JP2009069867A - Action route search device, method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an optimum solution in real time, and to perform optimum solution derivation not affected by teacher data. <P>SOLUTION: In this action route search device 10, a computer has: a solution candidate generation part 11 generating a plurality of solution candidates; an evaluation part 12 performing conformity evaluation of each the solution candidate; a selection part 13 selecting the solution candidate based on the conformity evaluation; and a genetic operation part 14 creating a next-generation solution candidate by performing genetic operation to the solution candidate selected by the selection part 13. In the action route search device 10, a parameter for determining movement is allocated to at least one non-terminal symbol constituting the solution candidate. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an action route search apparatus, method, and program for searching an action route using a genetic algorithm (hereinafter referred to as “GA”).

GA is a stochastic optimization algorithm that mimics the evolution of organisms. The outline of GA will be briefly described below.
First, solution candidates for the optimization problem are expressed as, for example, chromosomes that are primary strings of genes, and solution candidates having various genes are randomly selected to create an initial generation population (population). Next, the fitness of each solution candidate (the value of the objective function in the optimization problem) is calculated. Subsequently, a plurality of solution candidates are selected (replication selection) according to the fitness, and genetic crossover (crossover; see FIG. 5), mutation (see FIG. 6), inversion, etc. of these genes By performing the operation, a group of next-generation solution candidates is created. At this time, in selecting a solution candidate, a solution candidate having a higher degree of suitability to the environment (objective function) is selected so as to survive with a higher probability in the next generation. Then, by repeating this generation change, solution candidates with genes with low fitness are deceived, solution candidates with genes with high fitness survive, and finally obtain a solution that optimizes the objective function more Can do.
Masashi Iba, “Method of Evolutionary Computation”, The University of Tokyo Press, 1999

  By the way, when using the above-mentioned GA for actual optimization problems, it is necessary to generate an algorithm that can give various inputs as teacher data and output appropriate to those problems as a pre-stage of operation. There is. In other words, an optimization algorithm is obtained by performing optimization with the algorithm itself as an optimization target. At this time, conditional branches and operators are assigned to the non-terminal symbols constituting the solution candidates. The optimal algorithm obtained in this way is used in the subsequent operation of the optimization problem.

When the optimization is performed with the algorithm as the optimization target as described above, the input data given as the teacher data is limited. Therefore, when input data that is far from the teacher data is input during operation, There was a problem that the solution could not be obtained.
In addition, since it is necessary to perform calculation processing for generating an optimization algorithm as a stage before actual operation, there is a problem that it takes time until actual operation.

  The present invention has been made to solve the above problem, and is an action path search apparatus and method capable of obtaining an optimal solution in real time and realizing optimal solution derivation that is not affected by teacher data. The purpose is to provide a program.

In order to solve the above problems, the present invention employs the following means.
The present invention is an action path search device that includes a computer and searches for an optimal action path using a genetic algorithm, wherein the computer generates a plurality of solution candidates, and each of the solution candidates An evaluation unit that performs a conformity evaluation, a selection unit that selects the solution candidate based on the conformity evaluation, and a next-generation solution candidate by performing a genetic operation on the solution candidate selected by the selection unit A behavior path search device is provided in which a parameter for determining motion is assigned to at least one non-terminal symbol constituting the solution candidate.

  According to the present invention, the suitability evaluation of each solution candidate is performed by the evaluation unit, and a plurality of solution candidates are selected and duplicated by the selection unit based on this suitability evaluation. Then, genetic operations such as crossover rate, mutation rate, and inversion rate are performed on the selected and duplicated solution candidates by the genetic operation unit, and a solution candidate group in the next generation is generated. Then, the above-described units sequentially perform processing, so that generations of solution candidates are repeatedly changed, and the search for the optimum action route is advanced. In this case, since at least one non-terminal symbol constituting the solution candidate is assigned a parameter that determines the motion, the action path can be directly derived as the solution candidate, and the finally obtained solution is obtained. Candidates can be output as output data as they are. Thus, by devising the configuration of the solution candidates, it is possible to omit the calculation process of the optimal algorithm that was performed prior to operation as in the past, and to obtain the action route in real time. . Furthermore, since teacher data or the like is not necessary, it is possible to obtain an appropriate action route for any input data.

In the behavior route search apparatus, the parameter that determines the movement is, for example, a needle change or a shift.
In the behavior route search apparatus, a program for setting an initial position may be assigned to at least one non-terminal symbol constituting the solution candidate.

  The present invention is an action path search method for searching for an optimal action path using a genetic algorithm by a computer, the step of generating a plurality of solution candidates, the step of evaluating the suitability of each of the solution candidates, Selecting at least one solution candidate based on fitness evaluation, and creating a next-generation solution candidate by performing a genetic operation on the selected solution candidate. One non-terminal symbol is provided with a behavior route search method in which a parameter for determining movement is assigned.

  The present invention is an action path search program for causing a computer to execute a process for searching for an optimal action path using a genetic algorithm, a process for generating a plurality of solution candidates, A process of performing a fitness evaluation, a process of selecting the solution candidate based on the fitness evaluation, and a process of creating a next-generation solution candidate by performing a genetic operation on the selected solution candidate, An action route search program is provided in which a parameter for determining motion is assigned to at least one non-terminal symbol constituting the solution candidate.

  Moreover, the action route search apparatus, method, and program of the present invention can be applied to various technical fields. As an example, for example, search for an optimum route in car navigation, search for an optimum route such as an aircraft, helicopter, ship, and submarine can be cited.

  According to the present invention, it is possible to obtain an optimal solution in real time and to achieve optimal solution derivation that is not affected by teacher data.

Hereinafter, an embodiment of an action route search apparatus and method and a program according to the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of the action route search apparatus according to the present embodiment.
As shown in FIG. 1, an action route search apparatus 10 according to the present embodiment is a computer system (computer system), which includes a CPU (Central Processing Unit) 1, a RAM (Random Access).
Main memory device 2 such as Memory, ROM (Read Only Memory), HDD (Hard Disk)
Drive) and the like, an input device 4 such as a keyboard and a mouse, and an output device 5 such as a monitor and a printer.
Various programs are stored in the auxiliary storage device 3, and the CPU 1 reads out the programs from the auxiliary storage device 3 to the main storage device 2 such as a RAM and executes them to implement various processes.

FIG. 2 is a functional block diagram showing the functions provided in the action route search apparatus 10 in an expanded manner. As illustrated in FIG. 2, the behavior route search apparatus 10 includes a solution candidate generation unit 11 that randomly generates a plurality of solution candidates to create a population, an evaluation unit 12 that performs solution candidate suitability evaluation, and an evaluation. A selection unit 13 that selects a solution candidate based on the fitness evaluation by the unit 12, and a genetic that executes an algorithm such as a genetic operation such as duplication, crossover, mutation, inversion on the solution candidate selected by the selection unit 13 An operation unit 14 and an optimal solution determination unit 15 that determines and outputs an optimal solution are provided.
The behavior route search apparatus 10 searches for a solution candidate having a high degree of fitness by repeatedly performing evaluation, selection, and genetic operation of the solution candidate.

Next, the processing content performed in each part with which the action route search apparatus 10 mentioned above is provided is demonstrated with reference to FIG. 2 are realized by the CPU 1 reading out the action route search program stored in the auxiliary storage device 3 to the main storage device 2 and executing it. is there.
In the present embodiment, a case where an optimum route for an aircraft is searched will be described as an example.

  First, when an initial position, a target position, a relay point to be routed, speed, wind speed, wind direction, and the like are given as input data, the solution candidate generator 11 randomly selects a plurality of solution candidates based on these input data. Generate (step SA1 in FIG. 3). Here, an example of the solution candidate generated by the solution candidate generating unit 11 is shown in FIG. As shown in FIG. 4, the solution candidate is configured by a plurality of nodes represented by non-terminal symbols and terminal symbols, and at least one of the non-terminal symbols is assigned a parameter that determines motion. The parameter that determines the movement is information that determines what action the aircraft will take in the object to search for the action route, that is, in this embodiment. For example, as shown in FIG.

At least one child node is connected to the non-terminal symbol. These child nodes are arbitrarily assigned numerical values corresponding to parameters that determine the motion assigned to the non-terminal symbol. For example, child nodes 21a and 21b indicating what kind of shift is performed are connected to the non-terminal symbol 20a to which shift is assigned, for example, a child node to which a numerical value such as +1 kt / sec and 5 minutes is assigned. This indicates that the gear is shifted for 5 minutes at 1 knot per second. The other non-terminal symbols 20b to which the shift is assigned are connected to child nodes 22a and 22b to which a numerical value such as 0 kt / sec and 10 minutes is assigned, for example. This indicates moving without changing the speed.
In addition, child nodes 23a and 23b indicating what kind of change is performed, for example, -1 deg / sec, 7 minutes, etc. are assigned to the non-terminal symbol 20c to which the change is assigned. . This indicates that the course is changed for 7 minutes at -1 degree per second.

When a plurality of solution candidates having the above-described configuration are generated, these solution candidates are given to the evaluation unit 12 (see FIG. 2). The evaluation unit 12 performs the fitness evaluation of each solution candidate (step SA2 in FIG. 3). For example, the evaluation unit 12 performs the fitness evaluation of each solution candidate using an evaluation function using flight time, fuel consumption, and the like as evaluation parameters.
The selection unit 13 selects and duplicates solution candidates based on the fitness evaluation obtained by the evaluation unit 12 (step SA3). The genetic operation unit 14 performs genetic operations such as crossover, mutation, and inversion on the solution candidates selected and duplicated by the selection unit 13 (step SA4). Then, the fitness evaluation, solution candidate selection / duplication, and genetic operation are repeatedly performed in order to perform generation change of the solution candidates, deceptive action paths are deceived, and a more optimal path is selected. Left behind. Note that a known technique can be employed for such a solution candidate search method, and the method is not particularly limited.

  Next, when the above-mentioned solution candidate generation change is performed a predetermined number of times and the end condition is satisfied (“YES” in step SA5), the evaluation unit 12 determines the optimal solution from the evaluation unit 12 and the suitability evaluation value thereof. Is output to the unit 15. Here, the end condition is, for example, that the generation change has reached a predetermined generation set in advance, or the suitability evaluation value of any solution candidate has reached a preset final fit evaluation value Or the elapsed time from the start of processing has reached a predetermined time.

  The optimal solution determination unit 15 determines a solution candidate having the highest fitness evaluation value from the solution candidates acquired from the evaluation unit 12 as an optimal solution, outputs this optimal solution as output data (step SA6), and performs the processing. finish.

As described above, according to the action route search apparatus and method and the program according to the present embodiment, since a parameter for determining motion is assigned to at least one non-terminal symbol constituting the solution candidate, The action route can be directly derived as a solution candidate, and the finally obtained solution candidate can be output as output data as it is.
Thus, by devising the configuration of the solution candidates, it is possible to omit the calculation process of the optimal algorithm that was performed prior to operation as in the past, and to obtain the action route in real time. . Furthermore, since teacher data or the like is not necessary, it is possible to obtain an appropriate action route for any input data.

In the above-described embodiment, the case of searching for the optimum route of an aircraft has been described as an example. However, it goes without saying that the action route search apparatus, method and program of the present invention can be applied in various fields. Nor. In this case, input information, an evaluation function, etc. are given according to each application. For example, when searching for the optimal route of a vehicle such as a car navigation system, the current location, the destination, etc. are given as input data, and an evaluation function using the arrival time, fuel consumption, etc. as evaluation parameters is given. In the case of vehicles, unlike airplanes and ships, there is a restriction that the vehicle cannot travel where there is no road, so the parameters that determine the movement are unique to the vehicle such as turning right, turning left, and going straight at intersections. Parameters will be used. Thus, the parameters that determine the movement are determined according to the object.
In the solution candidate configuration illustrated in FIG. 4, a non-terminal symbol to which a program for setting an initial position is assigned may be further added. By adding the program for setting the initial position as described above, the program can be applied even when the initial position is unknown.

  As mentioned above, although embodiment of this invention was explained in full detail with reference to drawings, the specific structure is not restricted to this embodiment, The design change etc. of the range which does not deviate from the summary of this invention are included.

It is the block diagram which showed schematic structure of the action route search apparatus which concerns on one Embodiment of this invention. It is the functional block diagram which expand | deployed and showed the function with which the action route search apparatus which concerns on one Embodiment of this invention is provided. It is the flowchart which showed the procedure of the process performed by the action route search apparatus which concerns on one Embodiment of this invention. It is the figure which showed one structural example of the solution candidate. It is a figure for demonstrating crossing. It is a figure for demonstrating a mutation.

Explanation of symbols

1 CPU
2 Main storage device 3 Auxiliary storage device 4 Input device 5 Output device 10 Action path search device 11 Solution candidate generation unit 12 Evaluation unit 13 Selection unit 14 Genetic operation unit 15 Optimal solution determination unit

Claims (5)

An action path search device that includes a computer and searches for an optimal action path using a genetic algorithm,
The calculator is
A solution candidate generator for generating a plurality of solution candidates;
An evaluation unit that performs conformity evaluation of each of the solution candidates;
A selection unit that selects the solution candidates based on the fitness evaluation;
A genetic operation unit that creates a next generation solution candidate by performing a genetic operation on the solution candidate selected by the selection unit, and
A behavior route search device in which a parameter for determining motion is assigned to at least one non-terminal symbol constituting the solution candidate.   The action route search device according to claim 1, wherein the parameter that determines the movement is a needle change or a shift.   The action route search device according to claim 1 or 2, wherein a program for setting an initial position is assigned to at least one non-terminal symbol constituting the solution candidate. An action route search method for searching for an optimum action route using a genetic algorithm by a computer,
Generating a plurality of solution candidates;
Performing a fitness evaluation of each of the solution candidates;
Selecting the solution candidates based on the fitness evaluation;
Creating a next generation solution candidate by performing a genetic operation on the selected solution candidate,
A behavior route search method in which a parameter for determining motion is assigned to at least one non-terminal symbol constituting the solution candidate. An action path search program for causing a computer to execute processing for searching for an optimal action path using a genetic algorithm,
Processing to generate multiple solution candidates;
Processing for evaluating the suitability of each of the solution candidates;
Processing for selecting the solution candidates based on the fitness evaluation;
Generating a next generation solution candidate by performing a genetic operation on the selected solution candidate,
A behavior route search program in which a parameter for determining motion is assigned to at least one non-terminal symbol constituting the solution candidate.

Images