JP2006048472A - Self-travelling mobile vehicle - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a slef-travelling mobile vehicle which can flexibly cope with a travelling route in response to needs for a range wherein a user wants to move the vehicle preponderantly, or the like, by causing the vehicle to learn the running route while actually running the vehicle. <P>SOLUTION: A user uses an input means to input evaluations about a plurality of (n) routes actually traced by a moving means (S46 and S47) in initial travelling (S44) and travelling along routes after initial travelling (S45), and a control means learns travelling routes on the basis of user's evaluations to each travelling route (S54). This learning of routes evaluated by the user is repeated to obtain travelling routes which the user desires, such as a travelling route passing a range where the vehicle should travel preponderantly, several times in the existence of the range. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a self-propelled mobile vehicle including a control unit that controls a moving unit based on a detection result of a detecting unit.

The self-propelled cleaner currently marketed mainly cleans the room by repeating straight forward and backward rotation or on-site rotation mainly according to certain rules or randomly. In addition, in a conventional self-propelled mobile vehicle, there has been proposed a mobile robot that travels smoothly through the room by reciprocating while shifting the travel position by a predetermined interval (Patent Document 1). The mobile robot includes boundary detection means for detecting a boundary of a region around the main body, and movement control means for controlling the movement of the main body by controlling the traveling means and the movement direction changing means. Is controlled so that the main body reciprocates in the above-mentioned area by repeating straight movement and reversal, and moves when the boundary of the area that cannot advance further forward and laterally is detected at least twice. By completing the above, it is possible to travel in the region efficiently without performing complicated control with a simple configuration.
JP 2003-241835 A (paragraphs [0017] to [0018], FIGS. 1 to 4)

  With these conventional self-propelled cleaners, it is possible to mainly clean the room uniformly. On the other hand, there are cases where the user does not simply drive the purchased self-propelled cleaner in the room at a uniform rate, but wants to clean a specific range with a focus such as a corner of a living room at home. However, a running pattern corresponding to the user's needs is not generally prepared in a self-propelled cleaner, and it is difficult for the user to create such a running pattern according to the time.

  By the way, as a technique for obtaining an optimal solution or an approximate solution of a plurality of independent parameters, there is a technique called a genetic algorithm obtained by imitating the evolution of a living organism. For example, as described in paragraphs [0010] to [0012] of JP-A-2001-84285, the genetic algorithm considers each parameter as a gene and one parameter solution as an individual. Utilizing the mechanism of biological evolution in which individuals with higher adaptability to the environment survive survival competition by repeating operations that follow genetic phenomena such as superiority or inferiority, drought, gene crossover, and mutation. Thus, an optimal solution or an approximate solution thereof can be derived.

A cognitive map, such as a cleaning robot or a security robot, has a cognitive map that is spatial knowledge to express the environment, and can be flexibly handled even when there is an environmental change by stochastically expressing the cognitive map. For each component, cognitive map storage means for storing the definition information of the component and probability information indicating the probability of the component, and the definition information and probability of each component stored in the cognitive map storage means An environment recognition means for probabilistically recognizing the environment using information has been proposed (Patent Document 2). As the environment recognition means, the shortest distance from a certain starting point to the arrival point is determined by a genetic algorithm. It has also been proposed to use a cost path or a path that satisfies a predetermined constraint condition in a realistic time.
JP-A-9-222852 (paragraphs [0010] to [0014], [0037] to [0041], FIGS. 1 and 3)

In addition, when designing a route configuration of a mobile unit such as an unmanned transport system, the initial candidate selection unit uses a required specification from a route configuration candidate database that stores a plurality of route configuration candidates in advance in order to enable a design close to the optimum value. One or a plurality of route configuration candidates close to is selected, and a new route configuration candidate is generated by an optimization technique such as a genetic algorithm in the new candidate generation unit 3 using the selected route configuration candidate. Then, the optimum candidate selection unit selects the most desirable route configuration candidate from among all candidates including the newly generated route configuration candidate using an evaluation function or the like, and the correction unit configures the configuration of the route configuration candidate. There has been proposed a path configuration design apparatus for a moving body that obtains a path configuration closer to the required specification as an optimal solution by exchanging some of the components with other components based on a predetermined rule (Patent Document 3).
JP 2000-66723 A (paragraph [0008], FIGS. 1 to 2)

  In a self-propelled mobile vehicle, normally, map information of a travel range is not input to the self-propelled mobile vehicle at the time of starting the travel. Although a method of inputting a traveling area based on image information obtained by photographing a floor surface from a camera provided on a ceiling can be considered, it is expensive to introduce imaging equipment such as a camera and input / setting software. In addition, setting and inputting a travel restriction condition for a self-propelled mobile vehicle imposes a complicated operation on the user. Therefore, in a self-propelled mobile vehicle, there is a problem to be solved in that a travel route to be traveled while actually traveling the mobile vehicle is learned so that a user can determine a travel route desired by the user at low cost.

  An object of the present invention is to allow a user to learn a travel route while actually traveling a vehicle, for example, to flexibly correspond to a travel route according to needs for a range that the user wants to move intensively. It is to provide a self-propelled mobile vehicle that can do.

  In order to solve the above problems, a self-propelled mobile vehicle according to the present invention includes a moving means, a detecting means for detecting a self-position, a control means for controlling the moving means based on a detection result of the detecting means, and the control Input means to the means, the control means to learn a plurality of travel routes actually traced by the moving means based on the evaluation of each travel route of the user input using the input means It is characterized by.

  According to this self-propelled mobile vehicle, the user inputs evaluations using the input means for a plurality of travel routes actually traced by the moving means, and the control means determines the travel routes based on the user's evaluation for each travel route. To learn. Therefore, the travel route is a route desired by the user, such as a travel route that passes through the range when there is a range to be moved with priority, based on the user's evaluation. When the travel route is determined, the control means controls the movement means based on the detection result of the detection means for detecting the self position, and the mobile vehicle can automatically travel along the determined travel route.

  In the above-described self-propelled mobile vehicle, a genetic algorithm can be used for learning of the travel route. As described above, the genetic algorithm is a technique for obtaining an optimal solution or an approximate solution of a plurality of independent parameters by imitating the evolution of a living organism. In the genetic algorithm, the survival path obtained as an optimal solution or an approximate solution thereof is obtained by using a mechanism of biological evolution in which individuals with higher adaptability to the environment survive. It can be guided as a travel route to be determined.

  In the above-mentioned self-propelled mobile vehicle that uses a genetic algorithm to learn the travel route, the genetic algorithm is based on the number of coordinates of the passing points within the priority range that the self-propelled mobile vehicle traveled as the travel route. Thus, it can be applied to obtain the optimum approximate solution of the parameters by regarding the parameters as the operation, coordinates, and traveling direction of the traveling route and the parameter solutions thereof as genes and individuals, respectively. In other words, since the travel route can be regarded as a series of data with parameters such as coordinates and travel direction, each parameter can be regarded as a gene, and based on the number of coordinates of passing points within the priority range, Based on the evaluation that the user gives to the travel route, it is possible to obtain the parameter solution as the optimal gene sequence by repeating the operation following the genetic phenomenon such as individual superiority or inferiority, drought, gene crossover, mutation, etc. This is the optimal approximate solution of the route.

  In the self-propelled mobile vehicle to which the genetic algorithm is applied, the priority range is the initial value for each initial travel route of the user input using the input means for a plurality of initial travel routes actually traced by the moving means. It can be determined based on the evaluation. For a plurality of initial travel routes actually traced by the moving means, a priority range in which the mobile vehicle should travel is determined based on an initial evaluation input by the user using the input means.

  In the above-mentioned self-propelled mobile vehicle that determines the priority range based on the initial evaluation, the initial evaluation for each initial travel route consists of assigning points for each coordinate that each initial travel route follows, and the priority range is all points Can be obtained as a coordinate range in which the cumulative value for each coordinate of the travel route exceeds a predetermined threshold. That is, for each coordinate followed by each initial travel route, a score is assigned from the viewpoint of the user's evaluation, the score is accumulated for each coordinate for all travel routes, and the accumulated value is emphasized as a range of coordinates exceeding a predetermined threshold value. A range is defined. Therefore, the priority range is determined by reflecting the determination that the user preferably travels with priority.

  In the above-described self-propelled mobile vehicle that learns the travel route based on the user's evaluation, the travel route other than the highest evaluation route with the highest evaluation among the plurality of travel routes obtained by learning the travel route is moved. The user's evaluation when actually traced by the means is input using the input means, and the learning can be repeated based on the totaled result obtained by collecting the user's evaluation including the highest evaluation path. That is, the user's evaluation when actually traveling by the moving unit is input using the input unit for each of the traveling routes having the second or lower evaluation among the plurality of traveling routes obtained by learning. A higher-evaluation travel route can be obtained by repeating the learning of the travel route again based on the totaled result of the user evaluations including the highest evaluation route.

  In the above-described self-propelled mobile vehicle that learns the travel route based on the user's evaluation, the user may not necessarily input the evaluation. In such a case, a predetermined evaluation can be used as the above evaluation. By using a predetermined evaluation, it is possible to continue the learning of the travel route.

  In the above-described self-propelled mobile vehicle that learns the travel route based on the user's evaluation, the user may not necessarily input the evaluation. If you do, you can end learning. By ending the learning of the travel route, it is avoided that the learning operation mode of the control means continues longer than necessary. The learning can be ended when, for example, the number of learnings is counted and the count reaches a predetermined value or more and there is no input from the user.

  As described above, in a self-propelled mobile vehicle, it is possible to estimate and learn constraint conditions at the same time by learning the travel route while actually traveling the mobile vehicle. It is possible to flexibly adapt to the part that the user wants to move mainly according to the user's intention. Further, when another constraint condition is to be imposed, a route that satisfies the constraint condition can be determined only by inputting an evaluation for each run without re-inputting such a condition. Furthermore, in determining the travel route, the survival route obtained as an optimal solution or its approximate solution is obtained by using a mechanism of biological evolution in which individuals with higher fitness to the environment survive by using a genetic algorithm. Can be determined as the travel route to be selected.

  An embodiment of a self-propelled mobile vehicle according to the present invention will be described based on the drawings. FIG. 1 is an external view of an embodiment of a self-propelled mobile vehicle according to the present invention. Two traveling drive wheels 3 and 3 facing left and right are respectively attached to the bottom surface of the main body 1, and the main body 1 travels at an equal distance from the drive wheels 3 and 3 in the same manner as a commercially available cleaner. An auxiliary wheel 4 is provided to support the wheel. Inside the main body 1, a control unit 2 for controlling the drive wheels 3 and 3, the distance sensor 7 and the working unit 8 is provided. A distance sensor 7 is provided around the main body 1 so that the distance to the surrounding walls and obstacles can be measured. Each drive wheel 3 is driven by a drive motor 5, and an output rotation amount can be detected by an encoder 6 provided in each drive motor 5.

  According to the output of the control unit 2, the driving wheels 3 and 3 rotate simultaneously in the forward direction when moving forward, and rotate simultaneously in the backward direction when moving backward. Further, when the mobile vehicle turns, both drive wheels 3 and 3 rotate in different directions. The auxiliary wheel 4 rotates freely so that the main body 1 can smoothly travel along these movements, such as a ball caster. By the combination of the forward movement, the backward movement, the left and right turning, and the stop operation, the main body 1 can freely travel in the room. At this time, the encoder 6 transmits the output rotation amount of the drive motor 5 to the control unit 2, and the control unit 2 can calculate the movement amount of the main body 1 using this.

  A working unit 8 is provided in front of the main body 1, and for example, the floor surface can be cleaned by the working unit 8. The working unit 8 may be provided not only on the bottom surface of the main body 1 but also on the upper surface of the main body 1 when a monitoring camera is mounted. On the upper surface of the main body 1, an input unit 9 is provided for the user to input an evaluation for the travel route of the self-propelled mobile vehicle. In accordance with this input result, the mobile vehicle learns about the travel route thereafter and can travel on a route more in line with the user's intention. The input unit 9 may be a remote control type instead of the main unit 1.

  An example of the travel route of the main body 1 and a method by which the control unit 2 stores the route will be described with reference to FIGS. In FIG. 2, the outer periphery 21 is like a wall of a room. Moreover, the movement range 22 is a floor surface in the outer periphery 21, and the main body 1 can move around freely on this. The obstacle 23 is a range to be avoided because the main body 1 cannot enter, such as a chest. The main body 1 can move within the movement range 22 while avoiding the obstacle 23 using the information from the distance sensor 7.

  Consider a case in which the main body 1 starts from an arbitrary point O in the room and moves in the room by performing either a straight movement or a rotation on the spot. The control unit 2 always calculates and stores information on the self-position of the main body 1 by accumulating the outputs of the encoder 6 every certain fixed time such as 1/1000 second. This is a means for constantly grasping the position and direction of the main body 1, and the same thing can be performed by imaging the main body with a camera installed on the ceiling or the like.

  FIG. 3 is a diagram showing the route information from the point O to the point P in FIG. 2 in the form of a list (however, this is shown as an example, 2). Considering the X and Y coordinates with the point O in FIG. 2 as the origin, the direction of the main body 1 is set to 0 degree in the direction coinciding with the Y axis. The control unit 2 stores the route of FIG. 2 by dividing it into a plurality of moving steps 31 as shown in the table of FIG. Among the movement steps 31 in FIG. 3, those in which the operation column is a straight line appear as arrows in FIG. In the case of start, end, and rotation, since there is no movement of X and Y, it does not appear in FIG. For example, in step 1, the operation is started at the coordinates (0, 0) at an angle of 90 degrees (that is, toward the X-axis direction). In the movement step 2, it goes straight and reaches the coordinates (0, 1), and the angle becomes 90 degrees. Thereafter, the process ends in step 25 in the same manner.

  Next, the flow of operation and the flow of learning of the travel route in the self-propelled mobile vehicle according to the present invention will be described with reference to the flow chart of operation in FIG. The numbers with S in parentheses are the corresponding process numbers in FIG.

  When the user instructs start (step 41, abbreviated as “S41”, the same applies hereinafter), the route number is set to 1 (S42). Next, it is determined whether there is route data (S43). In the determination of S43, since there is no route data at the initial stage of learning, the main body 1 shifts to the initial travel (S44). This initial traveling may be performed according to a certain rule, or may be repeated straight and rotating at random, and is not specified as any one. For example, the vehicle 1 travels for a certain time, and when one initial travel is finished, the main body 1 returns to a fixed point. When there is route data as in the case where the initial travel has ended, the determination in S43 is Yes and the route travel is performed (S45). After the initial travel in S44 or the route travel in S45, the main body 1 requests the user's evaluation and enters a state of waiting for user evaluation (S46). In response to the request, the user can input an evaluation corresponding to the degree of satisfaction (a five-level evaluation of 1, 2, 3, 4, and 5 or a three-level evaluation of × Δ ◯). In this embodiment, it is assumed that three-stage evaluation (“◯” = 1, “Δ” = 0, “×” = − 1) is performed. At this time, for example, the route information may be illustrated on the main body 1 or the remote controller, or nothing may be displayed when evaluation is possible by checking the indoor conditions such as cleaning. .

  It is determined whether or not there is an evaluation input by the user (S47). If there is an evaluation input by the user in S47, it is determined whether or not the vehicle has traveled to route n (S50). If there is no evaluation input by the user in S47, it is determined whether or not a predetermined time has elapsed, that is, the elapsed time in the evaluation waiting state (S48). If the predetermined time has not elapsed in the determination of S48, the process returns to S46 and the evaluation waiting state is continued. If the predetermined time has elapsed in the determination of S48, a predetermined evaluation value determined in advance is added (S51), and the process proceeds to S50. If the vehicle has not traveled to the route n in the determination of S50, the route number is updated by adding 1 (S49), and the process returns to S43.

  If it is determined in S50 that the vehicle has traveled to the route n, it is determined whether or not the user has never been evaluated (S52). If the user has never been evaluated, learning is stopped because only the predetermined evaluation value is assigned in S51 (S55). If the user's evaluation is at least once, the evaluation is totaled (S53), and further route learning is performed (S54). After the route learning in S54 or after the learning stop in S55, the process returns to S42.

  FIG. 5 shows an example in which n is 9 in FIG. The numbers in the upper right of each figure are the evaluations at that time converted to values. In addition, the route numbers 1 to 9 in the upper left are abbreviated with numbers for convenience, such as the route 1 in the upper left route and the route 2 in the adjacent route in order to distinguish the travel routes in the figure (hereinafter referred to as the route numbers 1 to 9). (The same notation is used in FIGS. 6, 9, 10, and 11). A target range 61 (a portion hatched in the moving range in the figure) indicates a range in which the user wants to move the moving vehicle with priority. The target range 61 is a priority cleaning range in which cleaning is performed intensively when the working unit 8 performs cleaning, and is a focus monitoring place where monitoring is frequently performed when the working unit 8 is a monitoring camera.

  When the user is traveling within the range for one initial travel, “◯” (+1 point), and those far from the range are “△” (0 point) and “×”. (-1) Give a score.

  When the initial running is completed nine times, the control unit 2 adds up the evaluations and routes at that time, as indicated by S53 in FIG. First, as shown in FIG. 6, the product of each evaluation and the passing point is taken. On the map, 1 is the coordinate where the moving vehicle has passed, and 0 is the coordinate where the vehicle has not passed. Next, when the result of each product is added for each coordinate, a score is obtained for each coordinate. FIG. 7 is a graph of the total for each coordinate of each route in FIG. FIG. 8 is a graph in which the graph of FIG. 7 is divided by threshold values, and this makes it possible to specify the priority range 81. In FIG. 8, the range where three or more points are acquired is the priority range 81, which is close to the target range 61 intended by the user in FIG. In FIG. 8, the threshold value is set to 3, but the determination method of the threshold value may be the discriminant reference method (2 in the case of the embodiment) or the like other than the predetermined fixed value.

  Next, the travel route indicated by S54 in FIG. 4 is learned. FIG. 9 is a diagram in which the priority range 81 that appears as a result of FIG. 8 is overlapped with the travel route of FIG. Regarding the travel route, when an arrow with the same traveling direction continues, it is abbreviated with one arrow. The number on the upper right indicates the effective coordinate number 91 as the number of coordinates of the passing point included in the priority range 81. As the number of effective coordinates 91 of the passing points is larger, it can be considered that the route is more excellent. Based on this number, a genetic algorithm technique is applied to perform processing operations following genetic phenomena such as rearrangement of travel routes, crossover of wrinkles and route genes, and mutation.

  As shown in FIG. 10, first, the travel routes are rearranged in descending order of the effective coordinate number 91. Thereafter, the one having the highest effective coordinate number 91 is propagated and later genetically mated with another traveling route (route 1 and route 2 in FIG. 10). Since the travel route 1 is the highest evaluation route of nine initial travels, the travel route 1 is left as the travel route 1 and is propagated to the travel route 2 for comparison with the next highest evaluation route. Further, at this time, the one with the low effective coordinate number is replaced (ie, hesitated) with the one with the highest effective coordinate number 91, and the gene is mutated later. In the embodiment, the 0-point travel route is replaced with the 9-point travel route (travel routes 7, 8, 9 in FIG. 10).

  Next, the crossing of the route step 31 is performed with respect to the travel route that does not hesitate. Two pairs are created from the travel route 2 of FIG. 10, and the route steps 31 are replaced with each other. For example, a pair of the travel route 2 and the travel route 3 is formed and crossed at a portion that passes through the same portion of the travel route, that is, a circled portion on each route. As a result of the crossover, the travel route 2 and the travel route 3 are as shown in FIG. 11, and the dotted line portions in FIG. The same applies to the travel route 4 and the travel route 5. If there is a surplus as a result of making two pairs as in the travel route 6, a pseudo pair is formed with the travel route 7 and the same crossover is performed. However, since the traveling route 7 is a deceived traveling route and is propagated to mutate the traveling route 1, the traveling route 7 does not reflect the result of the crossover.

  Next, regarding the travel route (travel routes 7, 8, 9) that has been deceived and replaced with the result of the travel route 1, a route obtained by mutating the route step 31 is set as a new travel route. The point to be mutated is an arbitrary path step other than the effective coordinates. The results of these crossovers and mutations are the travel routes 7, 8, and 9 in FIG. The route that is mutated is indicated by a dashed line.

  Thus, learning of the travel route is completed (S54). The learning result in FIG. 11 is considered to reflect the user's intention because there are more travel routes passing through the target range 51 than in the initial travel in FIG.

  Next, when the moving vehicle is activated, the control unit 2 sequentially reads the travel routes 1 to 9 in FIG. 11 (the route 1 is unchanged from the state shown in FIG. 9), and the route number is set to 1 as shown in FIG. After setting (S42), the flow is resumed. In S43, since it is determined that there is route data, the process proceeds to S45 and the route travel is performed. For each actual travel, a new user evaluation is requested and waiting for evaluation (S46). Thereafter, when the travel route reaches a predetermined number of times (n), the evaluation is totaled (S53). When there is no user evaluation, it is the same as in the case of initial travel (S52, S55).

  By repeating the above steps many times, the self-propelled mobile vehicle can always move within a range that the user intends to move with priority.

In addition, when the learning reaches the user's requested level, there is a case where the evaluation is not performed even if the traveling is finished and the user is requested to input the evaluation. In that case, it is possible to eliminate the troublesomeness of the user to perform the evaluation permanently by always setting a certain value as the evaluation value (S51) or stopping the learning (S55).

The external view which shows the outline | summary of the self-propelled mobile vehicle by this invention. The movement example (coordinates, movement) of the self-propelled mobile vehicle by this invention, and a memory principle figure. The movement example (path | route step) of the self-propelled mobile vehicle by this invention, and a memory principle figure. The action | operation flowchart of the self-propelled mobile vehicle by this invention. The principle figure which shows the initial driving | running | working in the route learning of the self-propelled mobile vehicle by this invention. The principle figure of the route learning of the self-propelled mobile vehicle by the present invention (score assignment). The principle diagram of the route learning of the self-propelled mobile vehicle according to the present invention (total score and diagonal graph). The principle figure (priority range) of the route learning of the self-propelled mobile vehicle by this invention. The principle figure (the number of effective coordinates) of the course learning of the self-propelled mobile vehicle by the present invention. The principle diagram of the route learning of the self-propelled mobile vehicle by this invention. The principle diagram of the route learning of the self-propelled mobile vehicle by this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Main body 2 Control part 3 Drive wheel 4 Auxiliary wheel 5 Drive motor 6 Encoder 7 Distance sensor 8 Working part 9 Input part 21 Outer periphery 22 Movement range 23 Obstacle 31 Movement step 61 Target range 81 Important range 91 Effective coordinate number

Claims (8)

  A self-propelled mobile vehicle comprising a moving means, a detecting means for detecting a self-position, a control means for controlling the moving means based on a detection result of the detecting means, and an input means to the control means. The control means learns a plurality of travel routes actually traced by the moving means based on an evaluation of each travel route input by the user using the input means. A traveling vehicle.   2. The self-propelled mobile vehicle according to claim 1, wherein a genetic algorithm is used for learning of the travel route.   3. The self-propelled mobile vehicle according to claim 2, wherein the genetic algorithm is configured to calculate the travel based on the number of coordinates of a passing point within a priority range traveled by the self-propelled mobile vehicle as the travel route. A self-propelled mobile vehicle characterized in that it is applied to obtain an optimum approximate solution of the parameters by regarding the parameters of the movement, coordinates, and traveling direction of the route and the parameter solutions thereof as genes and individuals, respectively.   4. The self-propelled mobile vehicle according to claim 3, wherein the priority range is input using the input unit with respect to a plurality of initial travel routes actually traced by the moving unit. A self-propelled mobile vehicle characterized in that it is determined based on an initial evaluation of the vehicle.   5. The self-propelled mobile vehicle according to claim 4, wherein the initial evaluation for each initial travel route includes assigning a score for each coordinate followed by each initial travel route, and the priority range includes all of the points. A self-propelled mobile vehicle characterized in that an accumulated value for each coordinate of the travel route is obtained as a range of the coordinates exceeding a predetermined threshold.   2. The self-propelled mobile vehicle according to claim 1, wherein each of the travel routes other than the highest evaluation route with the highest evaluation among the plurality of travel routes obtained by learning of the travel route is actually used by the moving means. The user's evaluation at the time of tracing is input using the input means, and the learning is repeated based on a totaling result obtained by aggregating the user evaluation including the highest evaluation path. Moving car.   2. The self-propelled mobile vehicle according to claim 1, wherein when the user's evaluation is not input, the evaluation is a predetermined evaluation.   The self-propelled mobile vehicle according to claim 1, wherein the learning is terminated when the user's evaluation is not input.

Images