JP2024015640A - Multiple mobile robot control method, program, and multiple mobile robot control system - Google Patents

Abstract

The present invention provides a method for controlling multiple mobile robots and a system for controlling multiple mobile robots, which can improve search efficiency compared to conventional methods.
When the current position of the first robot 100 is unknown, the past movement position trajectory of the first robot 100, the estimated error range 307 in estimating the past position of the first robot 100, and the movement amount Based on this, a first search range for the first robot 100 is determined, a search command for the first search range is sent to a second robot 200 different from the first robot 100, and the first search range is determined based on the first search range. sensor information from the second robot 200 is received, and the current position of the first robot 100 is estimated.
[Selection diagram] Figure 2

Description

The present invention relates to a method for controlling multiple mobile robots and a system for controlling multiple mobile robots.

Patent Document 1 discloses that a management server maintains robot information of an autonomous mobile robot, periodically acquires position information from the autonomous mobile robot, and sends subordinate robots to help when a rescue notification is received from the autonomous mobile robot. The rescue robot that receives the rescue instruction searches for the robot in question, calculates the absolute position and direction of the rescued robot from its relative position to the rescued robot, and recovers the rescued robot. There is.

Japanese Patent Application Publication No. 2010-3240

Currently, multiple autonomous mobile robots are being used in factories, logistics centers, office buildings, etc. to transport goods, guide people, and provide security. Such autonomous mobile robots generally use laser sensors or cameras to estimate their own position.

An autonomous mobile robot using SLAM (Simultaneous Localization and Mapping) technology estimates its own position by sensing feature points in the surrounding environment, such as the shape of a wall. However, in a space with few features or in an environment where features change due to changes in the arrangement of surrounding objects, position estimation may fail and movement may become impossible.

Patent Document 1 has been proposed as an example of a recovery method for a failure in position estimation. In Patent Document 1, the management server side issues a search instruction to an autonomous mobile robot to perform the search, but there is room for improvement in the method of determining the area to be searched by the search robot, and in particular, there is room for improvement in search efficiency. It became clear that there was room for

The present invention provides a method for controlling multiple mobile robots and a system for controlling multiple mobile robots that can improve search efficiency compared to conventional methods.

The present invention includes a plurality of means for solving the above problems, and one example is a method for controlling a plurality of mobile robots equipped with sensors, in which the current position of the first robot is unknown. In this case, a first search range for the first robot is determined based on a past movement position trajectory of the first robot, error information in past position estimation of the first robot, and a movement amount. a step of transmitting a search command for the first search range to a second robot different from the first robot; and receiving sensor information from the second robot in the first search range. and estimating the current position of the first robot.

According to the present invention, since the search efficiency can be improved compared to the conventional method, it is possible to improve the work efficiency of an autonomous mobile robot. Problems, configurations, and effects other than those described above will be made clear by the description of the following examples.

FIG. 1 is a block diagram showing an example of the configuration of a multiple mobile robot control system according to an embodiment. 7 is a flowchart illustrating an example of position recovery control processing in the method for controlling multiple mobile robots according to the embodiment. FIG. 3 is a diagram illustrating an example of a method for determining a search range in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example of a method for determining a search range in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example of a method for determining a search range in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example of a method for determining a search range in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example of a method for determining a search range in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example of a method for determining a search range in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example of limiting a search range using obstacle sensor information in the multiple mobile robot control method of the embodiment. FIG. 6 is a diagram illustrating an example of a method for recognizing a first robot using a sensor of a second robot in a method for controlling multiple mobile robots according to an embodiment. FIG. 3 is a diagram illustrating an example in which sensor information received by a first robot in the multiple mobile robot control method of the embodiment is based on a fixed sensor.

FIG. 1 is a diagram illustrating an embodiment of a multiple mobile robot control method, a program, and a multiple mobile robot control system of the present invention for recovering position estimation using another autonomous mobile robot when an autonomous mobile robot fails in position estimation. This will be explained using FIGS. 1 to 11. In the drawings used in this specification, the same or corresponding components are given the same or similar symbols, and repeated description of these components may be omitted.

In a control method, program, or control system for controlling a plurality of autonomously movable first robots 100 and second robots 200 equipped with sensors, it is preferable to use the past movements of lost robots collected at regular intervals by a management server. The search range is determined using the trajectory, error information, and movement amount, the second robot 200 closest to the robot is selected as the rescue robot, and an object that appears to be a lost robot is detected from the sensor information of the rescue robot, and an area with a high probability of existence is determined. Find and recognize lost robots quickly by limiting the search range to It is possible to recover from a position estimation failure by transmitting the recognized absolute position of the lost robot to the lost robot and resetting the position.

In the following example, an in-factory article transport robot that travels in an indoor flat environment will be taken as an example. An example will be described in which, of the two mobile robots shown in FIG. 1, the first robot 100 is the robot whose position estimation has failed, and the second robot 200 is used as the rescue robot.

The present embodiment will be described below in the order of configuration and processing flow.

<Configuration>
FIG. 1 is a configuration diagram of a system according to this embodiment. Hereinafter, an overview of each of these devices will be explained first, and the details will be described later when explaining the processing flow.

A first robot 100 shown in FIG. 1 is equipped with a sensor 101 that acquires map information.

The information processing device 109 of the first robot 100 includes a sensor processing unit 105 that processes map information acquired by the sensor 101, a self-position calculation unit 106 that calculates the self-position based on the sensor 101 information and shared information, and a first robot 100. a position information acquisition unit 107 that acquires the self-position based on information such as a wheel encoder built into the robot 100, and a position estimation failure determination unit 104 that determines the failure of position estimation from the information of the sensor 101 and the self-position calculation unit. , has a travel control section 108 that controls the travel of the first robot 100. Each of these parts is configured by a program executed by the central processing unit 102 of the first robot 100.

The information transmitting/receiving unit 103 of the first robot 100 is connected to the network 8 in order to exchange map information with the management server 1.

The second robot 200 is equipped with a sensor 201 that acquires map information.

The information processing device 209 of the second robot 200 includes a sensor processing unit 205 that processes map information acquired by the sensor 201, a self-position calculation unit 206 that calculates the self-position based on the sensor 201 information and shared information, and a second robot 200. a position information acquisition unit 207 that acquires the self-position based on information such as a wheel encoder built into the robot 200, and a position estimation failure determination unit 204 that determines the failure of position estimation from the information of the sensor 201 and the self-position calculation unit. , has a travel control section 208 that controls the travel of the second robot 200. Each of these parts is configured by a program executed by the central processing unit 202 of the second robot 200.

The information transmitting/receiving unit 203 of the second robot 200 is connected to the network 8 in order to exchange map information with the management server 1.

The management server 1 has an information processing device 7. The information processing device 7 of the management server 1 includes an information transmitting/receiving section 6, a main storage device 5, a memory 4, and a map information display section 3.
It comprises a first search range determining section, a search command transmitting section, and an estimating section. The memory 4 includes a map information storage section 41, a position range calculation section 42, a search command management section 43, and a coordinate transformation processing section 44, which are executed by a central processing unit (CPU) 2 of the management server 1. This is a program to do this.

Map information exchanged between the network and the information transmitting/receiving section 6 of the management server 1 is written into the main storage device 5 by the map information holding section 41 . The main storage device 5 is a storage medium, and may be an HDD, SSD, or Flash memory. The network 8 may be configured by a wireless LAN, and the map information to be exchanged may use socket communication using TCP/IP.

Note that information communication regarding map information may be performed directly with other robots connected to the network, and there may be two or more robots connected to the network.

The sensor 101 of the first robot 100 and the sensor 201 of the second robot 200 are, for example, at least one of LiDAR, RGB-D camera, monocular camera, stereo camera, omnidirectional camera, infrared camera, GPS, etc. The configuration described above can be a combination of different types of sensors. It is assumed that the first robot 100 and the second robot 200 can perform highly accurate position estimation using these sensors 101 and 201.

Although the second robot 200 has the same configuration as the first robot 100 in this embodiment, it may have a different configuration. For example, this also applies to communication between a crawler-equipped mobile robot and a humanoid service robot.

Further, the management server 1 may also include a physical server for realizing its functions, or programs may be stored inside each robot, and the functions may be implemented by the first robot 100 and the second robot. It may also be executed in an autonomous and decentralized manner on the central processing units 102 and 202 mounted on the robot 200.

When each function of the management server 1 is inside the first robot 100 and the second robot 200, the ad-hoc network construction between the first robot 100 and the second robot 200 allows nearby robots to communicate with each other. Since the system can be completed by simply exchanging information, it can be operated without installing wireless network infrastructure in the equipment. Furthermore, it is possible to reduce the communication load by reducing the amount of communication information.

On the other hand, when the management server 1 has an independent configuration, effects such as simplification of the configurations of the first robot 100 and the second robot 200 can be obtained.

The map information to be shared is configured such that the map information display section 3 of the management server 1 allows the user to check it. Here, the map information sent and received includes the acquisition time, robot ID, robot appearance information, self-position coordinates, estimation error in self-position estimation, running movement variable by odometry, recognition ID of other robots, recognition reliability, and own robot. It includes relative position information, lost detection flag, sensor type, and obstacle map around the robot acquired by the sensor.

The contents of the map information vary depending on the environment and scene in which the robot runs. For example, in the case of a robot that moves on a well-maintained indoor plane in an office, warehouse, factory, etc., the self-position coordinates of the map information are three-dimensional self-position coordinates (x, y, θ). On the other hand, in an undeveloped indoor environment with falling objects or uneven ground, or outdoors, the self-position coordinates of the above map information are 6-dimensional self-position coordinates (x, y, z, roll, pitch, yaw). It will be done.

In addition, the estimation error of the above map information is, for example, the range in which a robot whose position estimation has failed is estimated to have moved, and uses the error ellipse obtained in the Monte Carlo position estimation calculation or the Kalman filter position calculation process. be able to.

This concludes the explanation of the system configuration of this embodiment. Next, the processing flow of this embodiment will be explained with reference to the information used therein.

<Processing flow>
The processing flow of this embodiment will be described below. FIG. 2 is a flowchart showing operation control processing for recovery from a failure in position estimation in this embodiment. It is preferably processed by the information processing device 7 of the management server 1 in FIG.

In addition, in this embodiment, the plurality of first robots 100 and second robots 200 use map information (their own past movement position locus, estimated error range 307 in past position estimation (Fig. reference) and movement amount) are preferably output to the management server 1 at regular intervals.

In FIG. 2, when the first robot 100 detects a failure in position estimation as in step S200, the process moves to step S201.

The position estimation failure determination unit 104 of the first robot 100 detects the position estimation failure in step S200. If the estimation error obtained by the self-position calculation unit 106 exceeds a preset threshold, it is determined that the position estimation has failed, a notification is sent to the management server 1, and the vehicle stops running. In addition to this, if the position estimation fails due to an error in calculation processing, it is also determined that the position estimation has failed.

Furthermore, when estimating the robot's self-position, the estimation result may converge to an incorrect position due to incorrect matching of feature points. If the estimated error does not represent the actual error, the method of determining lost using only estimated error information has the problem of erroneous detection. Therefore, it is desirable to have a method of determining a failure in position estimation by comparing estimation results obtained by different types of position estimation methods.

Next, in step S201, the management server 1 shares sensor information obtained from other robots other than the first robot 100, and uses the absolute position of the first robot 100 observed by the sensors of the other robots. An attempt is made to recover the position of the first robot 100, and in step S202 it is determined whether or not the position can be recovered using only the shared information. If the position can be recovered because the first robot 100 exists within the field of view of another robot's sensor, the process is completed.

On the other hand, if it is determined that position recovery is not possible due to reasons such as the first robot 100 not being within the field of view of the sensor of another robot, the second robot 100 is moved to a position where the lost robot can be observed in order to recover the position. robot 200 needs to move. This operation is called a search, and steps after step S203 represent the search operation. Since the estimation error in the lost state is often inaccurate, in step S203, a physically possible range is set as a search range (first search range).

The method for determining the search range will be described in detail below. 3 to 8 are diagrams showing an example of the flow of determining a search range and selecting a rescue robot.

First, as shown in FIG. 3, movement history information of the lost robot is used to determine the range in which the lost robot can physically exist. Since the estimation error is small at points where the position has been estimated with high accuracy, the estimation error history 302 is traced back from the current time, and a point where the estimation error 301 is less than a reliable threshold is set as a search reference point.

Furthermore, if the position of the lost robot at that time is the search center coordinate 304, the lost robot exists in an area 305 consisting of a circle whose radius is r and the total movement distance 303 traveled by odometry from that point to the current time. It can be said that it is.

Furthermore, an area with a high search priority is determined based on the estimated error range of the lost robot at the current time and the direction between the search center coordinates and the estimated position at the current time. To estimate the amount of movement, we use the amount of movement determined by wheel odometry, assuming that the robot is a mobile robot that is unlikely to be lifted by a person, such as a transport robot.

As shown in FIG. 4, the search range in the real space includes an area 305 where a lost robot may physically exist as an area S ₁ , an area 306 where there is a higher probability of a lost robot existing as an area S 2 , and an area 306 where a lost robot is likely to exist as an area S ₂ . The error range 307 is defined as a region S ₃ and the search range S _a 311 (see FIG. 7) in equation (1).

Here, _SM is a driveable area 310 (see FIG. 6) extracted from the map shape. The definitions of regions S ₁ , S ₂ , and S ₃ are shown below. The reliable position used as the search center coordinates 304 is the result of position estimation (x _t1 , y _t1 , θ _t1 ) at time t ₁ , which is the first point at which the estimation error falls below the threshold value, tracing back the error history from the current time t. The radius r, which is the total distance traveled 303 by odometry from the trusted position to the current position, is determined by equation (2).

An area 305 consisting of a circle with a radius r centered on the trusted position (x _t1 , y _t1 ) is an area S ₁ in which the robot can physically exist. Further, within the search range, a fan-shaped area 306 that is open by an angle φ and centered on a vector connecting the trusted position and the current position is set as a high-priority area _S2 , as an area with a high possibility of existence. Similarly, the range that overlaps with the estimated error range of the robot's position at the current time obtained by the position estimation calculation is set as a high-priority area _S3 307.

Since the range to be searched is the product set of the union set of these set areas S ₁ , S ₂ , S ₃ and the driveable area S _M 310, the search range is the search range S _a of formula (1). 311 is given. The search range S _a 311 is illustrated in FIG. 7 . In addition, by storing the map coordinates where position estimation failures occurred in the past, the management server 1 creates areas on the map where it is determined that lost locations are likely to occur, and includes them in area _S2 as candidates for the search range. May be used.

Next, "points of interest" 308 and 309 as shown in FIG. 5 are set as autonomous movement destination coordinates for causing the rescue robot to search the search range area obtained by equation (1). Points of interest 308 and 309 are discretely arranged at equal intervals within the search range area. The arrangement interval of points of interest is set based on the sensor's practical recognition distance so that the search range can be sufficiently sensed by the sensor mounted on the rescue robot.

In this embodiment, the distance at which the camera can recognize the other robot with high accuracy is defined as the distance between points of interest d, and the robots are arranged so as to fill the search range circle.

The priority of the point of interest is determined based on the overlap of the search ranges S. The points of interest 308 in areas _{S 2} and S ₃ of the search range Sa are searched first as the first priority, and the points of interest 309 in the area S ₁ of the search range S _a are searched as the second priority. do.

Next, in step S204, a rescue robot to search the search range determined in the above procedure is determined. The determination method will be explained in detail below.

At the current time t, the robot located at the coordinates closest to the center coordinates of the search range is extracted by the nearest neighbor method and assigned the role of a rescue robot. In this example, the Euclidean distance from the search center coordinates is used to select the nearest rescue robot, but if the space is not open, the nearest point is not the shortest in terms of actual travel distance, so a graph search algorithm etc. is used. It is desirable to assign tasks to the robot with the shortest travel route obtained through route planning.

Furthermore, in order to avoid affecting the transportation tasks of other robots, it is desirable to select a rescue robot based on the availability of tasks and the travel route.

Subsequently, in step S205, as shown in FIG. 8, the coordinates of the search target initially set for the rescue robot are the points of interest 308 located closest to the coordinates of the rescue robot, and the points of interest 312 of the first priority. and is determined by the nearest neighbor method. After receiving the coordinates, the rescue robot generates a travel route to the point of interest and moves autonomously.

During this autonomous movement search, the rescue robot continues to determine whether the lost robot has been successfully recognized based on the sensing results, as in step S206.

When the first robot 100 enters the field of view of the sensor due to sensing, the process transitions from step S206 to step S210, where the absolute position of the first robot 100 is corrected and recovered from the lost position.

On the other hand, if the first robot 100 cannot be recognized within the sensor field of view in step S206, the process moves to step S207, where it is determined whether the entire search range (second search range) has been searched, and step S208 The search range limiting process using the obstacle sensor information is executed until the search range is completed. The method for limiting the search range using the obstacle sensor information in step S208 will be described in detail below.

FIG. 9 is an example of limiting points of interest using obstacle sensor information of a rescue robot. As shown in Figure 9, in order for the rescue robot to discover the lost robot and recognize its position and orientation with high accuracy during the search operation, it is necessary to physically approach the lost robot so that the sensor can demonstrate sufficient recognition accuracy. There is.

Therefore, due to these limitations on approach distance and sensor viewing angle, it is inefficient to observe points of interest one by one. Therefore, an obstacle sensor with a wide observable area such as LiDAR is used to extract obstacles that were not recorded in the map shape initially obtained by SLAM, and the area occupied by the obstacle is determined by the appearance of the robot to be searched. Objects that are close to the occupied area of the shape (template) are recognized as ``objects that seem to be the robot to be searched,'' and points of interest are left only around them. The search range is limited by deleting points of interest in map areas where unoccupied areas have been confirmed based on the obstacle sensor information.

Specifically, new obstacles 401 that were not present at the time of map creation are extracted from the obstacle occupancy map 400 generated by the sensor of the rescue robot by taking the difference between the known map shapes 402. Furthermore, among the extracted unknown obstacles, an obstacle 110 having an occupied area close to the size of the robot to be searched is extracted and set as a candidate to be searched. Only the points of interest around the extracted robot-like unknown object 403 are left, and the points of interest in other unoccupied areas in the obstacle map are deleted.

In addition, pattern matching using point clouds and image recognition using deep learning can be used to extract objects with a high degree of similarity to the robot's appearance, and processes such as edge extraction using line Hough transform from point clouds can be performed to identify unknown objects. You can also get .

In this way, it is desirable that the sensor information from the second robot 200 in the second search range be the position information of the first robot 100 recognized by the second robot 200. Further, it is desirable that this positional information be information estimated by template matching using appearance information of the first robot 100, point cloud matching, AR marker image recognition, or recognition using deep learning.

Next, step S209 will be explained. If the target robot cannot be found even after searching the entire search range generated in step S203, the process moves from step S207 to step S209. Specifically, in step S209, the radius r (total movement distance 303) of the search area 305 is expanded, and then the process returns to step S203 to continue the search.

When the search range is wide, the search range may be divided into several areas and searched by multiple robots.

Additionally, if the robot group does not succeed in the search within a certain period of time, a support notification can be sent to the administrator.

Next, step S206 will be explained using FIG. 10. FIG. 10 is an example of recognizing the first robot 100 using the sensor results of the second robot 200.

Recognition of the first robot 100 includes recognition using deep learning using robot appearance information, LiDAR point cloud matching 501, pattern matching of image features, image recognition of AR markers 500, sound, light, and beacon recognition. You may use any of the recognition methods using These recognition methods require relative positions and orientations. Further, the recognition processing using these sensors may be performed by the sensor processing units 105 and 205 within each robot, or may be performed by the management server 1.

Furthermore, the sensor that recognizes the position of the first robot 100 may be a fixed sensor such as a surveillance camera as shown in FIG. 11, as long as the position of the sensor is known.

Finally, step S210 will be explained. In step S210, an absolute position is calculated by coordinate transformation from the relative position of the recognized rescue robot and the first robot 100, and is transmitted to the first robot 100. Taking AR marker recognition as an example, the position of the rescue robot recognized by the robot is converted into an absolute position using the following coordinate transformation formula, and then transmitted to the robot.

The coordinate transformation is performed by the coordinate transformation processing unit 44 of the management server 1. The coordinate transformation to calculate the absolute position p _A = (x, y, θ) of the first robot 100 using the map origin as the map uses a homogeneous transformation matrix T∈R ^(4×4) including translation and rotation. can be expressed as equations (3) and (4). B_sensor is the mounting position of the robot sensor, B_marker is the marker installation position, and B_detect_marker is the recognized marker position.

The observation from the rescue robot (B) to the first robot 100 (A) is defined as the following equation (3).

Furthermore, in addition to observation from the rescue robot, when the rescue robot can be observed by the sensor of the first robot 100, the absolute position can be calculated using the following equation (4).

In the observation from the first robot 100 (A) to the rescue robot (B),

The absolute positions obtained in these ways are transmitted to the first robot 100 by the management server 1. The first robot 100 stochastically fuses a plurality of received absolute positions, resets its own position to a plausible position, and recovers from a position estimation failure.

Although this embodiment has been described using two mobile robots, which are the minimum configuration capable of implementing the invention, if there are three or more robots, sensor information from a third robot is also used. For example, equations (3) and (4) hold true even if B is replaced by a third robot. Furthermore, depending on the positional relationship of the robots, the third robot can also serve as an exploration robot.

In this embodiment, it is assumed that the rescue robot can always estimate the position with high accuracy, but in reality, if the rescue robot enters a space with few features, the accuracy of the rescue robot's position estimation may deteriorate. Therefore, to prevent this, it is desirable to arrange a third robot that can observe the rescue robot.

In addition, rescue robots have different sensors and processing methods than lost robots, such as position estimation methods using GPS or ceiling features, so there is a possibility that rescue robots may also fail in position estimation during search. It gets lower.

Furthermore, the search range is assumed to be an area where position estimation failures are likely to occur. Therefore, by setting the travel route of the third robot excluding the first search range and the second search range and planning the route so that other robots do not approach as much as possible, it is possible to reduce the loss occurrence rate. In this case, even after the search is completed, the area in which the lost robot whose position estimation has failed can be set as a range from which the lost robot should not approach as much as possible.

This concludes the explanation of the processing flow of this embodiment.

Note that the present invention is not limited to this embodiment, and includes various modifications.

For example, the present invention includes a method of controlling an autonomous mobile robot, but this control method may be performed by the mobile robot itself, or by another device, or may be performed by the mobile robot and another device. It may be done in collaboration.

The present invention also includes a computer program for executing the mobile robot control method and a medium for storing the program.

Next, the effects of this embodiment will be explained.

The method of controlling a plurality of robots according to the present embodiment described above is a method of controlling a plurality of first robots 100 and second robots 200 equipped with sensors 101 and 201, and the current position of the first robot 100 is unknown. In this case, the first search range for the first robot 100 is determined based on the past movement position trajectory of the first robot 100, the estimated error range 307 in the past position estimation of the first robot 100, and the movement amount. determining, transmitting a search command for the first search range to a second robot 200 different from the first robot 100, receiving sensor information from the second robot 200 in the first search range, The current position of the first robot 100 is estimated.

For example, in the step of estimating the current position of the first robot 100, sensor information from the second robot 200 in the first search range is received, an unknown object is extracted, and a second search range is set around the unknown object. , transmits a search command for a second search range to the second robot 200, receives sensor information from the second robot 200 in the second search range, and determines the current position of the first robot 100. Estimate.

According to the present invention, when the first robot 100 fails to estimate its own position and recovers the position estimation process by itself through search in cooperation with the second robot 200, the probability that a lost robot exists is determined. Since the search range can be limited at an early stage to areas with high values, search efficiency can be improved. Therefore, the position of the lost robot can be recovered more quickly than before without requiring human intervention.

Furthermore, since the second robot 200 is the second robot 200 closest to the first search range or the second search range, or the second robot 200 with available tasks, the search efficiency can be further improved. or the decline in the robot's work efficiency can be minimized.

Further, the sensor information received by the first robot 100 is information from a fixed sensor provided in the area where the second robot 200, the third robot, or the first robot 100 and the second robot 200 operate. By doing so, it is possible to obtain a large amount of information at the time of resetting the self-position and to make it highly accurate, so that it is possible to further improve the accuracy of recovery from a failure in position estimation.

Furthermore, if the first robot 100 cannot be found even after searching the second search range, the first search range can be expanded to avoid human intervention as much as possible.

Furthermore, the plurality of first robots 100 and second robots 200 have a step of outputting their past movement position trajectories, estimated error ranges 307 in past position estimation, and movement amounts to the management server 1. , information for determining a search range when a position estimation failure occurs can be grasped in advance, and a search range with a high possibility of discovery can be determined more quickly.

Furthermore, the plurality of first robots 100 and second robots 200 can perform the step of outputting the movement position locus, estimated error range 307, and movement amount to the management server 1 at regular intervals, so that the search range can be improved. This eliminates bias in the information necessary for determining the location of robots, and it becomes possible to determine a search range that contributes to finding lost robots whose position estimation has failed with a high probability.

Further, the sensor information from the second robot 200 in the second search range is the position information of the first robot 100 recognized by the second robot 200, and in particular, the position information is based on the appearance of the first robot 100. Since the information is estimated by template matching using information, point cloud matching, AR marker image recognition, or recognition by deep learning, position estimation can be performed reliably with a small amount of information.

Furthermore, each step is executed in an autonomous decentralized manner on the independent central processing unit 2 or on the central processing units 102 and 202 mounted on the first robot 100 and the second robot 200, so that the movement It is possible to reduce the device configuration on the robot side or the management server and communication load.

In addition, by setting the travel route of the third robot excluding the first search range and the second search range, it is possible to suppress the occurrence of additional lost robots whose position estimation fails, and more reliably prevent a decrease in work efficiency. can be avoided.

<Others>
Note that the present invention is not limited to the above-described embodiments, and various modifications and applications are possible. The embodiments described above are described in detail to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to having all the configurations described.

1... Management server 2... Central processing unit 3... Map information display section 4... Memory 5... Main storage device 6... Information transmitting/receiving section 7... Information processing device (first search range determining section, search command transmitting section, estimating section)
8...Network 41...Map information holding section 42...Position range calculation section 43...Search command management section 44...Coordinate transformation processing section 100...First robot (mobile robot)
101, 201...Sensors 102, 202...Central processing unit 103, 203...Information transmitting/receiving unit 104, 204...Position estimation failure determination unit 105, 205...Sensor processing unit 106, 206...Self position calculation unit 107, 207...Position information acquisition Sections 108, 208... Travel control section 110... Obstacle 200... Second robot (mobile robot)
301...Estimation error 302...Estimation error history 303...Total movement distance 304...Search center coordinates 305, 306, 310, 311...Area 307...Estimation error range 308, 309, 312...Point of interest 400...Obstacle occupancy map 401...Obstacle Object 402...Map shape 403...Unknown object 500...AR marker 501...Point cloud matching

Claims (13)

A method for controlling multiple mobile robots equipped with sensors, the method comprising:
When the current position of the first robot is unknown, based on the past movement position trajectory of the first robot, error information in the past position estimation of the first robot, and the amount of movement, determining a first search range for the robot;
transmitting a search command for the first search range to a second robot different from the first robot;
A method for controlling multiple mobile robots, comprising the steps of: receiving sensor information from the second robot in the first search range and estimating the current position of the first robot. The method for controlling multiple mobile robots according to claim 1,
In the step of estimating the current position of the first robot,
Receive sensor information from the second robot in the first search range, extract an unknown object, limit a second search range to the vicinity of the unknown object, and send the second robot to the second robot. 2. Send a search command for the search range,
A method for controlling multiple mobile robots, comprising: receiving sensor information from the second robot in the second search range, and estimating the current position of the first robot. The method for controlling multiple mobile robots according to claim 2,
A method for controlling a plurality of mobile robots, wherein the mobile robot closest to the first search range or the second search range or the mobile robot with a vacant task is used as the second robot. The method for controlling multiple mobile robots according to claim 2,
The sensor information received by the first robot is information from a fixed sensor provided in an area where the second robot, the third robot, or the mobile robot operates. The method for controlling multiple mobile robots according to claim 2,
If the first robot cannot be found even after searching the second search range, the first search range is expanded. The method for controlling multiple mobile robots according to claim 2,
A method for controlling a plurality of mobile robots, comprising the step of the plurality of mobile robots outputting their past movement position trajectories, error information in past position estimation, and movement amounts to a management server. The method for controlling multiple mobile robots according to claim 6,
A method for controlling a plurality of mobile robots, wherein the plurality of mobile robots execute a step of outputting the movement position locus, the error information, and the movement amount to the management server at regular intervals. The method for controlling multiple mobile robots according to claim 2,
The sensor information from the second robot in the second search range is position information of the first robot recognized by the second robot. The method for controlling multiple mobile robots according to claim 8,
The position information is information estimated by template matching using appearance information of the first robot, point cloud matching, AR marker image recognition, or recognition by deep learning. A method for controlling multiple mobile robots. The method for controlling multiple mobile robots according to claim 2,
Each of the steps is executed in an autonomous and decentralized manner on an independent computer or on a computer mounted on the mobile robot. The method for controlling multiple mobile robots according to claim 2,
A method for controlling a plurality of mobile robots, wherein a traveling route of a third robot is set excluding the first search range and the second search range. A computer that controls multiple mobile robots equipped with sensors,
When the current position of the first robot is unknown, based on the past movement position trajectory of the first robot, error information in the past position estimation of the first robot, and the amount of movement, determining a first search range for the robot;
transmitting a search command for the first search range to a second robot different from the first robot;
A program comprising: receiving sensor information from the second robot in the first search range and estimating the current position of the first robot. A system for controlling multiple mobile robots equipped with sensors,
When the current position of the first robot is unknown, based on the past movement position trajectory of the first robot, error information in the past position estimation of the first robot, and the amount of movement, a first search range determination unit that determines a first search range for the robot;
a search command transmitter that transmits a search command for the first search range to a second robot different from the first robot;
A multiple mobile robot control system, comprising: an estimation unit that receives sensor information from the second robot in the first search range and estimates the current position of the first robot.

Images