix from the obtained state equation and an evaluation function represented by the set evaluation matrix and target state quantity When,
State vector input means for obtaining a state vector representing the in-process state of the token in the transport process;
The feedback gain matrix obtained by the gain matrix calculating means, the state vector representing the in-process state obtained by the state vector input means, and the target state quantity set in accordance with the state of the transport vehicle and the given transport command And an operation vector calculation means for obtaining an operation vector for the moving operation end of the transported object and the transport vehicle,
Based on the operation vector obtained by the operation vector calculation means, each transported object and each transport vehicle can be moved in the order in which a large positive operation amount in the operation vector is obtained for each moving operation end. Obtain the state vector after the operation by performing an operation to move as many as possible, and use the obtained state vector to find the solution to the vehicle allocation problem that assigns each transported object to each transport vehicle. Control means,
The vehicle allocation plan creation device, wherein the control means feeds back the obtained state vector after the operation to the state vector input means as a new state vector representing an in-process state of the token in the transport process.   The model construction means constructs a Petri net model by combining a Petri net model module for skid, where the transport vehicle performs work, and a relay Petri net model module, where the transport vehicle communicates. The vehicle allocation plan creation device according to claim 1. The model construction means constructs the Petri net model for each transport vehicle, uses a Petri net model for each transport vehicle to which the transport command has been assigned, and uses one Petri net model for transport vehicles to which the transport command has not been allocated. Use Petri net models together,
3. The vehicle allocation plan creation device according to claim 1, wherein the transfer command is assigned to the unassigned transfer vehicle.   The said control means is large with the positive value in the said integrated operation vector with respect to each moving operation end based on the operation vector which integrated the operation vector calculated for every said conveyance vehicle by the said operation vector calculation means. 4. The vehicle allocation plan creation device according to claim 3, wherein an operation is performed to move each transported object and transport vehicle by a movable number in the order in which the operation amount is obtained.   The vehicle allocation plan creation device according to any one of claims 1 to 4, further comprising state management means for controlling and managing state transition of the transport vehicle according to firing of the moving operation end. A cost function setting means for setting a cost function defined by the position and work state of each transport vehicle, and the transport source and transport destination of each transport object assigned to each transport vehicle;
Target value changing means for changing the target state quantity as a function of the coefficient ratio of the cost function;
A vehicle combination calculation means for obtaining a plurality of solutions for a vehicle allocation problem for allocating each transported object to each transport vehicle using a combination optimization method so that the value of the cost function is as small as possible,
Simulation means for performing a simulation for each of the solution of the vehicle assignment problem obtained from the control means and the solution of the vehicle assignment problem obtained by the vehicle combination calculation means, and calculating the value of the cost function;
Comparing the values of all cost functions calculated by the simulation means, and having a cost function comparison means for outputting a solution of a dispatch problem corresponding to the smallest value,
The operation vector calculation means includes a feedback gain matrix obtained by the gain matrix calculation means, a state vector representing an in-process state obtained by the state vector input means, and a target state quantity changed by the target value changing means. The vehicle allocation plan creation device according to any one of claims 1 to 5, wherein an operation vector with respect to a moving operation end of the transported object and the transport vehicle is obtained by using.   The said cost function setting means uses the linear sum of the monotonically increasing function of the empty travel time of each transport vehicle and the monotonically increasing function of the transport waiting time of each transport object as the cost function. Vehicle allocation plan creation device.   The cost function setting means includes a monotonically increasing function of an empty vehicle traveling time of each transport vehicle, a monotonically increasing function of a transport waiting time of each transport object, and a function that increases in value as the transport vehicles concentrate near a specific zone. The vehicle allocation plan creation device according to claim 6, wherein the linear sum is used as a cost function.   The vehicle allocation combination calculation means is configured so that the time required for each transport vehicle to travel to the nearest battery field and the nearest battery field so that the remaining battery level of each transport vehicle does not fall below a specified value. The vehicle allocation plan creation device according to any one of claims 6 to 8, wherein a solution of the vehicle allocation problem is obtained in consideration of a charging time of each of the transport vehicles.   The vehicle allocation combination calculation means divides the traveling path of the transport vehicle into a plurality of areas, and when the number of empty vehicles and unloading transport vehicles in the area exceeds the specified number, The vehicle allocation plan creation device according to any one of claims 6 to 9, wherein a solution of the vehicle allocation problem is obtained so that the transport vehicle is evacuated to another area so as to be equal to or less than a specified number. Expressing a transportation process with a Petri net model, a vehicle allocation plan creation method for determining allocation of a transportation command to a transportation vehicle from the Petri net model,
As a Petri net model of the transport process, a model construction step of constructing a Petri net model in which a traveling path of the transport vehicle is represented by a place and a moving operation end, and a transport object and a transport vehicle are represented by tokens;
A gain matrix calculation step for obtaining a state equation of the Petri net model constructed by the model construction step and obtaining a feedback gain matrix from the obtained state equation and an evaluation function represented by the set evaluation matrix and target state quantity When,
A state vector input step for obtaining a state vector representing the in-process state of the token in the transport process;
The feedback gain matrix obtained in the gain matrix calculation step, the state vector representing the work state obtained in the state vector input step, and the target state quantity set in accordance with the state of the transport vehicle and the transport command to be given And an operation vector calculation step for obtaining an operation vector for the moving operation end of the transported object and the transport vehicle, and
Based on the operation vector obtained in the operation vector calculating step, each transported object and each transport vehicle can be moved in the order in which a large operation amount with a positive value in the operation vector is obtained for each moving operation end. Obtain the state vector after the operation by performing an operation to move as many as possible, and use the obtained state vector to find the solution to the vehicle allocation problem that assigns each transported object to each transport vehicle. A control step,
In the state vector input step, the post-operation state vector obtained in the control step is acquired as a new state vector representing a token in-process state in the transport process. A computer program for representing a transfer process by a Petri net model and for causing a computer to execute assignment of a transfer command to a transfer vehicle from the Petri net model,
As a Petri net model of the transport process, a model construction step of constructing a Petri net model in which a traveling path of the transport vehicle is represented by a place and a moving operation end, and a transport object and a transport vehicle are represented by tokens;
A gain matrix calculation step for obtaining a state equation of the Petri net model constructed by the model construction step and obtaining a feedback gain matrix from the obtained state equation and an evaluation function represented by the set evaluation matrix and target state quantity When,
A state vector input step for obtaining a state vector representing the in-process state of the token in the transport process;
The feedback gain matrix obtained in the gain matrix calculation step, the state vector representing the work state obtained in the state vector input step, and the target state quantity set in accordance with the state of the transport vehicle and the transport command to be given And an operation vector calculation step for obtaining an operation vector for the moving operation end of the transported object and the transport vehicle, and
Based on the operation vector obtained in the operation vector calculating step, each transported object and each transport vehicle can be moved in the order in which a large operation amount with a positive value in the operation vector is obtained for each moving operation end. Obtain the state vector after the operation by performing an operation to move as many as possible, and use the obtained state vector to find the solution to the vehicle allocation problem that assigns each transported object to each transport vehicle. Control steps to the computer,
The state vector input step obtains the post-operation state vector obtained in the control step as a new state vector representing a token in-process state in the transport process. A computer-readable recording medium having the computer program according to claim 12 recorded thereon.

Images