JP4157731B2 - Robot cleaner and robot cleaner control program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a self-propelled robot cleaner that performs a cleaning operation according to cleaning route data automatically generated based on room map data.
[0002]
[Prior art]
In a robot that moves in a space with various obstacles, how to deal with obstacles when encountering obstacles is an important point. As a method for coping with obstacles, it is common to use stored data relating to obstacles.
[0003]
In the example disclosed in Japanese Patent Laid-Open No. 2001-129787, the robot holds image data of an obstacle. When there is an obstacle on the moving path, the robot takes in the obstacle with an image sensor and compares and determines whether the obstacle is registered in advance. If it is determined that the obstacle is a registered obstacle, an avoidance operation is performed according to an avoidance program corresponding to the obstacle. If it is determined that the obstacle is an unknown obstacle, the robot determines whether there is a person in the vicinity, and if the presence of the person is confirmed, asks for help by voice or gesture.
[0004]
In the example disclosed in Japanese Patent Laid-Open No. 9-222852, room map data is expressed as a cognitive map that is spatial knowledge, and component definition information and component probability information are stored. Then, route generation is performed using this cognitive map. When the robot encounters an obstacle different from the cognitive map on the movement route, the probability information of the corresponding component (route) is changed. As a result, the changed information is utilized in the next movement route generation.
[0005]
In addition to the image sensor described above, an ultrasonic sensor is used as a sensor provided in the robot for detecting an obstacle. The distance between the mobile robot and surrounding obstacles is measured by an ultrasonic sensor, and room map generation processing and robot position estimation processing are performed (for example, the IEEE document “IEEE Journal of Robotics and Automation (IEEE Journal) of Robotics and Automation) "(published in 1988), RA-3, pages 249 to 265).
[0006]
An example of a map storage method for determining a travel route in advance is disclosed in JP-A-8-16241. In this example, for each region of the room, the attributes of a part where the entire work vehicle can travel, a part where only the work part of the work vehicle can be inserted, and a part where travel and insertion are impossible, Each area with attributes is registered in the room map. A cleaning route is planned using the room map with attributes.
[0007]
[Problems to be solved by the invention]
Regarding the countermeasure against obstacle encounter, in the example of the above Japanese Patent Laid-Open No. 2001-129787, since the image data is a planar information source, it is difficult to guess the geometric structure of the obstacle, and the memory is stored. It is difficult for a moving robot to image an obstacle in exactly the same environment as the obstacle image data being performed. Therefore, there is a possibility of causing erroneous recognition that the obstacles are different even though they are the same obstacle. In addition, since no data update or learning process is performed, the same recognition error may occur during the next movement.
[0008]
In the example of Japanese Patent Application Laid-Open No. 9-222852, when an obstacle different from the cognitive map information is encountered, a dynamic map update is performed, but the update contents include probability information indicating the reliability of the corresponding route. The actual map / route shape is not changed. Therefore, route reconstruction cannot be performed in real time while moving, and if there is an unknown obstacle, the operation cannot be completed. Furthermore, since the shape of the obstacle is not stored in the cognitive map, it is difficult to precisely clean around the obstacle.
[0009]
In a robot that does not simply move but also cleans while moving, it is required to accurately grasp the shape of the obstacle. When an image sensor is used as the obstacle detection sensor for that purpose, the image data becomes a planar information source as described above, so that the shape grasp becomes insufficient. In addition, when an ultrasonic sensor is used, the distance to the obstacle can be measured to some extent. However, since the ultrasonic wave is spread, sufficient accuracy for grasping the shape cannot be obtained. In addition, if a relatively expensive sensor such as an ultrasonic sensor is used, the production cost of the robot is inevitably increased. When targeting robot cleaners for general households, it is desirable to manufacture them at the lowest possible cost.
[0010]
In the example of JP-A-8-16241, a cleaning route is generated depending on the attribute of the area in the room map. However, since dynamic data is not updated, an environment such as moving an obstacle is used. If a change occurs, the cleaning operation cannot be completed.
[0011]
A first object of the present invention is to provide a robot cleaner that accurately reconstructs a cleaning route and resumes cleaning when an obstacle is encountered, and a robot cleaner control program for controlling its operation. It is in.
[0012]
A second object of the present invention is to provide a low-cost robot cleaner that dynamically updates a room map with high accuracy and a robot cleaner control program for controlling its operation.
[0013]
A third object of the present invention is to provide a robot cleaner that cleans an obstacle by accurately and quickly recognizing an obstacle, and a robot cleaner control program for controlling the operation thereof.
[0014]
[Means for Solving the Problems]
The first object is to provide an obstacle detection sensor for detecting the presence or absence of an obstacle and recognizing its shape, room map data storing a room map formed by registering obstacles in a room, and room map data. And attribute data storing obstacle attributes (existence history, cleaning method, etc.) registered in the above, and cleaning route data storing a cleaning route created based on the room map data as a cleaning route, When cleaning is performed according to the cleaning route, the presence or absence of an obstacle is detected using an obstacle detection sensor. If an obstacle is detected, whether or not an obstacle is registered on the room map data at that position. The obstacle detected when the obstacle is not registered is determined as an unknown obstacle, the shape of the unknown obstacle is recognized using the obstacle detection sensor, and the shape that matches the recognized shape is detected in the room. map If the estimated obstacle can be estimated by referring to the data and the attribute data, the obstacle of the estimated shape is registered in the room map data, the cleaning route data is reconstructed, and the shape that matches the recognized shape is estimated. If this is not possible, it is achieved by determining the unknown obstacle as a newly appearing obstacle, registering it in the room map data, and reconstructing the cleaning route data.
[0015]
By adopting such means, when the robot cleaner encounters an obstacle, the distinction of known or new appearance of the obstacle is clearly distinguished, and the room map is updated at each obstacle encounter accordingly. Is done. Therefore, even when an unknown obstacle is encountered, the cleaning route can be accurately reconstructed, and cleaning can be resumed according to the reconstructed cleaning route.
[0016]
The second object is to acquire information from the infrared sensor and the tactile sensor for the robot traveling based on the route data stored in the storage means, and use the information to determine whether there is an obstacle on the route. It is achieved by detecting, recognizing the shape of the obstacle using information from the tactile sensor, and executing a map creation method for creating the route data using the detection result and the recognition result. .
[0017]
By using the infrared sensor and the tactile sensor, it is possible to accurately grasp the shape of the obstacle. Therefore, it is possible to create the cleaning route data with high accuracy and to create the room map based on the route data. Furthermore, both the infrared sensor and the tactile sensor can be configured using inexpensive sensor elements, and a low-cost robot cleaner can be realized.
[0018]
The third object is to provide at least a sensor, obstacle data, and cleaning route data stored as a cleaning route as a cleaning route, and from the sensor when cleaning is performed according to the cleaning route. This is achieved by obtaining information, reconstructing the cleaning route using the information and the obstacle data, and when cleaning route is reconstructed, continuing cleaning according to the reconstructed cleaning route. The
[0019]
Obstacle data, which is an attribute unique to the obstacle (existing history, cleaning method, etc.) is recorded in advance and reflected in the reconstruction of the cleaning route, so that the obstacle is recognized accurately and at high speed for cleaning. Can be performed.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, a robot cleaner according to the present invention and a robot cleaner control program for controlling its operation will be described in more detail with reference to the embodiments of the invention shown in the drawings.
[0021]
A system configuration example of a robot cleaner using the robot cleaner control program of the present invention is shown in FIG. The robot cleaner includes a fixed intelligent processing unit 200 and a robot unit 100 that performs cleaning. The intelligent processing unit 200 and the robot unit 100 are connected wirelessly via the wireless communication module 225 and the wireless communication module 150, and bidirectional wireless communication is formed. The intelligent processing unit 200 sends an operation command to the robot unit 100 wirelessly, and the robot unit 100 sends the sensor data to the intelligent processing unit 200 wirelessly.
[0022]
The intelligent processing unit 200 stores a CPU (central processing unit) 221 that operates based on the robot cleaner control program 235, a memory 500 that the CPU 221 uses during processing, a wireless communication module 225, a robot cleaner control program 235, and a database 400. The storage device 222 is connected to the internal bus 224. These units connected to the internal bus 224 form a computer. The database 400 includes room map data 401, attribute data 402, and cleaning route data 403.
[0023]
As will be described later, room map data 401 is data indicating where and how obstacles (cupboards, tables, chairs, walls, pillars, etc.) in the room to be cleaned are arranged. Yes, these are expressed as ordinate and abscissa coordinate values starting from the cleaning start point. Therefore, the position where the obstacle is arranged, the shape and direction of the obstacle can be known from the coordinate value.
[0024]
The attribute data 402 includes attributes of each obstacle, that is, the type of the obstacle (whether it is a wall or a general obstacle, whether it is a fixed obstacle or an object whose position changes daily), the ID of the obstacle (identification) Number), cleaning method (cleaning only around the obstacle or precision cleaning to clean the inside of the obstacle), number of history areas (number of areas where obstacles existed in the past), etc. The cleaning route data 403 is data representing a cleaning route determined from the room map data 401 and the attribute data 402 by the coordinate values.
[0025]
In the robot unit 100, the control unit 300 is connected to the wireless communication module 150, and the drive unit 160 and the sensor data acquisition unit 110 are connected to the control unit 300. The drive unit 160 includes a mobile robot drive unit 120 that drives a motor of a wheel, and a cleaning mechanism drive unit 130 that drives a suction unit that performs cleaning. The sensor data acquisition unit 110 includes an obstacle detection sensor 115 that detects the approach of the obstacle and the shape of the obstacle, and a wheel encoder 128 that measures the number of rotations of the wheel.
[0026]
The control unit 300 receives an operation command from the intelligent processing unit 200 and controls the movement of the driving unit 160. The control unit 300 also receives an operation command from the intelligent processing unit 200 and transmits the sensor data from the sensor data acquisition unit 110 to the intelligent processing unit 200.
[0027]
The computer determines the operation of the robot unit 100 using each data in the database 400 and the sensor data transmitted from the robot unit 100, and transmits an operation command to the robot unit 100. In this way, processing according to the robot cleaner control program 235 is executed.
[0028]
In the above, the robot cleaner is shown as being separated into the intelligent processing unit 200 and the robot unit 100, but the intelligent processing unit 200 can be mounted on the robot unit 100 and integrated. In this case, the wireless communication module 225 and the wireless communication module 150 are omitted, and the control unit 300 is directly connected to the internal bus 224.
[0029]
FIG. 2 is a top view showing the configuration of the robot unit 100. In FIG. 2, a main body 101 of the robot unit 100 is an independent drive wheel type mobile robot having a left drive wheel 121, a right drive wheel 122, and an auxiliary wheel 123. By adjusting the respective wheel speeds, the robot unit 100 can move in any direction. It is possible to move. Rotary encoders 126 and 127 are installed on both wheels 121 and 122, respectively, and the number of rotations of both wheels can be measured. The mobile robot drive unit 120 is configured by the wheels 121 and 122 and respective motors (not shown in FIG. 2) for driving them (FIG. 1).
[0030]
Infrared sensors 111, 112, and 113 are installed around the robot unit main body 101, and it is possible to detect the presence of relatively close obstacles on the front, side, and rear. In addition, a wireless communication module 150 that realizes wireless communication with the intelligent processing unit 200 is installed. Although not shown, the control unit 300 is installed adjacent to the wireless communication module 150.
[0031]
Furthermore, the robot unit main body 101 is provided with an arm part 131 that can be moved horizontally to the left and right and a suction part 132 that can be rotated about the tip part of the arm part 131 as a central axis. The suction unit 132 includes a dust collection unit 136 for collecting dust, a suction motor 135 for performing a suction operation by suction, and a tactile sensor 114. The tactile sensor 114 detects that an object has been touched.
[0032]
The arm part 131 and the suction part 132, which are cleaning mechanisms, can operate independently of the robot unit main body 101 and can be detached from the robot unit main body 101. Accordingly, it is possible to prepare several cleaning mechanisms having different shapes and operations, and perform work by switching the cleaning mechanism in accordance with the state of the cleaning environment.
[0033]
The infrared sensors 111, 112, 113 and the tactile sensor 114 constitute an obstacle detection sensor 115, and the rotary encoders 126, 127 constitute a wheel encoder 128 (FIG. 1). The infrared sensors 111, 112, and 113 are realized by a structure in which a light emitting diode and a photodiode are combined, and reflected light from an infrared obstacle emitted from the light emitting diode is detected by the photodiode. It is realized using a piezoelectric element or a small micro switch. All of these elements are very inexpensive, and the obstacle detection sensor 115 can be realized at low cost.
[0034]
FIG. 3 is a block diagram illustrating a relationship among the control unit 300, the sensor data acquisition unit 110, the mobile robot driving unit 120, and the cleaning mechanism driving unit 130 in the robot unit 100.
[0035]
When the operation command transmitted from the intelligent processing unit 200 is a drive command, the control unit 300 generates a control signal for operating the mobile robot driving unit 120 and the cleaning mechanism driving unit 130 based on the operation command, and periodically The control signal is transmitted to the mobile robot drive unit 120 and the cleaning mechanism drive unit 130. By this control signal, the rotation of the left wheel motor 124 and the right wheel motor 125 for driving the wheels 121 and 122 in the mobile robot drive unit 120 is controlled, respectively, and the cleaning mechanism drive unit 130 is further controlled. The cleaning mechanism drive unit 130 includes a suction motor 135, an arm unit slide motor 134 for moving the arm unit 131, and a suction unit rotation motor 133 for operating a rotation mechanism attached to the suction unit 132. The rotation of these motors is controlled respectively.
[0036]
In addition, the control unit 300 periodically acquires data from the sensor data acquisition unit 110. When the operation command transmitted from the intelligent processing unit 200 is a sensor data transmission request command, the control unit 300 transmits the sensor data to the intelligent processing unit 200 using the wireless communication module 150.
[0037]
FIG. 4 is a flowchart showing the processing contents of the control unit 300. First, the operation command received by the wireless communication module 150 is analyzed in step 301. Next, the process proceeds to step 302. If the analyzed operation command is a drive command, the process proceeds to step 308. On the other hand, if the analyzed operation command is a sensor data transmission request command, the process proceeds to step 303, and the latest sensor data stored in the robot unit 100 is transmitted to the intelligent processing unit 200 via the wireless communication module 150. In step 308, the drive target motor is set, the speed is set, and the movement distance is set based on the content of the drive command.
[0038]
The control unit 300 periodically executes an interrupt process 304 at intervals of 10 milliseconds to drive the motor and acquire sensor data. When the interrupt processing execution time is reached, in step 305, control signals for the drive motors of the wheel, arm, and suction unit are generated based on the values set in step 308 and transmitted to the drive motors of the drive unit 160. Next, the process proceeds to step 306, where the sensor data acquisition unit 110 acquires obstacle detection and wheel rotation number data as sensor data. The acquired sensor data is stored in the memory 307 as the latest sensor data. That is, the latest sensor data in the memory 307 is updated each time an interrupt process is performed.
[0039]
Next, the contents of processing executed by the computer of the intelligent processing unit 200 according to the robot cleaner control program 235 will be described. In the present invention, the robot cleaner being cleaned updates the room map data 401 on the spot when it detects that the situation of the room is different from the situation shown in the room map data 401, and performs cleaning. Reconstruct the route data 403. That is, whenever there is a difference, the room map data 401 is updated and the cleaning route data 403 is reconstructed. The difference is when an obstacle is moved away from the room or when a new obstacle that is not in the room is brought into the room.
[0040]
The respective update of the room map data 401 and the restructuring of the cleaning route data 403 each time the room map data 401 and the cleaning route data 403 are copied to the memory 500, respectively, and the temporary room map data shown later in FIG. When the cleaning is completed, the temporary room map data 501 and the temporary cleaning route data 502 are overwritten on the room map data 401 and the cleaning route data 403 and updated. The next cleaning is performed using the updated room map data 401 and cleaning route data 403.
[0041]
FIG. 5 is a flowchart showing the processing procedure of the intelligent processing unit 200 by the computer. When an obstacle is registered on the temporary room map data 501 at the location where the robot cleaner is moving, or when an obstacle is encountered while moving, first at step 201, the sensor data of the robot unit 100 is acquired. The sensor data from the unit 110 is analyzed. Specifically, the current position of the robot unit 101 is calculated based on data from the wheel encoder 128, and whether or not the tactile sensor 114 has touched an obstacle is detected. The infrared sensors 111 to 113 can roughly detect the presence of nearby obstacles and detect that they are approaching the obstacle, and the tactile sensor 114 can touch the obstacle in advance. You can foresee.
[0042]
Next, in step 202, the temporary room map data 501 and the sensor data analysis result are compared, and a difference is first determined based on the presence or absence of an obstacle. The determination of the presence or absence of an obstacle is performed based on the data from the tactile sensor 114 as described above. When there is an obstacle, the process proceeds to step 203, and subsequently, it is determined whether or not the obstacle is registered on the temporary room map data 501 at the position where the tactile sensor 114 touches the obstacle. If there is an obstacle registered, it is determined that there is no difference, and the process proceeds to step 211 to perform cleaning according to the temporary cleaning route data 502. Subsequently, if all the rooms have not been cleaned in step 212, the process returns to step 201 to continue the process.
[0043]
In step 202, when there is no obstacle at the current position, the process proceeds to step 204, and the obstacle registered on the temporary room map data 501 as the obstacle is moving is removed. In step 208, the temporary room map data 501 is updated by removing the obstacle, and in step 210, a cleaning route is reconstructed based on the updated temporary room map data 501. Next, in step 211, cleaning is performed according to the reconstructed temporary cleaning route data 502.
[0044]
If no obstacle is registered on the temporary room map data 501 at the position where the tactile sensor 114 touches the obstacle in step 203, the obstacle touched by the tactile sensor 114 is set as an unknown obstacle and the process proceeds to step 205. The shape of an unknown obstacle is recognized using the tactile sensor 114.
[0045]
Subsequently, in step 206, it is searched whether or not the temporary room map data 501 has an obstacle having a shape that at least partially matches the shape recognized by the unknown obstacle. If there is an obstacle with a matching shape, the process proceeds to step 207 to estimate the matching shape as a whole using the attribute data 402. In step 208, the estimated shape of the obstacle is registered in the temporary room map data 501. . Thereby, the temporary room map data 501 is updated.
[0046]
If there is no matching obstacle in step 206, the process proceeds to step 209, where the unknown obstacle appears as a newly appearing obstacle, grasps the entire shape of the unknown obstacle or the area where the unknown obstacle exists, The attribute data of the newly appearing obstacle is added to the attribute data 402. In step 208, the newly appearing obstacle whose shape or area is grasped is registered in the temporary room map data 501, and the temporary room map data 501 is updated.
[0047]
When all the cleaning is completed in step 212, the attribute data 402 is updated in step 213 (for example, when the obstacle that has been fixed is moved, the attribute is changed to an obstacle whose position changes), and the processing is performed. finish.
[0048]
As described above, in the present invention, the temporary room map data 501 and the temporary cleaning route data 502 are dynamically updated during cleaning, so that the cleaning can be completed even if the situation inside the room changes.
[0049]
The function realized by the above steps 202 to 208 is the (1) obstacle shape estimation function, the function realized by the step 213 is the attribute data learning function, and the function realized by the step 210 is (3) A cleaning route reconstructing function is added. In addition, functions realized by steps 209 and 208 are (4) a new registration function for unregistered data. Details of each will be described below.
[0050]
FIG. 6 shows a flow chart of the processing contents by the (1) obstacle shape estimation function. As described above, there are the following two situations in which a difference occurs between the sensor data and the temporary room map data 501. One is when an unknown obstacle is encountered, and the other is when there is no obstacle present in the temporary room map data.
[0051]
Which of the two factors has occurred is confirmed using the sensor data analysis result in step 601, and if there are no registered obstacles, the process proceeds to step 605 and the corresponding obstacle is searched from the attribute data 402. The shape of the corresponding obstacle is deleted from the temporary room map data 501. Thereby, the temporary room map data 501 is updated at the same time. Then, (3) the process is transferred to the cleaning route restructuring function.
[0052]
On the other hand, in step 601, if the cause of the data difference is due to encounter with an unknown obstacle, the process proceeds to step 602, where the shape of the unknown obstacle is recognized by the tactile sensor 114 installed in the suction unit 132, and the recognized shape is temporarily stored. Write to room map data 501. In step 603, the room map data 401 and the short-term memory attribute data 900 are referenced to search for shape data that at least partially matches the shape of the unknown obstacle.
[0053]
Here, the short-term memory attribute data 900 is the obstacle shape data and attribute data that existed on the room map data 401 and the attribute data 402 before, but these data are constant. Only the period is held in the storage device 222.
[0054]
If there is no shape data that at least partially matches the shape of the unknown obstacle in the room map data 401 or the short-term memory attribute data 900, it is assumed that the unknown obstacle is a newly appearing object. Moves processing to new registration of unregistered data.
[0055]
On the other hand, when there is a shape that at least partially matches the shape of the unknown obstacle, the number of obstacles that partially match the shape is confirmed. -Prioritize by referring to information other than shapes such as knowledge, and narrow down the matching shapes to one. In this way, if the number of matching shapes is narrowed down to one, the process proceeds to step 608, where the information on the obstacle with the matching shape is read from the attribute data 402 or the short-term memory attribute data 900, and the overall shape is estimated. In step 609, the temporary room map data 501 is updated based on the estimation result, and the process proceeds to (3) the cleaning route reconstruction function.
[0056]
Next, (4) new registration of unregistered data will be described. FIG. 7 shows the processing contents. In step 1001, it is determined whether the entire shape of the obstacle can be grasped or only a part of the grasp is grasped. In the case of grasping the entire shape, the obstacle type TYPE of the attribute data is registered in the attribute data 402 as a fixed obstacle and the other items are invalid, and the process is transferred to the (3) cleaning route reconstruction function.
[0057]
On the other hand, if only part of the shape is grasped, the process proceeds to step 1003, where a rectangular area is formed from the part of the grasped shape range, and is registered in the temporary room map data 501 as an obstacle. Then, the obstacle type TYPE of the attribute data is newly registered in the attribute data 402 as a fixed obstacle and other items are invalid, and the process is transferred to the (3) cleaning route reconstruction function.
[0058]
Next, (3) the cleaning route reconstruction function will be described. FIG. 8 shows the processing contents. As a result of the processing by the obstacle shape estimation function, when it is estimated that there are no obstacles present in the room map data 401, in step 702, a route is reconstructed for the unswept area of the temporary room map data 501 Write to temporary cleaning route data 502. This is shown in FIG.
[0059]
FIG. 9 a shows obstacles 414 and 415, cleaning route data 403, and a swept area 411 represented by coordinate values in the room map data 401. If the obstacles 414 and 415 in FIG. 9a are lost, the temporary room map data 501 is updated to the state without the obstacles 414 and 415 as shown in FIG. 9b, and the cleaning route 512 is reconstructed for the unswept area. Is done. Thereafter, the robot cleaner performs work according to the reconstruction route 512.
[0060]
On the other hand, the processing when an unknown obstacle is detected as a result of the processing by the obstacle shape estimation function and the overall shape thereof is estimated will be described. First, in step 703, area division is performed for the unswept area of the temporary room map data.
[0061]
FIG. 10 shows the processing contents of the area division. First, in step 1101, a vertical line is drawn from the vertices in the horizontal and vertical directions of the room map data 401 to perform region division. Next, in step 1102, a rectangular region is formed by integrating regions sharing sides, and only the region having the maximum area is selected. The region division is realized by performing the processing of step 1102 until there is no unselected region.
[0062]
After dividing the area in this way, the process proceeds to step 704, and a straight line route is created for each area. These straight line routes are integrated to generate a cleaning route, and this is written into the temporary cleaning route data 502.
[0063]
A schematic diagram of step 703 and step 704 is shown in FIG. FIG. 11 a shows obstacles 416 and 417, cleaning route data 403, and a swept area 411 represented by coordinate values in the room map data 401. When an unknown obstacle is detected on the cleaning route and the entire shape is estimated as the shape of the known obstacle 416, 417, it is assumed that the obstacle 416, 417 has moved to the position where the unknown obstacle is detected, and the temporary room map The data 501 is updated so that the obstacles 416 and 417 have moved as shown in FIG. Then, a perpendicular line is drawn from the vertex in the horizontal and vertical directions with respect to the unswept area, and the area is divided. Further, as shown in FIG. 11c, adjacent regions are integrated to form a rectangular region. Then, a straight route 418 is created for each region (FIG. 11d), and these are integrated to reconstruct the cleaning route 512 (FIG. 11e).
[0064]
Although not shown, when the shape of the unknown obstacle does not match the registered obstacle shape and it is determined that the unknown obstacle is a newly appearing obstacle, the obstacle shown in FIG. The objects 416 and 417 are left without moving. Therefore, in that case, a cleaning route different from the cleaning route 512 of FIG. 11e is reconstructed.
[0065]
Subsequently, in step 705, if a cleaning method that requires precision cleaning is registered in the attribute data of the obstacle, the process proceeds to step 706, and the cleaning method that requires precision cleaning is referred to as temporary cleaning route data 502. Integrate. FIG. 12 shows a cleaning method that requires precision cleaning, and FIG. 13 shows an integration method with temporary cleaning route data 502. In step 705, if the cleaning method that requires precision cleaning is not registered in the attribute data of the obstacle, the cleaning route 512 is written in the temporary cleaning route data 502.
[0066]
Since the suction part 132 of the robot cleaner shown in the present embodiment can operate independently of the robot unit main body 101, it can also enter an area through which the robot unit main body 101 cannot pass. The cleaning method registered in the attribute data 402 is a method of cleaning the inside of the obstacle by the suction unit 132.
[0067]
For example, as shown in FIG. 12a, in the obstacle 419 having a region 420 (the inside of the obstacle) through which the robot unit main body 101 cannot pass (a convex hull shape 421 surrounding the obstacle 419 is formed), the suction part 132 is inserted. A possible line (suction insertion line) 422 is obtained, and a point 423 (robot position / direction indication point) indicating the position / direction in which the robot cleaner inserts the suction part 132 as shown in FIG. 12B is calculated. From these points, the suction portion 132 is inserted to perform cleaning. The cleaning method registered in the attribute data 402 is the robot position / direction instruction point 423.
[0068]
As shown in FIG. 13a, the cleaning method is connected to the cleaning route 403 closest to the robot position / direction indicating point 423, and when the robot cleaner in operation reaches the connection point as shown in FIG. The direction is changed according to the direction indication point 423, and the inside of the obstacle 419 is cleaned by the suction part 132.
[0069]
Next, the processing contents of the (2) attribute data learning function will be described. FIG. 14 shows a flowchart of processing contents. (2) The attribute data learning function is executed every time the cleaning work is completed. First, the difference between the room map data 401 and the temporary room map data 501 is compared, and the obstacles that have moved or disappeared are confirmed against the obstacles in the previously created room map data 401 (step 801). .
[0070]
When performing this obstacle confirmation, it is necessary to consider the characteristics of the obstacle. For example, a leg of a chair or table constitutes one obstacle with a plurality of elements (obstacles). If this feature is not managed by the attribute data 402, an accurate obstacle confirmation cannot be performed. In this way, obstacles are grouped in order to search for one obstacle composed of a plurality of elements (step 802). FIG. 15 shows a specific example thereof.
[0071]
15, FIG. 15a shows room map data 401 generated by the previous cleaning operation, and FIG. 15b shows temporary room map data 501 generated by the current cleaning operation. When these two data are matched, the five obstacles shown in the group 424 of FIG. 15b are obstacles that do not match the data of FIG. 15a, so that the obstacle of FIG. 15a has moved or newly appears. It turns out that it is an obstacle.
[0072]
First, it is assumed that this group 424 constitutes one obstacle. Then, translation and rotation are performed while maintaining the relative coordinate values of the group 424, and a search is performed for an obstacle that matches the data in FIG. 15a.
[0073]
Subsequently, as shown in FIG. 15c, it is assumed that the group 424 is further divided and the group 425 and the group 426 constitute one obstacle, respectively, and the search is performed from the data of FIG. 15a. This grouping work is performed as long as the combination of obstacles allows. Then, by selecting the obstacle that matches the data of FIG. 15a and has the largest number of elements in the group, one obstacle consisting of a plurality of elements can be found (step 803).
[0074]
Walls are another characteristic obstacle. The shape of the wall changes each time the obstacle near the wall moves. Therefore, the obstacle cannot be confirmed accurately unless the wall and the obstacle at the wall are accurately distinguished. FIG. 16 shows such a wall shape search method.
[0075]
In FIG. 16, FIG. 16a shows the wall shape data 427 in the room map data 401 generated by the previous cleaning operation, and FIG. 16b shows the wall shape data 428 in the temporary room map data 501 generated by the current cleaning operation. ing. By integrating these two data, the shape having the maximum area is defined as the wall shape. By using the shape obtained from the difference between the wall shape and the wall shape data in the temporary room map data 501 of FIG. 16b as an obstacle, the wall 429 and the obstacle 430 can be distinguished as shown in FIG. Step 803).
[0076]
In consideration of such obstacle characteristics, the obstacle data learning function (2) is executed. If the obstacle search is completed, the process proceeds to step 804. As a result of the obstacle search in steps 801, 802, and 803, if there is an obstacle that exists only in the room map data 401 instead of the temporary room map data 501, This obstacle is regarded as a disappeared obstacle, and the attribute data of this obstacle is moved to the short-term memory attribute data 900 (step 805). Also, if there is an obstacle that exists only in the temporary room map data 501 but not in the room map data 401, this obstacle is a newly appearing obstacle. (4) Already by new registration of unregistered data The attribute data is registered in the attribute data 402.
[0077]
For obstacles not corresponding to step 804, the process proceeds to step 806 to narrow down the obstacles. In other words, if one obstacle in the temporary room map data 501 matches with a plurality of obstacles in the room map data 401 as a result of the obstacle search in steps 801, 802, and 803, the process proceeds to step 807. The matching obstacles are narrowed down to one by examining the existence history and knowledge of each obstacle based on the attribute data 402 corresponding to the plurality of obstacles. If one obstacle can be searched in this way, the position where the searched obstacle moves and the knowledge data calculated from the movement position are registered in the attribute data 402. By updating the attribute data 402 in this way, more accurate obstacle estimation and dynamic map updating can be performed in the next and subsequent sweep operations.
[0078]
A specific example of the attribute data 402 is shown in FIG. 17, and an example of the obstacle TYPE is shown in FIG. FIG. 17 is a diagram showing the attribute data 402 of three obstacles whose obstacle type is a wall, a fixed obstacle (chance), and a daily position change obstacle (chair foot), respectively. The attribute data 402 includes items of ID, integrated ID, obstacle TYPE, history area number, history area shape data, cleaning method, and knowledge data.
[0079]
Each item is explained as follows.
ID: Number to associate with each obstacle stored in the room map data
Integrated ID: For one obstacle (eg, chair foot) consisting of multiple elements (obstacles), each ID number associated with these elements
Obstacle TYPE: Indicates the type of obstacle. (See Figure 18)
Number of history areas, history area shape data: A summary of the number of areas and shape coordinate values where this obstacle existed in the past in room map data
Cleaning method: Memorize robot position / direction indication points when cleaning inside obstacles
Knowledge data: Describes knowledge and conditions specific to obstacles.
[0080]
In FIG. 17, the wall is ID 1, the fixed obstacle (chance) is ID 2, and the daily position change obstacle is ID 7 that integrates chair feet IDs 3, 4, 5, and 6
[0081]
Also, (2) specific examples of the attribute data learning function are shown in FIGS. FIG. 19 shows an example of updating the room map data 401, and FIG. 20 shows an example of updating the attribute data 402. As shown in FIGS. 20 and 21, in the first cleaning, all obstacles (ID1 walls, ID2 and 3 obstacles, ID4 obstacles) are fixed obstacles. In the second cleaning, data is updated by analyzing how the obstacles have changed (moved or disappeared) by the search using the grouping of obstacles described above.
[0082]
Here, the obstacles with IDs 2 and 3 are determined as one obstacle, and ID 5 is assigned. In addition, since the obstacles of IDs 2, 3, and 4 have changed positions, the obstacles TYPE become position-changing obstacles every day, and the areas 6 and 7 where the obstacles existed before (hatched portions in FIG. 19). ) As a history area.
[0083]
As for the cleaning method, the obstacles are not grouped in the first cleaning, so there is no obstacle inside, and the cleaning method is invalid, but the obstacles with ID4, 5 in the second cleaning are Since the inside is included, a robot position / direction indicating point as a cleaning method is calculated and registered in the attribute data 402. Furthermore, knowledge data determined from the state of the obstacle is stored.
[0084]
Thus, in the first data update, since all obstacles are single fixed obstacles, it takes time to update the data. However, when the attribute data 402 is constructed by overlapping the number of data updates, the update time is increased. Shortened.
[0085]
As described above, according to the present embodiment, it is possible to update the room map during the cleaning and to reconstruct the cleaning route, and it is possible to complete the cleaning of the room whose situation has changed. In addition, since infrared sensors and tactile sensors are used to detect obstacles, the shape of the obstacles can be recognized with high accuracy, and the room map can be updated and the cleaning route can be reconstructed with high accuracy. It was. Furthermore, since the attribute data is provided for each obstacle, the shape of the obstacle can be recognized with high accuracy and at high speed.
[0086]
【The invention's effect】
According to the present invention, when an obstacle is encountered during cleaning, it becomes possible to reconstruct the cleaning route while dynamically updating the room map. A robot cleaner that performs cleaning can be realized. Further, by providing the infrared sensor and the tactile sensor, it is possible to recognize the shape of the obstacle with high accuracy and at high speed, and to realize a precise cleaning operation. Furthermore, the cost of the robot cleaner can be reduced.
[Brief description of the drawings]
FIG. 1 is a system configuration diagram for explaining an embodiment of an invention of a robot cleaner according to the present invention.
2 is a top view for explaining a robot unit of the robot cleaner shown in FIG. 1; FIG.
3 is a configuration diagram for explaining a relationship between a control unit of the robot unit shown in FIG. 2 and a sensor system / drive system.
4 is a flowchart for explaining processing contents of the robot unit shown in FIG. 2; FIG.
FIG. 5 is a flowchart for explaining processing contents executed by a computer according to the robot cleaner control program of the present invention;
6 is a flowchart for explaining an obstacle shape estimation function in the processing content shown in FIG. 5;
7 is a flowchart for explaining new registration processing of unregistered data in the processing content shown in FIG.
8 is a flowchart for explaining a cleaning route reconstruction function in the processing content shown in FIG.
FIG. 9 is a diagram for explaining an example of cleaning route reconstruction.
FIG. 10 is a flowchart for explaining region division when a cleaning route is reconstructed.
FIG. 11 is a diagram for explaining another example of cleaning route reconstruction.
FIG. 12 is a diagram for explaining an example of a method for cleaning the inside of an obstacle.
FIG. 13 is a diagram for explaining an example of a cleaning route in which attribute data (cleaning method) is integrated.
14 is a flowchart for explaining an attribute data learning function in the processing content shown in FIG.
FIG. 15 is a diagram for explaining an example of grouping in obstacle search;
FIG. 16 is a diagram for explaining an example of a method for identifying a wall and an obstacle.
FIG. 17 is a diagram for explaining an example of attribute data.
FIG. 18 is a diagram for explaining a classification example of an obstacle TYPE.
FIG. 19 is a diagram for explaining an example of an obstacle attribute learning function in room map data.
FIG. 20 is a diagram for explaining an example of an obstacle attribute learning function in attribute data.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 100 ... Robot unit, 101 ... Robot unit main body, 110 ... Sensor data acquisition part, 111-113 ... Infrared sensor, 114 ... Tactile sensor, 126, 127 ... Encoder, 131 ... Arm part, 132 ... Suction part, 150, 225 ... Wireless communication module, 160 ... Drive unit, 200 ... Intelligent processing unit, 221 ... CPU, 222 ... Storage device, 224 ... Internal bus, 235 ... Robot cleaner control program, 300 ... Control unit, 400 ... Database, 401 ... Room map Data, 402 ... attribute data, 403 ... cleaning route data, 500 ... memory.

Claims (8)

A control program for controlling the operation of the robot cleaner by a computer,
The robot cleaner includes an obstacle detection sensor for detecting the presence or absence of an obstacle and recognizing the shape of the obstacle, room map data storing a room map formed by registering obstacles in a room, Attribute data storing the attribute of the obstacle registered in the room map data, and cleaning route data storing a cleaning route created based on the room map data as a cleaning route,
The control program is stored in the computer.
A procedure for detecting the presence or absence of an obstacle using the obstacle detection sensor when cleaning is performed according to the cleaning route;
A procedure for determining whether an obstacle is registered on the room map data at the position when an obstacle is detected;
A procedure for judging the detected obstacle as an unknown obstacle when no obstacle is registered;
Recognizing the shape of the unknown obstacle using an obstacle detection sensor;
A procedure for estimating a shape that matches a recognized shape with reference to the room map data and the attribute data;
If the shape can be estimated, register the obstacle of the estimated shape in the room map data and rebuild the cleaning route data,
If the shape cannot be estimated, the unknown obstacle is judged as a newly appearing obstacle, registered in the room map data, and a procedure for reconstructing the cleaning route data is executed. Robot cleaner control program to do.   The robot cleaner control program according to claim 1, wherein the obstacle detection sensor includes an infrared sensor and a tactile sensor. The room map data and the cleaning route data have temporary room map data and temporary cleaning route data, respectively.
The procedure for registering the estimated shape obstacle in the room map data and reconstructing the cleaning route data registers the estimated shape obstacle in the temporary room map data and reconstructing the temporary cleaning route data. The temporary room map data and the temporary cleaning route data are overwritten on the room map data and the cleaning route data, respectively, at the end of cleaning,
The procedure of determining the unknown obstacle as a newly appearing obstacle, registering it in the room map data, and reconstructing the cleaning route data determines that the unknown obstacle is a newly appearing obstacle, and Appearing obstacles are registered in the temporary room map data, and the temporary cleaning route data is re-registered. At the end of cleaning, the temporary room map data and the temporary cleaning route data are respectively stored in the room map data and the cleaning route data. The robot cleaner control program according to claim 1, wherein the program is an overwriting procedure.   4. The method according to claim 3, further comprising a step of learning and updating the attribute data by comparing the difference between the room map data before overwriting and the temporary room map data at the end of cleaning. Robot cleaner control program.   The obstacle cleaning method stored in the attribute data is used in the procedure of registering the estimated shape obstacle in room map data and reconstructing the cleaning route data. Robot cleaner control program. An obstacle detection sensor for detecting the presence or absence of an obstacle and recognizing the shape of the obstacle;
Room map data storing a room map formed by registering obstacles in a room, attribute data storing attributes of obstacles registered in the room map data, and cleaning created based on the room map data A storage device for storing cleaning route data in which the route is stored as a cleaning route;
A processing device for controlling cleaning,
The processor is
Detecting the presence or absence of obstacles using the obstacle detection sensor when cleaning is performed according to the cleaning route,
When an obstacle is detected, it is determined whether an obstacle is registered on the room map data at that position,
If no obstacle is registered, the detected obstacle is determined as an unknown obstacle,
Recognize the shape of the unknown obstacle using an obstacle detection sensor,
A shape that matches the recognized shape is estimated with reference to the room map data and the attribute data,
If the shape can be estimated, register the obstacle of the estimated shape in the room map data, rebuild the cleaning route data,
A robot cleaner characterized in that, when the shape cannot be estimated, the unknown obstacle is judged as a newly appearing obstacle, registered in the room map data, and the cleaning route data is reconstructed.   The robot cleaner according to claim 6, wherein the obstacle detection sensor includes an infrared sensor and a tactile sensor.   The robot cleaner according to claim 7, wherein the infrared sensor is provided in a robot unit having wheels for moving, and the tactile sensor is provided in a suction portion that performs a cleaning operation.

Images