JP3465725B2 - Mobile robot control device and method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mobile robot control apparatus and method suitable for use in controlling a mobile robot that carries articles in, for example, a factory.

[0002]

2. Description of the Related Art Mobile robots are sometimes used to transport articles in factories. This mobile robot transports goods (luggage) existing on the factory premises to a predetermined place (destination). In addition, it is preferable that the mobile robot should replenish its energy by itself, without human intervention.

When such a work is performed by a single mobile robot, the mobile robot can reach the destination by providing the mobile robot with a device for recognizing the destination. As a result, carry the luggage, unload the luggage at the destination, repeat the operation to leave the destination to carry other luggage, when energy becomes insufficient, move to the energy supply base and supply energy there After that, it is possible to relatively easily realize a mobile robot that repeats the load carrying operation again.

However, when there are many mobile robots that perform such work, a so-called deadlock may occur around the destination (or the energy supply base). For example, as shown in FIG. 22, when a deadlock occurs, a large number of mobile robots 1 gather around the destination 11, and the mobile robot 1 near the destination 11 unloads the baggage to the destination 11. Even if the user tries to move away from the destination 11, he or she cannot move away from the destination 11 because there are other mobile robots 1 around. Further, the mobile robot 1 located at a place where another mobile robot 1 exists between the destination 11 and the destination 11 cannot reach the destination 11 because the inner mobile robot 1 interferes.

The problem of this deadlock is, in essence,
This is a problem of exclusive control when a limited resource (destination 11) is shared by many users (mobile robot 1). In the conventional method, an introduction route and an exit route are provided around the destination 11 to restrict the route approaching the destination 11 or the route leaving the destination 11.

However, when the route is restricted in this way, a device for installing such a route around the destination 11 or recognizing the route is newly required. As a result, not only the extra cost is required, but also the facility and layout in the factory are largely restricted.

Therefore, the present applicant has filed Japanese Patent Application No. 6-14519.
As No. 9, it was previously proposed that the force acting on the mobile robot be a combined force obtained by combining the first to fourth forces, and the moving direction of the mobile robot is determined from this combined force. .

In the above proposal, the coefficient A _P for calculating the second to fourth forces among the first to fourth forces is calculated.
Two parameters α (= A _a defined by A _a and A _V
The average performance of the group of mobile robots is determined depending on / A _P ) and β (= A _V / A _P ). By careful selection of these parameters, a group of mobile robots with good average performance can be constructed.

[0009]

However, in the above proposal, it has been assumed that the mobile robot groups each have the same parameter value. However, in practice, it is possible that each mobile robot will have parameters of different values. In addition, it may not always be possible to check optimal parameters in advance.

In such a case, it is necessary to determine the values of the two parameters by some method.

In determining the values of these two parameters, it is assumed that, in the worst case, random two parameter values are assigned to each mobile robot, and for example, proposed by Holland, It is conceivable to use a genetic algorithm (Genetic Algorithm). This genetic algorithm is introduced, for example, on pages 2 to 6 of JSSST Tutorial "Basics of Artificial Life".

In this genetic algorithm, when a problem is given, the problem is encoded as a one-dimensional string by some means. Those encoded in this way are called chromosomes. Then, for the population of chromosomes, a fitness evaluation function (fitness function)
Ion) is applied, good ones are selected from them, they are crossed and propagated, and bad ones are deleted. In this way, the entire robot moving group searches for a good solution.

That is, each mobile robot is provided with this algorithm to search for a good parameter value.

However, this genetic algorithm is
A goodness-of-fit evaluation function is needed to evaluate the problem correctly, and is not suitable for problems whose nature is unclear. For example, it is possible to evaluate such that the combination of parameter values in the above proposal is better, but it is not always possible to perform an experiment to determine parameters in a dynamic environment in which many autonomous robots move around. Not that.

The present invention has been made in view of such a situation, and it is simple, reliable, low-cost, and has a relatively good average parameter combination (the two parameters that all mobile robots have are not necessarily the same). (It does not match the parameters of other mobile robots, but it is a combination of them considering the average value of each parameter)
Is to be able to obtain.

[0016]

Mobile robot controller of the present invention According to an aspect of the other of the mobile robot receives the parameters defined by the coefficients for calculating the force, a plurality of received
Select a new parameter from among the parameters of, select one of the selected parameter and the current parameter based on the probability, and select a part of the selected parameter.
Conversion as a new parameter, based on the new parameters
The moving direction of the mobile robot is determined based on the above.

From the process of receiving a parameter from another mobile robot to the process of changing the selected parameter to a new parameter , the amount of energy of the mobile robot is predicted.
If it is less than the standard value set for
The process of determining the moving direction of the mobile robot based on the new parameter can be performed when the amount of energy is larger than the reference value .

The other mobile robot that receives the parameter can be an adjacent mobile robot.

Parameters of a plurality of other mobile robots adjacent to each other
Parameters from the last time
The degree of conformity defined by the time until receiving
Select parameters for other mobile robots or
The parameters of a plurality of other mobile robot, the parameters
Depending on the time from when you received the last time until this time
It is possible to select with a probability corresponding to the defined goodness of fit value.

Of the number of times the movement direction is determined between LTIME and LTIME from the time the parameter is received last time to the time this time is received, the number of times when there is another mobile robot in the determined movement direction of the mobile robot is calculated. #S
When TOP is used, the goodness of fit fitness is fitness = (LTIME + 1) / (# STOP +
1) can be represented.

[0021] When the probability of selecting parameters of the other mobile robot is selected to R%, the probability of selecting the current parameters can be (100-R)%.

It is possible to generate a new parameter from a binary number obtained by inverting any bit of the parameter with a probability of K% for conversion and expressing the parameter with a binary number.

There are two parameters, each of which is converted into an n-bit binary integer and two n
The bit parameters are grouped into one to make a 2n-bit binary number, and any bit of the 2n bits is inverted with a probability of K%, and the inverted 2n-bit binary number is separated for every n bits. , The separated n-bit binary number can be converted into a floating point number to generate two new parameters.

The force acting on the mobile robot is generated by the first force received from the surrounding mechanical field, the second force received from another mobile robot located in the moving direction, and the movement of the other mobile robot. The third force received from the space before the movement and the fourth force received by the movement of another adjacent mobile robot are combined forces, and the coefficient when the second force is calculated is A _P , coefficient a _a time for calculating a third force, when the coefficient when calculating the fourth force and a _V, the alpha further each of the two parameters, the beta, alpha, beta are, alpha = a It may be expressed as _a / A _P β = A _V / A _P.

The mobile robot control method of the present invention receives a parameter defined by a coefficient for calculating a force from another mobile robot, selects a new parameter from the plurality of received parameters, and selects it. has been selected based on the parameter and the probability of one of the current parameters, and characterized in that the new parameters to convert a portion of the selected parameter to determine the moving direction of the mobile robot based on the new parameters To do.

[0026]

In the mobile robot control apparatus and method having the above structure, a plurality of parameters received from other mobile robots are used.
A new parameter is selected from the meters , and either the selected parameter or the current parameter is selected.
Selected based on probability . Then , some of the selected parameters are converted into new parameters, and the moving direction of the mobile robot is determined based on the parameters. Therefore, it becomes possible to control the mobile robot easily and reliably at low cost.

[0027]

[0028]

1 shows an example of the configuration of a mobile robot 1 of the present invention. In this embodiment, it is assumed that the mobile robot 1 has wheels 2 and 3 and moves in any one of eight directions, that is, a direction 1 to a direction 8 diagonally in the vertical and horizontal directions shown in FIG. That is, the movements of the mobile robot 1 are forward movement, rotation, and stop, and the rotation is ± 45 ° left and right, ±
It is either 90 °, ± 135 °, or 180 ° to the right. In other words, for convenience, the mobile robot 1 is shown in FIG.
To any position in the matrix, as shown in
The frames (blocks) are to be sequentially moved.

The mobile robot 1 has a robot sensor 4-1 for detecting another mobile robot in the direction 1. With this robot sensor 4-1, direction 1
It is possible to detect the direction in which another mobile robot located above is facing and the distance to the mobile robot.

Further, a potential sensor 5-1 is provided corresponding to the robot sensor 4-1 in the direction 1 so as to detect the mechanical potential. The potential sensor 5-1 is different depending on the type of dynamic field generated by the destination 11, but can be configured by, for example, a microphone, a gas sensor, an optical sensor, a temperature sensor, a VHF antenna, a UHF antenna, or the like.

In addition to this, the mobile robot 1 is provided with a robot sensor 4-2 and a potential sensor 5-2 in the direction 2, and also in the directions 3 to 8 as the robot sensors 4-3 to 4-8. , Potential sensors 5-3 to 5-8. As a result, the direction in which the other mobile robot is facing, the distance to that direction, and the mechanical potential are detected for each direction.

Further, the destination 11 is the dynamic field generator 12
And provides a dynamic field corresponding to the distance r from the destination 11 by sound, gas, light, temperature, radio waves, or the like.
This includes speakers, gas generators, lamps, heat sources,
It can be configured by a radio wave generator or the like.

In this embodiment, each mobile robot 1
A force vector F _i (e) that is received in each direction i is defined by the following equation.

[Equation 1]

Here, f _1i represents the force that the mobile robot 1 receives from the destination 11, and is a function of the dynamic field vector and the internal state of the mobile robot 1 (for example, the potential e corresponding to the remaining amount of energy). Determined by H, E, W.

As shown in FIG. 3, the function H is 1 when the potential e is smaller than the predetermined reference value S, and 0 when it is larger.
Is a function. Further, as shown in FIG. 4, the function E is a function that becomes 0 when the potential e is smaller than a predetermined reference value F (F> S) and 1 when it is large. Further, as shown in FIG. 5, the function W is a function that is 0 when the potential e is smaller than a predetermined reference value S, 1 when it is a value between the reference values S and F, and 0 when it is larger than F. .

The potential of this dynamical field depends on the destination 11
It gets bigger as it gets closer to. When the dynamic field generator 12 is composed of a light source, that is, when the dynamic field is given by light, its potential is inversely proportional to the square of the distance,
When the dynamic field generator 12 is composed of a radio wave generator, that is, when the dynamic field is given by a radio wave, the potential is inversely proportional to the cube of the distance.

This mobile robot 1 has a potential e
When the (energy) is smaller than the reference value S, a larger force is received in the direction of the destination 11, and when the energy is larger than the reference value F, a larger force is received in the direction away from the destination 11. The amount of energy is the reference value S and the reference value F.
When in between, the mobile robot 1 will receive equal force in all directions.

Φ _i in the above equation (1) represents the case where the potential is inversely proportional to the square of the distance.
₀ is the destination (in this example, the energy supply base)
11 represents the energy intensity of the dynamic field in 11.

When the force F _i (e) is constituted only by this force f _1i , if the mobile robot 1 is controlled so as to move in any one of the directions 1 to 8, the mobile robot 1 will It is possible to approach or leave the ground 11. However, in this case, the above-mentioned problem of deadlock cannot be solved. Therefore,
In this embodiment, the forces f _{2i to} f _4i are further combined.

Force f_2iIs the direction in which the mobile robot 1 moves
Force (pressure) received from another mobile robot 1 adjacent to i
Is represented. A in this section_PIs a predetermined constant
R, Q_iHas another mobile robot adjacent to direction i
It is a function that is set to 1 when 0 and 0 when not present. That is, in FIG.
As shown, the mobile robot 1 is adjacent to another direction 1 in the other direction.
If there is a mobile robot 1u, Q_iIs 1. direction
When there is no mobile robot 1 adjacent to 2 to 8
In that case, the value Q in those directions₂To Q₈ Is
Each becomes 0.

Further, d _iV is a vector of size 1 indicating the direction in which another mobile robot 1 adjacent to the direction i of the mobile robot 1 is facing. Further, n _iV is a normal vector in the direction i of the mobile robot 1 in the direction toward the mobile robot 1. d _iV · n _iV means the inner product of d _iV and n _iV . The force f _2i represents a component in the direction perpendicular to the mobile robot 1 in the direction in which each of the other mobile robots 1 adjacent to the directions 1 to 8 of the mobile robot 1 faces. If another mobile robot 1 adjacent to the mobile robot 1 is completely oriented in the vertical direction, the vector component for that direction is 1.

The coefficient A _V that gives the force f _3i is a predetermined constant. This force f _3i represents the force (viscous force) received when the adjacent mobile robot 1 moves. This power is
It is computed from the sum of the parallel component with respect to the normal vector n _iV other direction vector of the moving robot 1 adjacent in a direction having a vertical normal vector to the normal vector n _iV. That is, as shown in FIG. 7, for example, this force f _{3i tends} to move the mobile robot 1 in the same direction 1 when the adjacent mobile robot 1L moves in the upper direction (direction 1) in the figure. Means power.

[0043] In addition, the constant A _a for setting the force f _4i is a predetermined constant, Q _'i represents a one step previous Q _i, ¬ represents a NOT. ∧ represents AND, and Q ′ _i ∧¬Q _i means that the logical product of Q ′ _i and the inverted logic of Q _i is calculated.

That is, this force f _4i is a force (gravitational force) that is considered when the mobile robot 1 cannot move at the previous time.
When another mobile robot adjacent to the mobile robot 1 moves to a position that is not adjacent to the mobile robot 1 at the next time, it receives the force of A _{a in} the direction of the position before moving.

That is, as shown in FIG. 8, the mobile robot 1
_{When M} moves upward in the figure, an attractive force that attempts to move the mobile robots 1 _{A to} 1 _G adjacent to this space S _M in the direction of the space (the space before moving) S _M generated thereby. Is represented.

The mobile robot 1 is constructed, for example, as shown in FIG. 9 so that the moving direction can be determined in accordance with the above equation (1).

In this embodiment, the robot sensor 4
The outputs of (4-1 to 4-8) are supplied to the arithmetic circuit 21 via the gate 31, and the arithmetic circuit 21 determines that another mobile robot is within a certain distance (that is, an adjacent position) in each direction i. Information indicating whether or not the mobile robot is present and information indicating the direction in which the mobile robot is facing are separated when another mobile robot is adjacent to the mobile robot. Information about the presence or absence of the adjacent mobile robot is stored in the register array 22, and information indicating the direction in which the mobile robot is facing is stored in the register array 23.
To each output. Since these outputs exist for each of the directions 1 to 8, the register array 22 and the register array 23 are each composed of eight registers.

The output of the register array 22 is the gate 24.
Is supplied to a register array 25 composed of eight registers. That is, the register array 25 holds the original value stored in the register array 22. The output of the register array 22 is the function Q in the equation (1).
_i, and the output of register array 25 provides the function Q ′ _i . The output of the register array 23 gives d _i .

These register arrays 22, 23, 25
Is supplied to the force vector calculation circuit 26.

The force vector calculation circuit 26 is also supplied with the intensity φ _i given by the mechanical field generator 12 in the directions 1 to 8 which is detected and output by the potential sensor 5 (5-1 to 5-8). Has been done.

The energy check circuit 30 compares the energy amount of the mobile robot 1 with a preset reference value, and when the energy amount is larger than the reference value, the gate 3
2 is closed and the gate 31 is opened to supply the output of the robot sensor 4 to the arithmetic circuit 21. On the other hand, when the amount of energy is less than the reference value, the gate 31 is closed and the gate 32 is opened.

The communication device 34 uses the previous parameter α (= A _a / A _P ) held in the register 39, the previous parameter β (= A _V / A _P ) held in the register 40, and the degree of conformity. The conformity calculated by the calculation device 41 (calculated by the formula (2) described later) is received through the gate 42, and these values are transmitted to other mobile robots by radio waves, infrared signals, voice signals, or the like. Send. Also,
The parameters α and β and the goodness of fit output from another mobile robot are received, and the parameter α is input to the register array 35 and the parameter β is input to the register array 36 via the gate 32.
, And the fitness to the register array 37, respectively.

The gate 33 is the energy check circuit 3
Is controlled from 2, when the gate 32 is opened, it is by <br/> Uninasa opened simultaneously. When the gate 33 is opened, the data held in the register array 22 is transferred to the communication device 34.
Is output to the gate 42. The gate 42 outputs its parameters α and β and the fitness to the communication device 34, and when the adjacent mobile robot exists, the communication device 34 transfers the data input from the gate 42 to the mobile robot in the corresponding direction. Send.

The parameter selection device 38 selects a parameter of the mobile robot in a predetermined direction from the conformity input from the register array 37 by the first method or the second method described later.

In the first method, among the adjacent mobile robots, the one having the maximum fitness is selected.

Here, fitness is expressed by the following equation. fitness = (LTIME + 1) / (# STOP + 1) (2) Here, the LTIME is from the time when the adjacent mobile robot receives the previous supply of the parameters α and β and the fitness until the time when a new input is received again. Represents time, and also #S
TOP represents the number of collisions that occurred during LTIME. The number of collisions is the number of times the moving direction is determined.
The number of times when another mobile robot is present adjacent to the determined moving direction is shown.

In the second method, the robot i is selected with the probability P _i according to the value of the fitness of the i-th adjacent mobile robot.

The probability Pi in this case is expressed by the following equation.

[Equation 2]

The parameter selection device 38 further selects, with a predetermined probability, one of α and β of the selected mobile robot and one of the current α and β of itself, as shown in a flowchart of FIG. 11 described later. , Predetermined conversion processing is performed to obtain new α and β. Therefore, the parameter selection device 38
Is supplied with the output of the random number generation circuit 27. Then, the new parameter α is held in the register 39, and the new parameter β is held in the register 40. The parameters α and β held in the registers 39 and 40 are supplied to the gate 42 and the force vector calculation device 26, respectively.

[0060] force vector calculation circuit 26, the output d _i of the register array 23, output Q _i of the register array 22, output Q _'i of the register array 25, the output phi _i of the potential sensor 5, and the output of the register 39, 40 From α and β,
The force F _i (e) is calculated according to the equation (1) described above.
Then, the calculation result is output to the movement direction determination circuit 28.

The moving direction determining circuit 28 is supplied with the random number output from the random number generating circuit 27. The moving direction determining circuit 28 determines the moving direction from the force F _i (e) supplied from the force vector calculating circuit 26 and the random number output from the random number generating circuit 27, and outputs a signal corresponding to the moving direction to the drive circuit 29. It is designed to output to. Drive circuit 29
Is the output of the moving direction determination circuit 28 and the register array 2
The wheels 2 and 3 are finally driven from the output Q _{i of} 2. The gate 24 is controlled by the output of the moving direction determining circuit 28.

Next, the operation of the embodiment of FIG. 9 will be described with reference to the flowcharts of FIGS. First, in step S1, the energy check circuit 30
Determines whether the amount of energy is equal to or greater than a preset reference value. When the amount of energy does not reach the reference value, the process proceeds to step S6, and the robot sensor 4 determines whether or not there is an adjacent mobile robot.

If it is determined in step S6 that there is no adjacent mobile robot, step S6
Go to 8. In step S8, a process of receiving the fitness and the parameters α and β from all the adjacent mobile robots is executed.

That is, when the energy does not reach the reference value, the energy check circuit 30 opens the gate 33 and the gate 32. As a result, the information on the presence / absence of the adjacent robot held in the register array 22 is input to the communication device 34 via the gate 33. From the communication device 34 to all mobile robot adjacent in response to this information, conform degree fitness and the parameter alpha, beta
To be sent.

The parameters α and β of the adjacent mobile robots and the fitness degree fitnes received by the communication device 34.
s is a register array 35 to 37 through the gate 32.
Are respectively supplied to and retained.

Next, in step S9, the mobile robot selection process is executed by the above-described first method or second method. That is, the parameter selection device 38 uses the fitness degree fitnes held in the register array 37.
The one having the largest value among s is selected (first method). Alternatively, the mobile robot i is selected with the probability P _i according to the value of the fitness fitness (second method).

Then, the process proceeds to step S10 and step S9.
The parameters α and β of the mobile robot selected in step 1 have an R% probability, and their own parameters α and β are (100−
R)% is selected as the new parameters α and β of oneself.

For example, as shown in FIG. 12, [0%, 10
0%] ([0,1]), R% and (100-R
%) Section, 0 to 1 (0% to 100%)
The values between are generated by random numbers, and the parameters α and β in the interval corresponding to the random numbers are selected.

Further proceeding to step S11, the parameter selecting device 38 executes the conversion process of the parameters α and β selected in step S10. Details of this conversion process are shown in FIG.

That is, first, in step S21,
The process of encoding α and β into n-bit binary integers is executed. For example, when the parameters α and β are respectively 8-bit integers, α and β are multiplied by (2 ⁸ −1). Then, the binary integer represented by these two n bits is coded as one 2n-bit binary number as shown in FIG.

Next, in step S22, the value of any bit of the 2n-bit binary number is inverted with a probability of K%. That is, the logic 1 is inverted to 0 and the logic 0 is inverted to 1.

Next, the process proceeds to step S23, and step S2
The 2n-bit binary number obtained by inverting by 2 is again separated into n-bit binary numbers, each of which is decoded (for example, divided by (2 ⁸ -1)), The parameters α and β are represented.

Next, in step S12 of FIG. 11, the parameters α and β obtained by the conversion in step S11 are used as own parameters. That is, the parameter selection device 38
Is a register 39, 40 for the parameters α and β obtained by the conversion in step S11 (steps S21 to S23).
Set to. Α set in these registers 39 and 40,
β is supplied to the communication device 34 via the gate 42, transmitted to other mobile robots as its own parameters α and β, and also supplied to the force vector calculation device 26.

As described above, when there are adjacent mobile robots, new self parameters α and β are set by the processes of steps S8 to S12, and then the process proceeds to step S7 to receive energy supply. The process is executed. On the other hand, if it is determined in step S6 that there is no adjacent mobile robot, the fitness degree fitness and the parameters α and β cannot be supplied, so that the processes in steps S8 to S12 are not performed. If not, the process immediately proceeds to step S7, and the energy supply process is executed.

After the energy supply process is executed in step S7, the process proceeds to step S2.

When the energy supply is received in step S7, the energy reaches the reference value or more, and therefore, as in the case where it is determined in step S1 that the energy is above the reference value, step S2
Then, the process of step S2 is executed.

In step S2, the above (1)
The force F received by the mobile robot 1 is calculated according to the formula.

That is, the robot sensor 4-i detects another mobile robot 1 in the corresponding direction i. The arithmetic circuit 21 detects each direction i from the output of the robot sensor 4-i.
First, it is determined whether or not the other mobile robots 1 are adjacent to each other (whether or not they are located within a predetermined distance range), and the determination result is output to the register array 22. When another mobile robot 1 adjacent to the mobile robot 1 exists, the direction in which the mobile robot 1 is facing is determined and data indicating the direction is output to the register array 23.

The data held in the register array 22 is 1 if there is a mobile robot adjacent to the direction i, and 0 if it does not exist. This is (1) above
Corresponds to Q _i in the equation. In addition, the register array 23
The data held by is a data having a size of 1 and representing the direction in which the adjacent mobile robot 1 is facing, and is d _i in the above equation (1).

The data held in the register array 22 is transferred to and held in the register array 25 subsequent to the register array 22 via the gate 24 in response to the control signal from the movement direction determining circuit 28. It is a thing. That is, in the register array 25, the Q one step before
_i , that is, Q ′ _i in the above equation (1) is held.

On the other hand, the potential sensor 5-i detects the distance r to the destination 11 and divides a predetermined constant φ ₀ by the distance r ² _i to obtain a value φ _i (= φ ₀ / r ² _i ) is calculated and output to the force vector calculation circuit 26.

The force vector calculation circuit 26 uses the outputs of the register arrays 22, 23, 25, the potential sensor 5, and the registers 39, 40 according to the above equation (1).
The force F _i (e) is calculated for each direction i. Then, the calculation result is output to the movement direction determination circuit 28.

In the case of this embodiment, the coefficient A _P is fixed to a predetermined value. Therefore, α (= A _a / A _P ), β (=
By multiplying A _V / A _P ) by A _P , A _a and A _V can be obtained.

Next, in step S3, the moving direction of the section to which the generated uniform random number belongs among the partial sections obtained by dividing the [0,1] section for each moving direction i in proportion to the force. select.

That is, for example, as shown in FIG.
The absolute values of the forces F _{1 to} F _{8 in} the direction 8 are obtained, and the sections from 0 to 1 are assigned in descending order according to the absolute values. For example, in the embodiment shown in FIG. 15, the absolute value of the force F ₃ is the largest, and in the following, F ₂ , F ₁ , F ₆ , F ₅ , F ₈ , F ₄ ,
Since the order is F ₇ , the section from 0 to 1 is assigned in this order corresponding to the magnitude of the absolute value.

When the random number generating circuit 27 generates a value between 0 and 1 by a random number, the generated random number is used as the moving direction determining circuit 2
8 are supplied. As shown in FIG. 15, the moving direction determination circuit 28 selects the force F _i in the section corresponding to the random number supplied from the random number generation circuit 27 in the section from 0 to 1, and selects the selected force F _i. Is output to the drive circuit 29.

The drive circuit 29 also includes the register array 2
2 outputs Q _i are provided. As described above, this Q _i is set to 1 when another mobile robot 1 is present adjacent to the direction i, and is set to 0 when it is not present. Therefore, in step S4, another mobile robot 1 is adjacent to the selected moving direction i from this Q _i.
Is present.

In the moving direction decision circuit 28, when another moving robot is adjacent to the decided direction i, the moving robot 1 which is going to decide the moving direction may move in the direction i. Can not. Therefore, in this case, the process returns to step S2, the above-described operation is repeated, and the process of selecting a new direction is executed.

On the other hand, when Q _i is 0 (when there is no other mobile robot 1 adjacent to the direction i), the process proceeds to step S5, and the direction of the mobile robot 1 is changed to the selected movement direction. Change and move forward in that direction by one step (one block). That is, the drive circuit 29 drives the wheels 2 and 3 to rotate the mobile robot 1 in the direction i,
Move one step in direction i.

The above operation is independently performed for each mobile robot 1, and each mobile robot 1 moves to the destination 11.

Next, the experimental results will be described. 100 mobile robots 1 having the configuration shown in FIG. 9 were prepared, and the processing shown in FIGS. 10 and 11 was repeated 300 times for each of the 100 mobile robots 1. Further, as the position of the dynamic field generator 12, four types were prepared. And K
Experiments were performed four times by changing the combination of the% and R% values and the selection method of the mobile robot. Each experiment number is represented by numbers 1 to 4.

The evaluation in each experiment was evaluated by using the evaluation value represented by the following formula (this evaluation value is an index showing how fairly and many shared resources are reached). Evaluation value = average number of energy replenishments / standard deviation The average number of energy replenishments here represents the number of times that each mobile robot reached a shared resource while avoiding other mobile robots, and the standard deviation was the energy replenishment of all mobile robots. Calculated from the number of times.

An experiment in the case where all the mobile robots have the values of α = 1.0 and β = 0.0, which are considered to be a good combination in the applicant's previous proposal, is called BASIC, and The value when the evaluation value (standard evaluation value) for the result is normalized as 1 (that is, evaluation value /
The evaluation was performed by the standard evaluation value).

[0094]

[Table 1]

Table 1 shows R which determines which of the pair α and β of the selected mobile robot or the pair α and β of itself is selected in step S10 of FIG.
The normalized evaluation value is shown when% is 0%. Figure 1
If the value of K% for inverting the bit in step S22 of 2 is 0.0%, first, α of each mobile robot,
When β is randomly generated according to the normal distribution as an initial state, the experiment is performed without changing the state at all. The normalized evaluation value in this case is a low value. Hereinafter, the normalized evaluation value when the value of K is increased to 0.2% and further to 20% is shown.

Further, in Table 1, for comparison, R = 1
The normalized evaluation value when 00% and K = 0% is shown. In this case, step S9 of FIG.
The maximum adaptability (first method) is adopted as the method of selecting the mobile robot in (1). In this case, it can be seen that the performance comparable to the standard evaluation value is obtained.

FIG. 16 shows the time evolution of the coefficients α and β of Experiment No. 4 among the experiments in which this maximum performance was obtained. As shown in the figure, the value of α is large, and β
It can be seen that the tendency of the value of becomes smaller quickly converges.

That is, by setting the condition that R = 100% and K = 0% and the mobile robot having the maximum fitness is selected, the situation in which each mobile robot 1 has completely random parameters α and β. , It is found that a good pair of parameters α and β can be found quickly.

In FIG. 17, R = 100% and K% is 0.0
%, 0.2%, or 20%, and as a mobile robot selection method in step S9 of FIG.
The normalized evaluation value when the first method or the second method is adopted is shown. Also in FIG. 18, similarly,
The normalized evaluation value when R = 90% is shown. In any case, when K = 0.0% in the first method,
Good evaluation values are obtained.

FIG. 19 shows R of Experiment No. 1 among them.
= 90% and K = 0.2%, the parameters α and β are shown over time. Similarly, FIG. 20 shows the time evolution of the parameters α and β when R = 90% and K = 0.0% of the experiment number 2. Further, FIG.
Indicates the time evolution of the parameters α and β in the case of R = 100.0% and K = 0.2% of Experiment No. 4.

As apparent from comparison of FIGS. 19 to 21 with FIG. 16, in the experimental example shown in FIG.
It can be seen that the parameters α and β converge to a large value or a small value in an extremely short time.

In the above description, the moving direction of the mobile robot 1 is determined by the random number with the probability corresponding to the absolute value of the compatibility force. However, the moving direction is determined by applying the principle of neural net, fuzzy, etc. It is also possible to decide.

[0103]

As described above, according to the present invention , a new parameter is selected from a plurality of parameters received from other mobile robots, and one of the selected parameter and the current parameter is used as a probability. Since a part of the selected parameter is converted based on the selected parameter and the converted parameter is used as a new parameter, the moving direction can be determined easily, reliably, and at low cost.

[0104]

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a mobile robot of the present invention.

FIG. 2 is a diagram illustrating a moving space of a mobile robot.

FIG. 3 is a diagram illustrating a function H used for calculation performed by a mobile robot.

FIG. 4 is a diagram illustrating a function E used in a calculation performed by a mobile robot.

FIG. 5 is a diagram illustrating a function W used in a calculation performed by a mobile robot.

FIG. 6 is a diagram illustrating a pressure received by a mobile robot.

FIG. 7 is a diagram illustrating a viscous force received by a mobile robot.

FIG. 8 is a diagram illustrating an attractive force received by a mobile robot.

FIG. 9 is a block diagram showing an example of the internal configuration of a mobile robot.

FIG. 10 is a flowchart illustrating processing performed by a mobile robot.

11 is a flowchart following FIG.

FIG. 12 is a diagram illustrating a process of step S10 of FIG.

13 is a flowchart illustrating a more detailed process of step S11 of FIG.

FIG. 14 is a diagram illustrating a process of step S21 of FIG.

15 is a diagram illustrating the process of step S3 of FIG.

FIG. 16 is a diagram showing a temporal expansion of parameters α and β of experiment number 4.

FIG. 17 is a graph showing the experimental results for each experimental number when R = 100%.

FIG. 18 is a graph showing the experimental results for each experimental number when R = 90%.

FIG. 19 is a diagram showing a temporal expansion of parameters α and β of experiment number 1.

FIG. 20 is a diagram showing temporal expansion of parameters α and β of experiment number 2.

FIG. 21 is a diagram showing the temporal expansion of parameters α and β of experiment number 4.

FIG. 22 is a diagram illustrating a deadlock of a mobile robot.

[Explanation of symbols]

1 Mobile robot A few wheels 4, 4-1 to 4-8 Robot sensor 5, 5-1 to 5-8 Potential sensor 11 destinations 12 Mechanical field generator 21 Operation circuit 22,23 register array 24 gates 25 register array 26 force vector calculation circuit 27 Random number generator 28 Moving direction decision circuit 29 Drive circuit 30 energy check circuit 34 Communication device 35 to 37 register array 38 Parameter selection device 39, 40 registers 41 Fitness calculation device

Claims (12)

(57) [Claims] 1. A mobile robot control device for obtaining a force acting on a mobile robot and determining a moving direction of the mobile robot from the obtained force, which is defined by a coefficient for calculating the force from another mobile robot. It receives a parameter, selects one of the received plurality of parameters as a new parameter, selects one of the selected parameter and the current parameter based on a probability, and selects a part of the selected parameter. A mobile robot control device, wherein the mobile robot controller is configured to be converted into a new parameter, and the moving direction of the mobile robot is determined based on the new parameter. 2. The process from the process of receiving the parameter from the other mobile robot to the process of setting the selected parameter as a new parameter is such that the amount of energy of the mobile robot is based on a preset reference value. The movement according to claim 1, wherein the processing that is performed when the amount of energy is small and determines the movement direction of the mobile robot based on the new parameter is performed when the amount of energy is greater than the reference value. Robot controller. 3. The mobile robot controller according to claim 1, wherein the other mobile robot that receives the parameter is an adjacent mobile robot. 4. A parameter of the other mobile robot having a maximum degree of conformity defined by the time from the last reception of the parameter to the current reception, out of a plurality of parameters of the other mobile robots adjacent to each other. The mobile robot controller according to claim 3, wherein the mobile robot controller is selected. 5. The time from the last reception of the parameter to the present reception of the parameter is LTIME, and among the number of movement direction determinations performed during the LTIME, the other mobile robot is determined to move in the determined movement direction of the mobile robot. When the number of times when it exists is #STOP, the conformance fi is
tness is fitness = (LTIME + 1) / (# STOP +
1) The mobile robot controller according to claim 4, wherein 6. A parameter of a plurality of the other mobile robots adjacent to each other is selected with a probability corresponding to a value of a goodness of fit defined by a time from the last reception of the parameter to the current reception. The mobile robot controller according to claim 3. 7. The time from the previous reception of the parameter to the current reception of the parameter is LTIME, and among the number of determinations of the movement direction performed during the LTIME, the other mobile robot is determined to move in the determined movement direction of the mobile robot. When the number of times when it exists is #STOP, the conformance fi is
tness is fitness = (LTIME + 1) / (# STOP +
1) The mobile robot control device according to claim 6, characterized in that 8. The probability of selecting the current parameter is (100-R)% when the probability of selecting the parameter of the other selected mobile robot is R%. The mobile robot controller according to item 1. 9. The parameter is represented by a binary number, an arbitrary bit thereof is inverted with a probability of K% to perform the conversion, and the new parameter is generated from the binary number obtained by the inversion. The mobile robot controller according to claim 1. 10. The number of parameters is two, the two parameters are each converted into an n-bit binary integer, and the two n-bit parameters are combined into one 2n.
It is a binary number of bits, any bit of 2n bits is inverted with a probability of K%, the 2n-bit binary number after inversion is separated every n bits, and the separated n-bit binary number is floated The mobile robot controller according to claim 9, wherein the two new parameters are generated by converting into a decimal point number. 11. The force acting on the mobile robot is a first force received from a surrounding mechanical field, a second force received from the other mobile robot located in a moving direction, and the other mobile robot is moved. The second force is calculated by using a third force received from the space before movement generated by the above and a fourth force received by the movement of the other adjacent mobile robot as a combined force. When the coefficient is A
_P , the coefficient when calculating the third force is A _a , the coefficient when calculating the third force
_Let AV be the coefficient for calculating the force of, and α and β be the two parameters, α and β, respectively.
11. The mobile robot controller according to claim 10, wherein is represented by α = A _a / A _P β = A _V / A _P. 12. A mobile robot control method of a mobile robot control device, wherein a force acting on a mobile robot is obtained, and a moving direction of the mobile robot is determined from the obtained force, wherein the force is calculated from another mobile robot. Receive a parameter specified by a coefficient, select a new parameter from the received plurality of parameters, select one of the selected parameter and the current parameter based on probability, and select the selected parameter. A mobile robot control method comprising: converting a part of a parameter into a new parameter, and determining a moving direction of the mobile robot based on the new parameter.