JP4809690B2 - Autonomous mobile device and planar obstacle recognition method - Google Patents

Description

  The present invention relates to an autonomous mobile device, and more particularly to a planar obstacle recognition method mounted on an autonomous mobile device and suitable for autonomous movement.

Conventionally, a laser radar is used to scan a horizontal plane at a fixed angle or a fixed interval, receive a reflected wave from an object, obtain a group of scan points distributed in a two-dimensional plane, and from these scan points Select scan points under specific conditions and connect the selected scan points to form a set of line segments. The set of line segments recognizes the surface position of the wall (planar obstacle) distributed in space. There is a known method (see Non-Patent Document 1, for example).
Li Zhang & Bijoy K.K. Ghosh, “Line Segment Based Map Building and Localization Using 2D Laser Rangefinder”, IEEE Int. Conf. On Robotics & Automation, pp. 2538-2543,2000

  However, in the conventional planar obstacle recognition method, when a specific scan point is selected from a large number of scan points to form a set of line segments, for example, the distance from a scan point to a line segment The calculation process is complicated, and the calculation process is complicated, and the calculation process takes time. Therefore, when it is installed as an obstacle recognition method for an autonomous mobile device, there arises a problem that the moving speed of the autonomous mobile device has to be slowed down for safe and reliable autonomous movement. For this reason, a planar obstacle recognition method using a simpler procedure is required for mounting on an autonomous mobile device. Moreover, in the conventional planar obstacle recognition method, in the case of a planar obstacle with a gap such as a fence made of a net or a lattice, a reflected wave from an object existing behind the planar obstacle is received. Therefore, it is difficult to accurately recognize a planar obstacle having a gap. Therefore, a method capable of recognizing a planar obstacle having a gap has not been put to practical use as a planar obstacle recognition method suitable for an autonomous mobile device.

  The present invention has been made to solve the above-described problems, and can accurately recognize the position of a planar obstacle as an obstacle, in particular, a planar obstacle having a gap, by a simple procedure. An object of the present invention is to provide an autonomous mobile device capable of autonomous and safe autonomous traveling and a planar obstacle recognition method suitable for it.

In order to achieve the above object, an autonomous mobile device according to claim 1 is provided by:
Scanning a given space using electromagnetic waves or sound waves at a certain angle or interval, receiving reflected waves from objects in the space, and obtaining the coordinates of multiple scan points that reflected the electromagnetic waves or sound waves Point acquisition means;
Of the plurality of scan points, an element vector having the previous scan point as the start point and the subsequent scan point as the end point is formed with respect to two scan points whose order obtained by the scan point acquisition unit is before and after in time series Element vector forming means to perform,
Among a plurality of element vectors, the order obtained by the element vector forming means is a plurality of element vectors that are continuous in time series and are continuous, and each element vector has a length equal to or less than a first predetermined length. Element vector selection means for selecting one element vector having a deflection angle formed by other element vectors continuous to the first predetermined angle or less and an integrated value of the deflection angles being a second predetermined angle or less. When,
A scan segment vector forming means for combining a plurality of selected consecutive element vectors with each other to form one scan segment vector;
A line segment defined by the start point and end point of the formed scan segment vector is compared with a second predetermined length, and when the line segment is equal to or longer than the second predetermined length, a planar obstacle along the scan segment vector comprising a recognizing obstacle recognition means but is present,
The element vector forming means uses an arbitrary scan point as a first scan point, and an element vector formed with a scan point continuous in time series as the second scan point is longer than the first predetermined length. In some cases, the element vector is formed using another scan point that is not continuous in time series with the first scan point as a new second scan point .

  According to a second aspect of the present invention, in the autonomous mobile device according to the first aspect, the element vector forming means forms an element vector using only scan points within a predetermined distance from an arbitrary origin.

According to a third aspect of the present invention, in the autonomous mobile device according to the first or second aspect , even if the element vector forming unit forms the element vector a predetermined number of times for the arbitrary scan point, the first predetermined When an element vector having a length less than or equal to the length cannot be obtained, an element vector is formed using a scan point that is continuous in time series with the arbitrary scan point as a new first scan point.

The invention of claim 4 is the autonomous mobile device according to claim 1 or 2,
Element vector formation unit, the element vectors forming the scanning point continuous time series in the second scan points as the third scanning points,
An element vector having a second scan point as a start point and a third scan point as an end point is larger than a first predetermined angle with respect to an element vector having the first scan point as a start point and the second scan point as an end point. In this case, the third scan point is excluded, and an element vector is formed using another scan point that is not continuous in time series with the second scan point as a new third scan point.

According to a fifth aspect of the present invention, in the autonomous mobile device according to the first aspect, when one scan segment vector is formed by the scan segment vector forming means, it is used for forming a scan segment vector among a plurality of scan points. It is characterized by trying to form a new scan segment vector using the scan points that did not exist.

According to a sixth aspect of the present invention, in the autonomous mobile device according to the fifth aspect , when two or more scan segment vectors are formed, an angle formed by the second scan segment vector with respect to the other first scan segment vectors is When the angle is equal to or smaller than the first predetermined angle and the start point of the second scan segment vector is within a predetermined distance from the end point of the first scan segment vector, the scan segment vector forming means starts the end point of the first scan segment vector. A vector having the start point of the second scan segment vector as an end point is synthesized between the first scan segment vector and the second scan segment vector to form a new scan segment vector.

The invention according to claim 7 is the autonomous mobile device according to claim 1,
The scan segment vector forming means stores information on scan points that contributed to the formation of the scan segment vector,
When the length of the line segment defined by the start point and the end point of the scan segment vector is equal to or longer than the second predetermined length, the obstacle recognition means uses the stored scan point coordinates by the least square method, A line segment that optimally fits the scan point is calculated, and this line segment is recognized as a planar obstacle.

The invention according to claim 8 is the autonomous mobile device according to any one of claims 1 to 7 ,
Storage means for storing a map segment that is a set of line segments of planar obstacles existing on a map of an operation area of the autonomous mobile device;
It functions as a segment verification unit that compares the scan segment with the map segment, and further includes a self-position specifying unit that specifies the position of the planar obstacle on the map and the self-position of the autonomous mobile device on the map It is characterized by.

The invention according to claim 9 is the autonomous mobile device according to claim 8 , wherein the self-position specifying means includes the position of the planar obstacle on the map and the planar obstacle from the origin on the autonomous mobile device. The self-location on the map is specified based on the distance and direction.

A planar obstacle recognition method according to a tenth aspect of the present invention comprises:
Scanning a given space using electromagnetic waves or sound waves at a certain angle or interval, receiving reflected waves from objects in the space, and obtaining the coordinates of multiple scan points that reflected the electromagnetic waves or sound waves A point acquisition step;
Of the plurality of scan points, an element vector having the previous scan point as the start point and the subsequent scan point as the end point is formed for two scan points whose order obtained by the scan point acquisition step is before and after in time series An element vector forming step,
Among a plurality of element vectors, the order obtained by the element vector forming step is a plurality of element vectors that are continuous in time series and are continuous, and each element vector has a length equal to or less than a first predetermined length. An element vector selection step of selecting one element vector having a deflection angle formed by another element vector continuous thereto equal to or smaller than a first predetermined angle and an integrated value of the deflection angles equal to or smaller than a second predetermined angle. When,
A scan segment vector forming step of combining a plurality of selected consecutive element vectors with each other to form one scan segment vector;
A line segment defined by the start point and end point of the formed scan segment vector is compared with a second predetermined length, and when the line segment is equal to or longer than the second predetermined length, a planar obstacle along the scan segment vector An obstacle recognition step for recognizing that
In the element vector formation step, an element vector formed with an arbitrary scan point as a first scan point and a scan point continuous in time series with the first scan point as a second scan point is longer than the first predetermined length. In some cases, the element vector is formed using another scan point that is not continuous in time series with the first scan point as a new second scan point.

The eleventh aspect of the present invention is the planar obstacle recognition method according to the tenth aspect , wherein the element vector forming step forms an element vector using only scan points within a predetermined distance from an arbitrary origin. And

According to a twelfth aspect of the present invention, in the planar obstacle recognizing method according to the tenth or eleventh aspect , even if the element vector forming step forms the element vector a predetermined number of times for the arbitrary scan point, the element vector forming step When an element vector having a length equal to or shorter than the first predetermined length cannot be obtained, an element vector is formed using a scan point that is continuous in time series with the arbitrary scan point as a new first scan point.

The invention of claim 13 is the planar obstacle recognition method of claim 10 or 11 ,
The element vector forming step forms an element vector using a scan point that is continuous in time series with the second scan point as a third scan point,
An element vector having a second scan point as a start point and a third scan point as an end point is larger than a first predetermined angle with respect to an element vector having the first scan point as a start point and the second scan point as an end point. In this case, the third scan point is excluded, and an element vector is formed using another scan point that is not continuous in time series with the second scan point as a new third scan point.

The invention of claim 14 is the planar obstacle recognition method of claim 10 ,
When one scan segment vector is formed by the scan segment vector formation step, an attempt is made to form a new scan segment vector using a scan point that was not used to form the scan segment vector among a plurality of scan points. It is characterized by.

The invention of claim 15 is the planar obstacle recognition method according to claim 14 ,
When two or more scan segment vectors are formed, the angle formed by the second scan segment vector with respect to the other first scan segment vectors is equal to or less than the first predetermined angle, and the start point of the second scan segment vector Is within a predetermined distance from the end point of the first scan segment vector, the scan segment vector forming step performs the first scan with a vector having the end point of the first scan segment vector as the start point and the start point of the second scan segment vector as the end point A new scan segment vector is formed by combining between the segment vector and the second scan segment vector.

A sixteenth aspect of the present invention is the planar obstacle recognition method according to the tenth aspect ,
The scan segment vector formation step stores the information of the scan points that contributed to the formation of the scan segment vector,
In the obstacle recognition step, when the length of the line segment defined by the start point and the end point of the scan segment vector is equal to or longer than the second predetermined length, using the coordinates of the saved scan point, the least square method is used. A line segment that optimally fits the scan point is calculated, and this line segment is recognized as a planar obstacle.

According to the first or tenth aspect of the present invention, the lengths of the plurality of continuous element vectors are each equal to or less than the first predetermined length, and the deflection angle formed by another element vector continuous to one element vector is the first angle. When the integrated value of the deflection angle is equal to or less than the first predetermined angle and the integrated value of the deflection angle is equal to or less than the second predetermined angle, the scan points used to form these element vectors are regarded as points on the same obstacle. Can do. Then, when the length of the scan segment vector formed by vector synthesis of these element vectors is equal to or longer than the second predetermined length, the obstacle can be further regarded as a planar obstacle. Since this scan segment vector represents the position, direction, size, etc. of a planar obstacle, the autonomous mobile device accurately identifies its position on the map based on the direction and distance to the planar obstacle. Accordingly, when the autonomous mobile device performs autonomous movement, it can perform safe autonomous movement by traveling while avoiding the planar obstacle specified by the scan segment vector. In addition, since the recognition of the planar obstacle is performed by a relatively simple process of scan point acquisition, element vector formation, element vector selection, scan segment vector formation, and scan segment vector comparison, The planar obstacle can be recognized in a shorter time than the method. As a result, an efficient autonomous movement becomes possible, such as the traveling speed of the autonomous mobile device can be increased. Also, when scan points that are consecutive in time series are too far apart to be suitable for forming an element vector, the scan point is skipped and an element vector is formed using another scan point. Can do. Therefore, for example, when a planar obstacle has a gap, even if a reflected wave from an object behind the obstacle is obtained as a scan point through the gap, the scan point is skipped. It is possible to accurately recognize a planar obstacle having a gap.

According to the second or eleventh aspect of the present invention, it is possible to reduce the amount of calculation by excluding the scan points of the obstacles that are relatively far away from the object of the obstacle recognition processing, for example, more judgment for autonomous movement Can be done quickly.

According to the third or twelfth aspect of the present invention, when the first scan point for which an element vector is to be formed is not appropriate, the scan point can be skipped to form an element vector for the next scan point. . As a result, it is possible to prevent a trouble that a planar obstacle cannot be recognized without forming an element vector.

According to the invention of claim 4 or 13 , when the angle formed by two element vectors formed by three time-sequential scan points is too large to be suitable as an element vector forming a scan segment vector It is possible to skip the third scan point and form an element vector using other scan points. Therefore, for example, when a planar obstacle has a gap, even if a reflected wave from an object behind the obstacle is obtained as a scan point through the gap, the scan point is skipped. It is possible to accurately recognize a planar obstacle having a gap.

According to the invention of claim 5 or 14 , for example, when a planar obstacle having a gap exists in parallel with a planar obstacle having no gap, Since a new planar obstacle recognition process is performed using a scan point that was not used to form a scan segment vector representing one planar obstacle, other planar obstacles existing behind one planar obstacle It becomes possible to accurately recognize a planar obstacle. And since it can grasp | ascertain correctly which of the position of several planar obstruction is recognized, the self position at the time of the autonomous movement of an autonomous mobile apparatus can be pinpointed correctly.

According to the invention of claim 6 or 15 , when a single planar obstacle is originally represented by a plurality of scan segment vectors for some reason, a new vector is added to the interrupted part under a certain condition. By combining, it can be combined into one scan vector. As a result, even when the planar obstacle is, for example, a fence composed of a plurality of support columns and a wire mesh held between the pillars, the planar obstacle can be accurately recognized.

According to the invention of claim 7 or 16 , when extracting a scan segment from a scan segment vector, these scan points are obtained by a least square method using the position coordinates of a plurality of scan segments contributing to the formation of the scan segment vector. Since the line segment that optimally fits is calculated, information closer to the actual position and direction of the planar obstacle can be obtained.

According to the invention of claim 8 , when the autonomous mobile device moves autonomously, the position of the planar obstacle on the map is specified by collating the scan segment with the map segment, and the autonomous mobile device on the map The self position can be specified.

According to the invention of claim 9 , the self-position on the map is determined based on the position of the planar obstacle on the map that is known in advance and the distance and direction from the origin on the autonomous mobile device to the planar obstacle. Can be accurately identified.

  Hereinafter, an autonomous mobile device and a planar obstacle recognition method suitable for the autonomous mobile device according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a block configuration of an autonomous mobile device 1 of the present invention. The autonomous mobile device 1 includes a storage unit 11 that stores map information of an operation area and various parameters for traveling, an environment information acquisition unit 12 that acquires environment information for specifying the position of an obstacle and its own position, A traveling device 13 for traveling, a self-position recognition unit 14 that collates the map information stored in the storage unit 11 and the environment information acquired by the environment information acquisition unit 12, and identifies its own position; The traveling device 13 is controlled while avoiding obstacles while identifying its own position based on a human interface 15 for inputting a destination, a parameter for traveling, and the like and a result of identification by the self-position recognition unit 14. The traveling control unit 17 is provided.

  The traveling device 13 includes a motor driven by the power of the battery 13a. The motor is provided with an encoder for measuring the rotation speed and rotation speed. The traveling control unit 17 of the autonomous mobile device 1 can know the moving distance and the moving direction using the output of the encoder, and performs dead reckoning (dead reckoning estimated navigation) based on them. The human interface 15 includes a touch panel and a keyboard that can be directly operated by a person, or a communication device that can be remotely operated by a person.

  Further, as the map information, a set of line segments representing planar obstacles such as walls, fences, and fences in the operation area is set in advance as map segments and stored in the storage unit 11. Furthermore, the environment information acquisition unit 12 uses a specific position of the main body of the autonomous mobile device 1 as the origin, scans the space facing from the origin at a constant angle or a constant interval, and exists in that direction for each direction from the origin. Acquires the position coordinates of a point on an obstacle including a planar obstacle such as a wall as environment information, and obtains a scan point consisting of a set of a plurality of position coordinates ordered according to the scan time series, for example, a laser radar Scanning point acquisition unit 12a.

  The scan point acquisition unit 12a is disposed, for example, in front of the main body of the autonomous mobile device 1, and scans the front space at a predetermined height and a predetermined angle to detect a planar obstacle such as a wall and other environmental objects. The reflected wave is received, and the position coordinates of the object reflected by the laser beam are acquired as a scan point. Further, as the scan point acquisition unit 12a, an ultrasonic sensor capable of electronic scanning, which is a combination of ultrasonic transmission elements and ultrasonic reception elements arranged in an array, can be used. Furthermore, a normal ultrasonic sensor for obstacle detection can be used for obstacle detection by the environment information acquisition unit 12. Note that the scan by the scan point acquisition unit 12a is not necessarily horizontal and may be inclined. Further, the irradiated laser beam or the like may have a certain spread.

  The self-position recognition unit 14 further includes an element vector formation unit 14a, an element vector selection unit 14b, a scan segment vector formation unit 14c, and an obstacle recognition unit 14d. The element vector forming unit 14a forms an element vector having the previous scan point as the start point and the subsequent scan point as the end point for two scan points whose acquired order is before and after the chronological order. . The element vector selection unit 14b includes a plurality of continuous element vectors from among the formed element vectors, and each element vector has a length equal to or less than a first predetermined length. The one in which the deflection angle formed by other successive element vectors is equal to or smaller than the first predetermined angle and the integrated value of the deflection angles is equal to or smaller than the second predetermined angle is selected. The scan segment vector forming unit 14c combines a plurality of selected continuous element vectors to form one scan segment vector. The obstacle recognition unit 14d compares the line segment defined by the start point and the end point of the formed scan segment vector with the second predetermined length, and when the line segment is equal to or longer than the second predetermined length, the scan segment vector The plane obstacle is determined to exist along the line, and the scan segment extracted from the scan segment vector is compared with the map segment to specify the position of the plane obstacle on the map. The self-position recognition unit 14 accurately specifies the self-position of the autonomous mobile device 1 based on the distance and direction to the planar obstacle with respect to the self-position.

  The storage unit 11, the self-position recognition unit 14, the element vector formation unit 14 a, the element vector selection unit 14 b, the scan segment vector formation unit 14 c, the obstacle recognition unit 14 d, and the travel control unit 17 include a CPU, a memory, and an external storage device The function is achieved by executing a predetermined program in a computer having a general configuration including a display device, an input device, and the like.

  Next, with reference to FIG. 2, a description will be given of a procedure in which the autonomous mobile device 1 autonomously travels while performing self-localization or recognition using the planar obstacle recognition method. The traveling control unit 17 knows the approximate position of the autonomous mobile device 1 based on the encoder information of the drive motor that drives the traveling device 13. The autonomous mobile device 1 performs dead reckoning based on the information on its own position at the start of traveling (# 1). The scan point acquisition unit 12a acquires a group of scan points in front of the traveling direction under the control of the environment information acquisition unit 12 (# 2). Next, the element vector forming unit 14a forms an element vector from the scan points, the element vector selecting unit 14b selects a plurality of element vectors satisfying the above requirements, and the scan segment vector forming unit 14c vectorizes the scan segment vector SSV. Synthesize (# 3).

  In the subsequent step, the obstacle recognizing unit 14d extracts a line segment having a second predetermined length or more from the line segment defined by the start point and end point of the scan segment vector SSV, and the extracted line segment exists in the traveling environment. It is determined that the scan segment SS is a line segment representing a planar obstacle. Further, the obstacle recognizing unit 14d compares the scan segment SS with a map segment formed by a set of line segments of the planar obstacle in the map information, and specifies the position of the planar obstacle on the map. The self-position recognition unit 14 specifies the self-position of the autonomous mobile device 1 based on the position of the planar obstacle on the map and the distance and direction to the planar obstacle with respect to the origin (self-position) (# 4).

  In response to the result of the self-position identification by the self-position recognition unit 14, the travel control unit 17 corrects the self-position of the autonomous mobile device 1 traveling by dead reckoning to a more accurate position, and performs travel control. And the traveling control part 17 complete | finishes driving | running | working, when the autonomous mobile apparatus 1 arrives at the destination (it is YES at # 5). If the destination has not been reached (NO in # 5), the above steps # 1 to # 5 are repeated at predetermined intervals.

  Next, with reference to FIGS. 3A and 3B, FIG. 4, and FIGS. 5A to 5D, a scan point acquisition method and definition of terms in the planar obstacle recognition method will be described. Each scan point is irradiated with a laser beam at a predetermined angle or at a fixed interval in front of the autonomous mobile device 1 or in a direction of interest by the laser radar that is the scan point acquisition unit 12a as described above, and is a flat obstacle such as a wall. It is the position coordinate information from the origin of the part of the obstacle which received the reflected wave reflected by the obstacle containing and reflected the laser beam. A group of scan points is a set of a plurality of position coordinates ordered according to time series. When the position of the planar obstacle on the map is specified as environmental information, if the planar obstacle can be recognized, it can be used to identify the self-position of the autonomous mobile device 1.

  Examples of the planar obstacle include a continuous building wall 21 as shown in FIG. 3 (a), a discontinuous outdoor lattice fence as shown in FIG. 3 (b), a wire mesh fence, or an armor plate. 22 etc. are included. The coordinate data of the scan point is obtained, for example, as data of a region 10 at a front 180 ° within a predetermined distance with a specific position of the main body of the autonomous mobile device 1 as the origin, as shown in (a). Here, the wall 21 is detected between the angle θ1 and the angle θ2.

  FIG. 4 shows an example of the spatial arrangement of the measured scan points. In this example, the scan point S (i) (i = 0 to 360) is measured every angle θ = 0.5 ° in the angle range 0 to 180 ° by the scan point acquisition unit 12a. Each scan point S (i) is grouped according to a certain rule to be described later, and a specific group of them constitutes a scan segment vector SSV (i, j) representing a planar obstacle. In the following, arguments such as i and j are integers, and the magnitude thereof represents the measurement order in the time series of scan point measurement.

For example, the scan points S (1) to S (4) constitute a scan segment vector SSV (1, 4). Further, the scan points S (i) to S (i + 6) except the scan point S (i + 4) constitute a scan segment vector SSV (i, i + 6). The scan point S (i + 4) and the scan point S _UN surrounded by the broken line G are scan points belonging to a group excluded from the groups constituting the scan segment vector SSV.

Here, the element vector V will be described. The element vector V is a vector that is combined with each other to form the scan segment vector SSV, and the scan point S is the start point and the end point. Accordingly, the scan points S constituting the scan segment vector SSV form an element vector V with each other. On the other hand, the scan point S _UN that does not constitute the scan segment vector SSV, for example, S (i + 4) does not form the element vector V. In addition to the case where the element vector V is formed by two scan points S that are continuous in time series, the element vector V may be formed by two scan points S that are not necessarily continuous in time series. For example, in the case of a discontinuous planar obstacle such as a fence, the scan point S (i + 4) is considered to be an obstacle point on the other side of the planar obstacle that has reflected the laser beam through the gap. .

  Next, procedures and rules for forming the scan segment vector SSV from the scan point S will be described. FIG. 5A shows an element vector V (j−1, j) defined by two scan points S (j−1) and S (j). The direction of the element vector V is from the scan point (j-1) obtained earlier by scanning to the scan point S (j) obtained later. In order for the vector defined by the two scan points S to be the element vector V, the distance between the two scan points S, that is, the vector length L (j−1, j) is the first predetermined value. It must be less than or equal to the length ML. Such formation of the element vector V is performed by the element vector forming unit 14 a provided in the self-position recognition unit 14.

  FIG. 5B shows a state where a new element vector V (j, j + 1) is defined by adding a new scan point S (j + 1) to the element vector V (j-1, j) described above. In this case, the length of the new element vector V (j, j + 1) is equal to or shorter than the first predetermined length ML, and the element vector V (j, j + 1) is another element vector V (j−1, The angle θ (j, j + 1) made with respect to j) needs to be equal to or smaller than the first predetermined angle Mθ.

  FIG. 5C shows the angle θ (j, j + 1) formed by the vector defined by the scan points S (j) and S (j + 1) with respect to the element vector V (j−1, j) that follows the vector. A case where the scan point S (j + 1) is excluded from the element vector formation points by being larger than the first predetermined angle Mθ is shown. In that case, an element vector V (j, j + 2) is formed using the next scan point S (j + 2) skipped by one. Such selection of the element vector V is performed by the element vector selection unit 14b provided in the self-position recognition unit 14.

FIG. 5D shows an example of a scan segment vector SSV (j, j + 5) defined when a plurality of element vectors V are formed continuously. The scan segment vector SSV is a vector formed by vector synthesis of a plurality of element vectors V. Therefore, this is a vector having the start point of the first element vector V of the continuous element vectors as the start point and the end point of the last element vector V as the end point. The scan segment vector SSV has two such scan points S (j) and S (j + 5) even if there are skipped scan points S _SK among the plurality of scan points S forming the combined element vector V. ) Can be defined.

  Since a planar obstacle is to be expressed by the scan segment vector SSV, it is desirable to provide a restriction that assumes the shape of a planar obstacle that actually exists in the vector synthesis. Therefore, for example, when the scan segment vector SSV is formed by vector synthesis of a plurality of element vectors V, a condition is set for the angle. When the sum of the angles θ formed by two adjacent element vectors V is equal to or smaller than the second predetermined angle MΣθ, this combined vector is set as a scan segment vector SSV. The total angle Σθ is obtained by defining and adding the angles between the element vectors V included in the scan segment vector SSV with positive and negative signs. This condition corresponds to the case where a continuous planar obstacle is assumed.

  Further, the scan segment SS is defined from the scan segment vector SSV under a predetermined restriction. That is, when the line segment defined by the start point and the end point of the scan segment vector SSV is not less than the second predetermined length TL0, this line segment is defined as the scan segment SS. After a plurality of scan segment vectors SSV are finally formed from a large number of scan points S, a specific one is extracted from these scan segment vectors SSV, and a scan segment SS having a length equal to or longer than the second predetermined length TL0 is obtained. can get. Such extraction of the scan segment SS is performed by the obstacle recognition unit 14 d provided in the self-position recognition unit 14.

  The obstacle recognizing unit 14d further functions as a segment collating unit, and collates with such a scan segment SS and a map segment made up of a set of line segments of the planar obstacle in the map information to determine the position of the planar obstacle. It specifies on a map, and also specifies the self-location of the autonomous mobile device 1. In this way, if the position of the planar obstacle is stored in advance as environmental information (segment map) on the map, the position of the planar obstacle is determined by a simple procedure by comparing the scan segment SS with the map segment. Therefore, the autonomous mobile device 1 can autonomously travel efficiently and safely.

As a reference example, the extraction of the scan segment SS will be described in detail with reference to the flowchart shown in FIG. The extraction processing described here is processing when the scan point S is not skipped when the element vector V is formed. The processing when the skip is permitted will be described later.

  The extraction of the scan segment SS starts with preparation of scan points S (i) (i = 0, 1,... N) composed of (n + 1) points acquired by the scan point acquisition unit 12a (# 11 ). The start point of the scan segment vector SSV is assumed to be the first scan point S (0), and the number m of the scan point S (m) that is the start point of the scan segment vector SSV is initialized as m = 0 (# 12). If this scan point S (0) does not form the element vector V, m is updated one after another. In this sense, the start point number is “assumed”, and m is determined when the scan segment SS including this point is determined.

  Next, the angle integration variable Σθ and the vector composition point number j are initialized (# 13). The angle integration variable Σθ is a variable for integrating the angle between the two element vectors V with a positive / negative sign. The vector synthesis point number j is the number of the scan point S (j) where the two consecutive element vectors V (j−1, j) and V (j, j + 1) in FIG. 5B are synthesized. Since m = 0 at this point and j = m + 1, j = 1. That is, the second scan point S (1) is a vector synthesis point.

  Next, the first element vector Va having the start point and the end point as the scan points S (j−1) and S (j) is formed (# 14). At this time, j = 1, and the start point and end point of the first element vector Va are the scan points S (0) and S (1), respectively. In the following processing, since the two element vectors V are combined, the first element vector Va is set as the second element vector Vb.

  Next, it is checked whether or not the first element vector Va is equal to or smaller than the first predetermined length ML. If the vector absolute value | Va | is equal to or smaller than the first predetermined length ML (| Va | ≦ ML) (YES in # 15) ), The element vector Va is held, and the scan segment vector SSV, which is the next stage of vector synthesis, is initialized (# 18). That is, the scan segment vector SSV = Va.

  When the length of the first element vector Va is longer than the first predetermined length ML and | Va |> ML (NO in # 15), the process proceeds to step # 16, and the element vector V can be formed and formed. It is checked whether there is a scan point S that allows the synthesized element vector V to be vector-synthesized. If there is no such scan point S (NO in # 16), the scan segment extraction process ends. On the other hand, when such a scan point S exists (YES in # 16), the value m of the scan point S (m) that is the start point of the current scan segment vector SSV is discarded, and a new m = j Is set (# 17). Thereafter, the above-described processing proceeds from step # 13.

  After the scan segment vector SSV is initialized in step # 18, the second element vector Vb is formed (# 19), and the deflection angle θ formed by the second element vector Vb with respect to the first element vector Va is the vector It is obtained using the inner product and the absolute value (# 20). Next, it is checked whether or not the deflection angle θ is equal to or smaller than the first predetermined angle Mθ described above. If it is equal to or smaller than the first predetermined angle Mθ (YES in # 21), the length of the second element vector Vb is checked. (# 22). If the length of the second element vector Vb is equal to or shorter than the first predetermined length ML (YES in # 22), the first element vector Va and the second element vector Vb are combined, and the scan segment vector SSV is updated ( # 23). Subsequently, the angle integration of the deflection angle θ is performed (# 24).

  Next, the angle integrated value Σθ is compared with the second predetermined angle MΣθ, and if the angle integrated value is equal to or smaller than the second predetermined angle MΣθ (YES in # 25), the addition of the element vector V to the scan segment vector SSV is completed. To do. If there is another scan point S (YES in # 26), the process proceeds to vector synthesis of the next point. That is, the second element vector Vb newly added to the scan segment vector SSV is replaced with the first element vector Va (# 31), the vector composition point number is advanced, and j = j + 1 is set (# 32). The process is repeated from step # 19 described above.

  If the condition is not satisfied in any of the above steps # 21, # 22, # 25, and # 26 and the result is NO, the vector synthesis is stopped and the process proceeds to step # 27. At this time, the start point and end point of the scan segment vector SSV are the scan points S (m) and S (j), respectively. Next, it is checked whether or not the length | SSV (m, j) | of the scan segment vector SSV (m, j) is equal to or longer than the second predetermined length TL0 (# 28). If the length | SSV (m, j) | of the scan segment vector SSV (m, j) is equal to or longer than the second predetermined length TL0 (YES in # 28), the scan segment SS is calculated from the scan segment vector SSV (m, j). (M, j) is obtained (# 29).

  The scan segment SS is a line segment defined by two scan points S (m) and S (j). In this specification, such a line segment is represented by a symbol SS (m, j). It is shown for convenience.

  If the length | SSV (m, j) | of the length SSV (m, j) of the scan segment vector SS is shorter than the second predetermined length TL0 in step # 28 (NO in # 28), the scan segment vector SSV ( Since the scan segment SS (m, j) cannot be obtained from m, j), the process proceeds to step # 30. For example, in an extreme case, when the scan segment vector SSV (m, j) is formed only by vector synthesis of two element vectors V, the length | SVV of the length SSV (m, j) of the scan segment vector SS Since (m, j) | is too short, it cannot be considered that the scan segment SS represents a planar obstacle.

  In step # 30, the number of remaining scan points S is checked. If the scan point S necessary for vector composition does not remain (NO in # 30), the scan segment extraction process ends. On the other hand, if the number of scan points S necessary for vector composition remains (YES in # 30), the process proceeds to step # 17. In step # 17, the number j of the last scanned scan point S is set as the start point number m of the new scan segment vector SSV. Thereafter, the above steps are repeated from step # 13.

As another reference example, with reference to the flowchart of FIG. 7, the scan segment extraction in the case where the skip of the scan point S is permitted when the element vector V is formed will be described in detail. In the extraction process described here, skipping of the scan point S is permitted up to a predetermined number MK. As an example of skipping of the scan point S, for example, in FIG. 4, the scan point S (i + 4) is skipped. An element vector skipped once is formed by the scan points S (i + 3) and S (i + 5).

  The flowchart shown in FIG. 7 is substantially the same as the flowchart shown in FIG. 6 except for the processing related to skipping of the scan point S. Therefore, the difference between the two will be described, and the description of the overlapping points will be omitted. To do. In the process shown in this flowchart, a skip variable k for counting the number of skips is introduced. The skip variable k is initialized at step # 43.

  Thereafter, the same steps # 44 to # 50 as in the process without skipping are passed, and the swing angle θ of the element vector V is checked in step # 51. Here, when the deflection angle θ exceeds the first predetermined angle Mθ (NO in # 51), the next scan point S is set to the element vector V without stopping vector synthesis immediately, unlike the case of no skip. (# 63, # 64, # 65). Then, the process returns to step # 49.

  In step # 63, the skip count is checked. In step # 64, the presence or absence of the scan point S is checked. In step # 65, the skip variable is incremented (updated). In step # 49, since the number of the end point scan point S (j + k) defining the second element vector Vb, that is, its argument is j + k, it is skipped (k-1) times (one time if k = 2). You can see that it has been skipped.

  Thus, the step in which the skip variable k is used is different from the case of no skip shown in FIG. For example, in step # 62, the next vector synthesis point is (j + k). If k = 1, there is no skip, and if k = 2, it is skipped once. In step # 66 after step # 62, the skip variable k is initialized.

  Next, the process at step # 47 will be described. In step # 47, the skip variable k is not involved, but since the vector synthesis process is newly started, the number m of the scan point S (m) that is the start point of the new scan segment vector SSV is set as m = j. Is set. For example, j, which is the last vector composition point number at the time of step # 47, is set as the start point number m. For example, the scan point S () of the last vector composition point number j at the corner where the planar obstacle intersects is used. j) Because there is a case.

  Next, an example of synthesizing a plurality of scan segment vectors SSV will be described with reference to FIGS. 8, 9A to 9C, and FIG. As an example of a planar obstacle from which the scan point S is acquired, a fence 3 in which a flat fence body 32 is provided between fence posts 31 is assumed as shown in FIG. Such a fence 3 has a structure in which a column 31 partially protrudes from a plane (fence body) 32 as shown in FIG. Therefore, when the scan segment vector SSV is formed from the scan point S obtained by scanning the fence 3 with a laser radar or the like, the scan segment vector SSV is obtained from the fence column 31 as shown in FIG. By the way, they are separated from each other.

  Therefore, as shown in FIG. 9C, a plurality of scan segment vectors SSV are connected to each other within a range satisfying a predetermined condition to form a new large scan segment vector SSV0. The connection conditions will be described with reference to FIG. An angle φ formed by two scan segment vectors SSV1 and SSV2 selected from a plurality of scan segment vectors SSV obtained from the scan point S is equal to or smaller than a second predetermined angle, and the start point of one scan segment vector SSV2 and the other When the distance d of the end point of the scan segment vector SSV1 is equal to or less than a predetermined distance, these two scan segment vectors SSV1 and SSV2 are added as vectors from the end point to the start point, and become a new one scan segment vector SSV. Synthesized.

  By combining a plurality of scan segment vectors SSV in this way, one long scan segment vector SSV0 can be formed, and thereby a long scan segment SS0 can be extracted. As a result, the planar obstacle can be recognized more accurately, and the self position of the autonomous mobile device 1 can be specified more accurately.

  Next, FIG. 11 shows a flowchart at the time of traveling of the autonomous mobile device 1 that travels while performing such processing. Since the processing for synthesizing the scan segment vector SSV in step # 74 is the same as that in the flowchart shown in FIG.

  Next, the formation of the scan segment vector SSV for a planar obstacle having a gap will be described with reference to FIGS. 12 and 13A to 13C. The operation area of the autonomous mobile device 1 is assumed to be an outdoor factory site, a park, a theme park, and the like in addition to indoors such as hospitals and production factories. In such an outdoor operation region, as shown in FIG. 12, a fence 30 having a fence body 32 having a large transmittance (gap) such as a wire mesh between fence posts 31 is provided as a boundary of a traveling path. There is. Therefore, as an example of a planar obstacle having a gap with such a fence 30 existing in the operation area, a set of line segments to be expressed as a map segment composed of a set of line segments of planar obstacles in map information The information is stored in the storage unit 11 of the autonomous mobile device 1.

  As shown in FIG. 13 (a), the cross section of the fence 30 has a fence post 31 portion that blocks or reflects the laser beam of the laser radar, and a fence main body 32 portion that transmits most of the laser beam. Then, when the fence 30 composed of a plurality of fence posts 31 and the fence body 32 provided between the fence posts 31 is scanned in a substantially horizontal plane by the scan point acquisition unit 12a, it is mainly reflected by each fence post 31. The reflected wave is received by the scan point acquisition unit 12a. As a result, the position coordinates of the point on the fence post 31, the fence body 32, or the object 36 on the other side of the fence 30 are acquired as the scan point S. As shown in FIG. 13B, the obtained scan point S is a set of points distributed at positions corresponding to these objects.

  The element vector formation unit 14a of the self-position recognition unit 14 uses the flow for skipping scan points shown in FIG. 7 to scan points S (i) and S (i + 1) at points on each fence post 31 and fence body 32. ), S (i + 3), S (i + 5), S (I + 6)..., Element vectors V (i, i + 1), V (i + 1, i + 3), V (i + 3, i + 5), V (i + 5) i + 6)... Further, the element vector selection unit 14b and the scan segment vector formation unit 14c combine these element vectors V to form a scan segment vector SSV.

  Further, the obstacle recognizing unit 14d determines whether or not the formed scan segment vector SSV can be recognized as a planar obstacle, and when it can be recognized as a planar obstacle, the fence further stored in the storage unit 11 The map segment representing 30 is compared with the scan segment SS defined by the obtained scan segment vector SSV. Thereby, the position of the fence 30 on the map is specified, and the self position of the autonomous mobile device 1 is specified based on the position.

  In this way, by recognizing the fence column 31 and the fence body 32 as the scan point S, the fence 30 including such a fence body 32 can be detected as a scan segment vector even when the gap ratio of the fence body 32 is high. A scan segment SS that can be expressed as SSV and collated with a map segment can be extracted. In addition, when acquiring such a fence support | pillar 31 selectively as the scan point S, an ultrasonic sensor whose spatial resolution is lower than a laser radar can be used.

  Next, a method for more reliably acquiring the position of the fence post 31 of the fence 30 as the scan point S will be described with reference to FIGS. 14 and 15A to 15C. In this case, as shown in FIG. 14, a reflector mark 33 is pasted on the fence post 31 of the fence 30 in advance. FIG. 15A shows the cross section of the fence 30 and the state of the reflector mark 33. FIG. 15B shows the acquired scan point S, and FIG. 15C shows the formed element vector V and scan segment vector SSV.

  When this fence 30 is scanned in a substantially horizontal plane using a laser radar, the intensity of the reflected wave reflected by the reflector mark 33 becomes higher than the intensity of the reflected wave reflected from other points, so that a large number of scans are performed. Only the scan point S on the fence post 31 can be extracted from the point S. According to such a method using the reflector mark 33, the fence post 31 can be easily recognized, and hence the fence 30 can be easily recognized.

  Next, a case where the scan point S to be processed is limited to a predetermined distance will be described with reference to FIGS. By reducing the number of scan points S acquired by the scan point acquisition unit 12a, it is possible to reduce the amount of calculation for forming the scan segment vector SSV performed by the self-position recognition unit 14, thereby enabling autonomous movement. Judgments can be made faster.

  As shown in FIG. 16, in accordance with the operating environment of the autonomous mobile device 1, the origin (the position of the scan point acquiring unit 12a) is included in the region 10 where the object coordinates can be measured by the scan point acquiring unit 12a of the autonomous mobile device 1. ) Is set within a predetermined distance. Then, an element vector V is formed using the scan point S in the region 10a.

  As a method for determining such a region 10a, for example, it can be used that the intensity of the reflected wave of the laser beam emitted from the laser radar changes according to the distance to the object. First, as shown in FIG. 17, the beam 41 of the laser beam LB at the position where the wire 41 is located with respect to the wire 41 constituting the fence at a certain distance from the position of the scan point acquisition unit 12 a provided with the laser radar. The area of the beam spot BS is calculated based on the diameter of the spot BS. Next, the ratio of the area occupied by the wire 41 to the area of the beam spot BS is calculated. A ratio obtained by multiplying the ratio by the reflectance specific to the material of the wire 41 is taken as a measured reflectance for the laser beam of the laser radar. As the distance increases, the area of the beam spot BS increases in proportion to the square of the distance. However, since the area of the wire 41 only increases in proportion to the distance, the measured reflectance value decreases as a result. . Therefore, the scan point acquisition unit 12a sets the measurement reflectance according to the distance that defines the region 10a, and performs a filtering process so as not to receive the reflected light that is equal to or less than the measurement reflectance, thereby performing the scan to be processed. The point S can be limited to a predetermined distance.

  Next, a situation in which the autonomous mobile device 1 autonomously travels while recognizing a planar obstacle such as a wall or a fence will be described with reference to FIGS. FIG. 18A shows a state in which the autonomous mobile device 1 is moving in a state in which a warning area 4 around the autonomous mobile device 1 and a caution area 5 in a region extending forward in the traveling direction are set. When the autonomous mobile device 1 detects the obstacle 23 in the caution area 5, the autonomous mobile device 1 slows the moving speed, or generates a display or warning sound for alerting the outside. Further, the autonomous mobile device 1 moves so that the obstacle does not enter the warning area 4, and immediately stops when an obstacle is detected in the warning area 4. As described above, the autonomous mobile device 1 efficiently moves autonomously by setting the control areas divided in stages around the autonomous mobile device 1.

  By the way, as shown in FIG. 12, a side groove 35 may be provided in front of the wall or fence, or a block 34 may be provided at the base of the fence column 31. Such a step due to the side groove 35 or the block 34 exists at a low position close to the traveling road surface, and therefore cannot be recognized by obstacle detection in the front horizontal direction. Therefore, as shown in FIG. 18B, for a specific obstacle 24 such as a wall or a fence, an enlarged alert area 4a and an enlarged alert area are respectively provided in the alert area 4 and the alert area 5 around the autonomous mobile device 1. An area 5 a is added so as not to approach a specific obstacle 24. As a result, the autonomous mobile device 1 can efficiently perform autonomous movement by preventing in advance occurrence of problems during traveling such as derailment or falling.

  The above-described enlarged warning area 4a and enlarged attention area 5a may be set inside the visual field range of the specific obstacle 24 as shown in FIG. In this case, in the area where the specific obstacle 24 is not recognized, the normal traveling movement shown in FIG. 18A can be performed, and the autonomous traveling can be efficiently performed.

  Here, the case where the planar obstacle which has a space | gap is assumed as the specific obstacle 24 is demonstrated, referring FIG. 19 (a) and (b). FIG. 19A schematically shows the arrangement of the planar obstacle, and FIG. 19B shows map segment data corresponding to the arrangement of the planar obstacle. In FIG. 19A, a solid line indicates a planar obstacle having no gap, and a broken line indicates a planar obstacle having a gap. The data shown in FIG. 19B is map segment data composed of a set of line segments of planar obstacles in map information stored in advance. For example, the left and right end points of a planar obstacle that can be detected by a laser radar are used. It is a coordinate value with the start and end points. This data includes information on the gap attribute (presence / absence of gap) for identifying whether or not the planar obstacle existing in the operation area has a gap, and the attribute is the flat obstacle having no gap. 0, a flat obstacle having a gap is set to 1.

  As described above, when a planar obstacle having a gap is assumed as the specific obstacle 24, the autonomous mobile device 1 can recognize the planar obstacle having the gap efficiently and reliably. Enable safe and efficient autonomous driving. Therefore, a method of extracting a scan segment SS by expressing a planar obstacle such as a wall or a fence with a scan segment vector SSV is applied.

As a method for recognizing a planar obstacle having a void, first, the element vector V is formed without skipping the scan point S according to the flowchart shown in FIG. 6, and the first group is selected from the element vector V selected under a predetermined condition. Scan segment vector SSV ₁ is formed. When the planar obstacle has a gap, for example, as shown by a broken line in FIG. 5C, the deflection angle θ formed by two consecutive element vectors V (j−1, j) and V (j, j + 1) is The probability of becoming larger than the first predetermined angle Mθ is high, and there is almost no possibility that the first group of scan segment vectors SSV ₁ is formed.

Next, according to the flowchart shown in FIG. 7, the element vector V is formed while skipping the scan point S as appropriate, and the second group of scan segment vectors SSV ₂ is formed from the element vector V. In this case, as shown by the solid line in FIG. 5C, the scan point S (j + 1) where the deflection angle θ is larger than the first predetermined angle Mθ is skipped, and therefore the planar obstacle has a gap. A second group of scan segment vectors SSV ₂ is also formed.

Next, when the first group of scan segment vectors SSV ₁ and the second group of scan segment vectors SSV ₂ obtained as described above are compared, both scan segment vectors corresponding to a planar obstacle having no air gap are both. However, the scan segment vector corresponding to the planar obstacle having a gap is included only in the second group of scan segment vectors SSV ₂ . Accordingly, the element vector forming the first group scan segment vector SSV ₁ is subtracted from the element vector forming the second group scan segment vector SSV ₂ , and the element vector V formed later is formed. The third group of scan segment vectors SSV ₃ corresponds to a planar obstacle having a gap. By extracting the scan segment SS from the third group of scan segment vectors SSV ₃ obtained in this way, a planar obstacle having a gap can be recognized.

  Next, a method for determining the first element vector V in forming the scan segment vector SSV will be described with reference to FIGS. 20 (a) to 20 (c) show examples that may occur regarding vector synthesis of the first element vector V. FIG. In the example shown in FIG. 20A, at the start of vector synthesis, the first element vector Va is formed by two scan points S (m) and S (m + 1) that are continuous in time series, and the second element vector Vb is formed by two scan points S (m + 1) and S (m + 4) that skip the scan points S (m + 2) and S (m + 3), but are not continuous but are continuous. This situation is handled without problems by the flow shown in FIG. 7 in which the skip variable k is introduced.

  In the example shown in FIG. 20 (b), a back row composed of scan points S (m), S (m + 2), S (m + 4)... And scan points S (m + 1), S (m + 3). There is a previous row consisting of ... In the case of such an array of scan points S, a situation occurs where the first element vector Va is not determined according to the flow shown in FIG. Specifically, in the situation shown in FIG. 20B, when the first element vector Va is formed by the scan points S (m) and S (m + 1) at the start of vector synthesis, the time series continues. The length of the vector formed by the two scan points S (m) and S (m + 1) exceeds the first predetermined length ML. Therefore, in the flow shown in FIG. 7, the first element vector Va is formed by the next two consecutive scan points S (m + 1) and S (m + 2). The two scan points S (m + 1) and The length of the vector formed by S (m + 2) also exceeds the first predetermined length ML. When such a situation continues, the first first element vector Va is not determined. What is important here is that the first element vector Va belonging to the same series, for example, the scan points S (m) and S (m + 2) in the subsequent column in FIG. 20B can be overlooked. .

  Therefore, as shown in FIG. 20C, if the second scan point S (m + 1) inappropriate for forming the first element vector Va is skipped with respect to the first scan point S (m), the first scan point S (m) The composition of the element vector V after the element vector Va is formed can be processed by the flow of FIG. Therefore, when an initial skip variable p is introduced to form the first element vector Va, inappropriate second and subsequent scan points can be skipped a predetermined number of times p. If the first element vector Va cannot be formed even after skipping the predetermined number of times p, the first element vector Va is shifted again by one scan point and an attempt is made to form the first element vector Va again. To do.

  21 and 22 show a flow for dealing with such a case by introducing the initial skip variable p. The flowcharts shown in FIGS. 21 and 22 are substantially the same as the flowchart shown in FIG. 7 except that the initial skip variable p is used. Therefore, only the differences between the flowcharts will be described, and overlapping points will be described. Is omitted.

  As shown in FIG. 21, after the number m of the scan point S (m) that is the start point of the scan segment vector SSV is initialized at step # 42, the initial skip variable p is initialized at step # 70. Thereafter, the first element vector Va is defined (# 71), and the length of the first element vector Va is inspected (# 45). When the first element vector Va exceeds the first predetermined length ML, the presence / absence of another scan point S (# 74) and skip possibility (# 75) are further examined. If skipping is possible (YES in # 75), one-point skip is performed (# 76), and the steps from step # 71 are repeated again.

  If skipping is not possible (NO in # 75), a new scan segment vector starting point number m is set (# 47), and the above steps are repeated again from initialization step # 70. When the first element vector Va is determined (YES in # 45), the skip variable k is initialized (# 73). The subsequent steps shown in FIG. 22 are the same as the flow shown in FIG. 7 except for the following points.

  In steps # 52 to # 55 of FIG. 7, the vector synthesis is stopped when the second element vector Vb exceeds the first predetermined length ML or the integrated angle Σθ exceeds the second predetermined angle MΣθ. In the flow shown in FIG. 22, the vector synthesis is attempted by skipping the scan point S (if # 52 or # 77 is NO, the process proceeds to # 63). As described above, when the scan segment extraction process according to the flow shown in FIGS. 21 and 22 is performed, the state shown in FIG. 20B can be avoided, and as shown in FIG. By combining, a scan segment vector SSV can be formed.

  Next, an example will be described in which scan segment extraction processing is performed a plurality of times retrospectively with respect to a plurality of scan points S preceding and following the time series. FIG. 23 shows how to measure the position coordinates of a double structure of an obstacle in which planar obstacles W1 and W2 having a gap exist in front of a planar obstacle W3 such as a wall having no gap. FIG. 24 shows how a scan segment vector SSV representing a planar obstacle is formed from a plurality of scan points S obtained by scanning the double structure of the obstacle shown in FIG. 23 with a laser radar. FIG. 27 shows three procedures for forming the scan segment vector SSV, respectively.

  In the operation area of the autonomous mobile device 1, as shown in FIG. 23, a fence W <b> 1 including an array of columns, a fence W <b> 2 having a transmission part, and the like are formed along the wall surface W <b> 3 of the building. There can be a double structure. In this case, two series of scan points S as shown in FIGS. 20B and 20C are obtained. When the scan segment extraction process is performed in the order obtained in time series for the plurality of scan points S obtained by the scan point acquisition unit 12a under the obstacle double structure, the previous column And the scan point S in one of the two series of scan points S in the rear column is skipped by the introduction of the initial skip variable p and the skip variable k. In the case of a double structure of obstacles, in particular, when an obstacle having a gap such as a fence W1 or a fence W2 existing in the foreground cannot be recognized, any planar obstacle is determined when the autonomous mobile device 1 locates itself. Whether it is detected is unknown, and as a result, the self-positioning accuracy may be lowered. In order to prevent such a problem, as shown in FIG. 24, it is important to form scan segment vectors SSV that overlap in the front-rear direction and to extract scan segments SS that represent obstacles that overlap in the front-rear direction.

  The extraction of the scan segment SS in the case of the double structure of the obstacle is, for example, the remaining scan points that are not used for the extraction of the scan segment SS after the extraction of the scan segment SS regarding the planar obstacle existing in the back is completed For S, the steps from the formation of the element vector V to the extraction of the scan segment SS are performed again.

In the flow shown in FIG. 25, every time one scan segment SS is extracted, the scan segment SS is extracted again for the scan point S _SK that was referred to when the scan segment SS was extracted but was skipped. Do. First, a group of scan points S1 is prepared (# 81), and processing up to the determination of the scan segment SS is performed (# 82, # 83). This process is performed, for example, from the first step # 41 of the flow shown in FIG. 7 to the step # 59 of determining the scan segment SS, or from the first step # 41 in the flow shown in FIG. 21 to the flow shown in FIG. This corresponds to the processing up to step # 59 for determining the scan segment SS in FIG. The scan point S _SK that has been skipped and unused among the scan points S1 that were referred to in the first-stage scan segment SS extraction is recorded as the first-stage unused scan point S _UN1 (# 84). .

Next, the second-stage scan segment SS extraction process is similarly performed on the first-stage unused scan point _SUN1 (# 85, # 86). Then, the scan point _SNR that has not been referred to in the process so far is prepared as a new first stage scan point S1 (# 87). Here, if there is no _unreferenced scan point _SNR , that is, if all the scan points S1 are referred to, the process ends (NO in # 88), and if there is an _unreferenced scan point _SNR (# 88). YES), the process returns to step # 82 and the above steps are repeated.

Also, in the flow shown in FIG. 26, after extracting all the scan segments SS that can be extracted for all the scan points S1, the scan segment SS is extracted again for the skipped scan points _SSK. I do. First, a group of scan points S1 is prepared (# 91), and the process until the determination of the scan segment SS is repeated until all the scan points S1 are referred to (# 92, # 93). This process corresponds to, for example, performing the flow shown in FIG. 7 or FIGS. 21 and 22 from the start to the end. However, for each scan point S1, a record of whether or not it contributed to the formation of the scan segment vector SSV is left. Then, a scan point _SUN that does not contribute to the formation of the scan segment vector SSV is prepared (# 94).

Thereafter, the scan segment SS extraction process is repeated in the same manner as described above until all the scan points _SUN prepared in step # 94 are referred to (# 95, # 96). Thus, by extracting the scan segment SS, it is possible to recognize a double-structured obstacle.

In the flow shown in FIG. 27, when one scan segment SS is extracted, the scan segment SS is extracted for the scan point S _SK that was not used in the previous process for extracting the scan segment SS. . First, a group of scan points S1 is prepared (# 101), and processing up to scan segment extraction is executed (# 102, # 103). This process is performed, for example, from the first step # 41 of the flow shown in FIG. 7 to the step # 59 of determining the scan segment SS, or from the first step # 41 in the flow shown in FIG. 21 to the flow shown in FIG. This corresponds to the processing up to step # 59 for determining the scan segment SS in FIG. When one scan segment SS is extracted, a group of scan points S2 that are referenced but not used in the previous process for extracting the scan segment SS and a group of scan points S3 that are not referred to are prepared. (# 104). Then, for example, in step # 47 in the flow shown in FIG. 7, by setting the number of the vector synthesis point to the number j of the first scan point S that was referred to but not used, in time series, The scan point S1 is replaced with a new scan point obtained by combining the scan points S2 and S3 (# 105).

  For example, in the double structure of the obstacle in FIG. 23, when the scan segment vector SSV1 representing the planar obstacle (fence) W1 is first extracted, the scan segment vector SSV2 representing the planar obstacle (fence) W2 is extracted. The scan segment vector SSV3 representing the planar obstacle (wall) W3 can be extracted first. According to such a configuration, since the length of the scan segment vector SSV3 can be detected accurately, the self-position of the autonomous mobile device 1 on the map can also be specified accurately.

  Next, a method for performing the extraction of the scan segment SS with higher accuracy and stability will be described. FIG. 28 shows how the scan segment SS is obtained by the least square method based on the scan segment vector SSV formed by any one of the above methods, and FIG. 29 shows the processing flow thereof. According to the method of extracting the scan segment SS, as shown in FIG. 28, the scan segment vector SSV (m, j) and the scan segment SS (m, j) that are finally determined are the scan point S ( m) and the end scan point S (j). That is, the positions of a large number of scan points S between the start point and the end point are not considered, and the position accuracy of the scan segment SS (m, j) depends strongly on the position detection accuracy of the start point and the end point scan point S. Yes.

  Therefore, in the step of forming the scan segment vector SSV (m, j), information on the scan point S that contributed to the synthesis of the scan segment vector SSV (m, j) is stored, and a large number of points between the start point and the end point are stored. The position information of the scan point S is statistically processed and reflected in the extraction of the scan segment SS (m, j) to ensure the accuracy improvement and stability of the extraction result. As a statistical process, for example, a least square method is used. That is, as shown in FIG. 29, in the step of extracting the scan segment SS (m, j), if the line segment defined by the start point and end point of the scan segment vector SSV (m, j) is greater than or equal to a predetermined length. (YES in # 58), by applying the least square method to the stored coordinate value of each scan point S, a line segment optimally fitted to the scan point S is calculated (# 79). Are extracted as scan segments SS (# 59).

  The scan segment SS (m, j) thus obtained is a scan point S between the start point and end point of the scan segment vector SSV (m, j) as shown in FIG. It reflects the position information of all the scan points S that contributed to the synthesis of SSV (m, j). The flow shown in FIG. 29 can be incorporated into the flow shown in FIG. 6, FIG. 7, FIG. 21, FIG.

  Next, another method for determining whether or not to synthesize a vector when the element vectors V are combined to form the scan segment vector SSV will be described. FIG. 30 shows another method of vector synthesis of element vectors V when forming a scan segment vector SSV representing a planar obstacle using scan points S in the planar obstacle recognition method according to the present invention. . FIG. 31 shows another criterion for determining the validity of the scan segment vector SSV.

  In the method shown in FIG. 30, instead of the deflection angle θ formed by two continuous element vectors V, the end point of the element vector Vb = V (j, j + 1) and the scan segment vector SSV (m, j) from the end point. The distance h from the lowered vertical leg is used for determining whether or not vector synthesis is possible. That is, vector synthesis is performed when h ≦ Mh with respect to the predetermined value Mh.

  In the method shown in FIG. 31, the integrated value Σh of the distance h between the end point of the element vector V and the vertical leg drawn from the end point to the scan segment vector SSV is used for determining whether or not vector synthesis is possible. That is, vector synthesis is performed when Σh ≦ MΣh with respect to the predetermined value MΣh. According to such a determination method, there is no time for calculating an angle, and it becomes easy to input the predetermined values Mh and MΣh.

  The present invention is not limited to the above-described configuration, and various modifications can be made. For example, in order to handle curved curved obstacles with a large radius of curvature, a map segment in which the obstacle surface is represented by a polygonal line and a combination of multiple scan segment vectors that are curved in one direction within a predetermined angle A scan segment vector may be used.

FIG. 1 is a block diagram showing a configuration of an autonomous mobile device according to the present invention. FIG. 2 is a flowchart of autonomous traveling of the autonomous mobile device using the planar obstacle recognition method according to the present invention. FIG. 3A is a plan view showing a state of recognizing a planar obstacle having no gap by the above-described planar obstacle recognition method, and FIG. 3B is a planar obstacle having a gap by the method. It is a top view which shows a mode that an object is recognized. FIG. 4 is an explanatory diagram showing a state in which a scan segment vector representing a planar obstacle is formed from a scan point by the planar obstacle recognition method. FIG. 5A is an explanatory diagram of the element vector V in the above-described planar obstacle recognition method, FIGS. 5B and 5C are diagrams illustrating vector synthesis of the element vector V, and FIG. ) Is an explanatory diagram of a scan segment vector SSV formed by vector synthesis of a plurality of element vectors V. FIG. FIG. 6 is a flowchart showing a procedure of the above-described planar obstacle recognition method when the scan point skip is not executed. FIG. 7 is a flowchart showing the procedure of the above-described planar obstacle recognizing method when scanning point skip is executed. FIG. 8 is a perspective view showing a wall having columns as an example of a planar obstacle. FIG. 9A is a plan view of a wall having columns shown in FIG. 8, FIG. 9B is a diagram showing a scan segment vector SSV representing the wall having columns, and FIG. It is a figure which shows one long scan segment vector SSV0 formed by combining the scan segment vectors SSV. FIG. 10 is a diagram for explaining conditions for synthesizing a plurality of scan segment vectors SSV. FIG. 11 is a flowchart of autonomous traveling using the planar obstacle recognition method of the autonomous mobile device of the present invention, including a procedure for synthesizing scan segment vectors. FIG. 12 is a perspective view showing a fence having a gap as another example of a planar obstacle. 13A is a plan view of the fence having the gap shown in FIG. 12, FIG. 13B is a diagram showing the scan point S corresponding to the fence post, and FIG. 13C is formed from the scan point S. It is a figure which shows the made element vector V and the scan segment vector SSV. FIG. 14 is a perspective view showing an example in which a reflector mark is pasted on a fence post having a gap as another example of a planar obstacle. FIG. 15A is a plan view of the fence shown in FIG. 14, FIG. 15B is a view showing the scan point S corresponding to the fence post, and FIG. 15C is formed from the scan point S. It is a figure which shows element vector V and scan segment vector SSV. FIG. 16 is a plan view for explaining how the scan point S to be processed is limited to a predetermined distance in the planar obstacle recognition method according to the present invention. FIG. 17 is a diagram for explaining a method of limiting scan points to be processed within a predetermined distance by using a change in the amount of reflected light of the laser radar depending on the distance. FIG. 18A is a plan view showing a state where the autonomous mobile device performs normal autonomous traveling, and FIG. 18B is an enlarged alert region and an enlarged alert region in the alert region and the alert region around the autonomous mobile device, respectively. It is a top view which shows a mode that it added so that it might not approach a specific obstruction. FIG. 19A is a diagram showing an example of the arrangement of planar obstacles existing in the operating area of the autonomous mobile device, and FIG. 19B is a diagram showing an example of map segment data expressing the planar obstacle. . 20 (a) to 20 (c) are diagrams for explaining vector synthesis of element vectors in the planar obstacle recognition method according to the present invention. FIG. 21 is a flowchart showing a planar obstacle recognition method according to the present invention. When a scan segment vector representing a planar obstacle is formed from scan points, the first element vector is formed by forming a scan point vector. It is a flowchart in case a skip is considered. FIG. 22 is a flowchart showing a continuation of the flowchart of FIG. FIG. 23 is a plan view showing a double structure of a planar obstacle in which a fence or fence having a gap is provided in front of a wall having no gap as another example of the planar obstacle. FIG. 24 is a plan view showing a state in which a plurality of scan segment vectors representing a planar obstacle are formed using scan points obtained from the double structure of the planar obstacle shown in FIG. FIG. 25 is a flowchart showing a procedure for forming a plurality of scan segment vectors representing a planar obstacle using scan points obtained from the double structure of the planar obstacle shown in FIG. FIG. 26 is a flowchart showing another procedure for forming a plurality of scan segment vectors representing a planar obstacle using scan points obtained from the double structure of the planar obstacle shown in FIG. FIG. 27 is a flowchart showing still another procedure for forming a plurality of scan segment vectors representing a planar obstacle using scan points obtained from the double structure of the planar obstacle shown in FIG. FIG. 28 is a plan view for explaining how the scan segment is obtained by the least square method based on the formed scan segment vector in the planar obstacle recognition method according to the present invention. FIG. 29 is a flowchart showing a procedure for obtaining a scan segment using the least square method shown in FIG. FIG. 30 is a diagram for explaining another example of a method for synthesizing element vectors when forming a scan segment vector representing a planar obstacle from a scan point in the planar obstacle recognition method according to the present invention. FIG. 31 is a diagram for explaining the determination of the validity of the scan segment vector formed by the element vector synthesis method shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Autonomous mobile device 3,30 Fence 11 Memory | storage part 12 Scan point acquisition part 13 Traveling apparatus 14 Self-position recognition part 14a Element vector formation part 14b Element vector formation part 14c Scan segment vector formation part 14d Obstacle recognition part 23, 24 Obstacle 31 Fence post 32 Fence body 33 Reflector mark 34 Block 35 Side groove 36 Object S, S _SK , S _UN , S (i) on the other side of the fence Scan point V, Va, Vb, V (i, i + 1) .. Element vector SS, SS0 Scan segment SSV, SSV0, SSV1 ... Scan segment vector Mθ First predetermined angle MΣθ Second predetermined angle ML First predetermined length TL0 Second predetermined length

Claims (16)

Scanning a given space using electromagnetic waves or sound waves at a certain angle or interval, receiving reflected waves from objects in the space, and obtaining the coordinates of multiple scan points that reflected the electromagnetic waves or sound waves Point acquisition means;
Of the plurality of scan points, an element vector having the previous scan point as the start point and the subsequent scan point as the end point is formed with respect to two scan points whose order obtained by the scan point acquisition unit is before and after in time series Element vector forming means to perform,
Among a plurality of element vectors, the order obtained by the element vector forming means is a plurality of element vectors that are continuous in time series and are continuous, and each element vector has a length equal to or less than a first predetermined length. Element vector selection means for selecting one element vector having a deflection angle formed by other element vectors continuous to the first predetermined angle or less and an integrated value of the deflection angles being a second predetermined angle or less. When,
A scan segment vector forming means for combining a plurality of selected consecutive element vectors with each other to form one scan segment vector;
A line segment defined by the start point and end point of the formed scan segment vector is compared with a second predetermined length, and when the line segment is equal to or longer than the second predetermined length, a planar obstacle along the scan segment vector With obstacle recognition means to recognize that
The element vector forming means uses an arbitrary scan point as a first scan point, and an element vector formed with a scan point continuous in time series as the second scan point is longer than the first predetermined length. In some cases, an autonomous mobile device is characterized in that an element vector is formed by using another scan point that is not continuous in time series with the first scan point as a new second scan point.   2. The autonomous mobile device according to claim 1, wherein the element vector forming means forms an element vector using only scan points within a predetermined distance from an arbitrary origin.   If the element vector forming means forms the element vector a predetermined number of times with respect to the arbitrary scan point, and the element vector of the first predetermined length or less is not obtained, the element vector forming means is time-sequential to the arbitrary scan point. The autonomous mobile device according to claim 1, wherein an element vector is formed with a continuous scan point as a new first scan point. The element vector forming means forms an element vector using a scan point that is continuous in time series with the second scan point as a third scan point,
An element vector having a second scan point as a start point and a third scan point as an end point is larger than a first predetermined angle with respect to an element vector having the first scan point as a start point and the second scan point as an end point. In this case, the third scan point is excluded, and an element vector is formed using another scan point that is not continuous in time series with the second scan point as a new third scan point. 2. The autonomous mobile device according to 2.   When one scan segment vector is formed by the scan segment vector forming means, an attempt is made to form a new scan segment vector using a scan point that was not used for forming the scan segment vector among a plurality of scan points. The autonomous mobile device according to claim 1.   When two or more scan segment vectors are formed, the angle formed by the second scan segment vector with respect to the other first scan segment vectors is equal to or less than the first predetermined angle, and the start point of the second scan segment vector Is within a predetermined distance from the end point of the first scan segment vector, the scan segment vector forming means uses the first scan segment vector end point as the start point and the second scan segment vector start point as the end point. 6. The autonomous mobile device according to claim 5, wherein a new scan segment vector is formed by combining between the scan segment vector and the second scan segment vector. The scan segment vector forming means stores information on scan points that contributed to the formation of the scan segment vector,
When the length of the line segment defined by the start point and the end point of the scan segment vector is equal to or longer than the second predetermined length, the obstacle recognition means uses the stored scan point coordinates by the least square method, The autonomous mobile device according to claim 1, wherein a line segment that optimally fits the scan point is calculated, and the line segment is recognized as a planar obstacle. Storage means for storing a map segment that is a set of line segments of planar obstacles existing on a map of an operation area of the autonomous mobile device;
It functions as a segment verification unit that compares the scan segment with the map segment, and further includes a self-position specifying unit that specifies the position of the planar obstacle on the map and the self-position of the autonomous mobile device on the map autonomous mobile apparatus according to any one of claims 1 to 7, characterized in. The self-position specifying means specifies the self-position on the map based on the position of the planar obstacle on the map and the distance and direction from the origin on the autonomous mobile device to the planar obstacle. The autonomous mobile device according to claim 8 . Scanning a given space using electromagnetic waves or sound waves at a certain angle or interval, receiving reflected waves from objects in the space, and obtaining the coordinates of multiple scan points that reflected the electromagnetic waves or sound waves A point acquisition step;
Of the plurality of scan points, an element vector having the previous scan point as the start point and the subsequent scan point as the end point is formed for two scan points whose order obtained by the scan point acquisition step is before and after in time series An element vector forming step,
Among a plurality of element vectors, the order obtained by the element vector forming step is a plurality of element vectors that are continuous in time series and are continuous, and each element vector has a length equal to or less than a first predetermined length. An element vector selection step of selecting one element vector having a deflection angle formed by another element vector continuous thereto equal to or smaller than a first predetermined angle and an integrated value of the deflection angles equal to or smaller than a second predetermined angle. When,
A scan segment vector forming step of combining a plurality of selected consecutive element vectors with each other to form one scan segment vector;
A line segment defined by the start point and end point of the formed scan segment vector is compared with a second predetermined length, and when the line segment is equal to or longer than the second predetermined length, a planar obstacle along the scan segment vector An obstacle recognition step for recognizing that
In the element vector formation step, an element vector formed with an arbitrary scan point as a first scan point and a scan point continuous in time series with the first scan point as a second scan point is longer than the first predetermined length. In some cases, the planar obstacle recognition method is characterized in that an element vector is formed by using another scan point that is not continuous in time series with the first scan point as a new second scan point. 11. The planar obstacle recognition method according to claim 10 , wherein the element vector forming step forms an element vector using only scan points within a predetermined distance from an arbitrary origin. In the element vector forming step, if an element vector of the first predetermined length or less is not obtained even if the element vector is formed a predetermined number of times for the arbitrary scan point, the time series is applied to the arbitrary scan point. The planar obstacle recognition method according to claim 10 or 11 , wherein an element vector is formed with a continuous scan point as a new first scan point. Element vector formation step, the element vectors forming the scanning point continuous time series in the second scan points as the third scanning points,
An element vector having a second scan point as a start point and a third scan point as an end point is larger than a first predetermined angle with respect to an element vector having the first scan point as a start point and the second scan point as an end point. If, with the exclusion of third scan point, the other scanning points noncontiguous time series to the second scan points as the new third scanning point, claim 10 or, characterized in that for the unit vector planar obstacle recognition method according to 11. When one scan segment vector is formed by the scan segment vector formation step, an attempt is made to form a new scan segment vector using a scan point that was not used to form the scan segment vector among a plurality of scan points. The planar obstacle recognition method according to claim 10 . When two or more scan segment vectors are formed, the angle formed by the second scan segment vector with respect to the other first scan segment vectors is equal to or less than the first predetermined angle, and the start point of the second scan segment vector Is within a predetermined distance from the end point of the first scan segment vector, the scan segment vector forming step performs the first scan with a vector having the end point of the first scan segment vector as the start point and the start point of the second scan segment vector as the end point 15. The planar obstacle recognition method according to claim 14 , wherein a new scan segment vector is formed by combining between the segment vector and the second scan segment vector. The scan segment vector formation step stores the information of the scan points that contributed to the formation of the scan segment vector,
In the obstacle recognition step, when the length of the line segment defined by the start point and the end point of the scan segment vector is equal to or longer than the second predetermined length, using the coordinates of the saved scan point, the least square method is used. The planar obstacle recognition method according to claim 10 , wherein a line segment that optimally fits the scan point is calculated, and the line segment is recognized as a planar obstacle.