Method for selecting a robot cluster leader based on field of view openness evaluation

Abstract

The application belongs to the field of robot cluster autonomous navigation, and particularly relates to a robot cluster leader selection method based on field of view openness evaluation. The method selects candidate leaders by means of a convex hull and obtains the field of view openness of each candidate leader by using a Gaussian distribution. Meanwhile, the unit vectors between the position center points of all candidate leaders and target points are calculated. According to the projection of each candidate leader and the position center point vector on the unit vector, the arrival cost of each candidate leader is obtained. The comprehensive evaluation function value of each candidate leader is obtained by weighting and summing the openness and the arrival cost, and the best leader is selected. The cluster leader switching method proposed by the application overcomes the problems of the traditional method, such as over-reliance on the initial leader and inability to cope with complex dynamic scenes. The method can select a robot with an open field of view and high target accessibility in the cluster as a new leader in a relatively short time, and can meet the demand of robot cluster rapid navigation transfer.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of robot swarm autonomous navigation, and particularly relates to a robot swarm leader selection method based on field of view openness evaluation. BACKGROUND

[0002] With the continuous development of robot technology, the application scenarios of robots are increasingly diversified, and robot swarm collaborative work has become an important research direction in the field of robots. However, in the process of robot swarm navigation, there are often problems of different degrees of field of view limitation of robots, including field of view obstruction of other robots and surrounding obstacles. Therefore, how to select a reliable leader in a complex dynamic field of view limitation environment to achieve rapid and safe transfer of the robot swarm becomes a key problem.

[0003] Some research has proposed a feedback control strategy under the condition of field of view visibility constraint, in which the leader is responsible for ensuring that the robot swarm avoids obstacles when navigating towards the target point, and the follower is responsible for ensuring the visibility maintenance with the leader and collision avoidance between vehicles. However, this strategy only considers the problem of field of view limitation of the follower in a static structured environment, and when facing complex scenes such as dead-end environments due to dynamic obstacles or corridor ends, the leader cannot complete the navigation task because most of its field of view is obstructed, so the swarm cannot achieve safe transfer. In view of this, some research has proposed a scheme to replace the leader, but it only considers the transformation of the leader according to the deviation of the formation in an open environment, which leads to great limitations of the scheme in complex environments.

[0004] Therefore, there is an urgent need for a robot swarm leader selection method that can cope with complex dynamic scenarios, which can not only ensure that the leader has an open field of view but also lead the entire robot swarm to safely and quickly transfer to the destination at a low arrival cost. SUMMARY

[0005] In order to overcome the defects of the prior art, the present application proposes a robot swarm leader selection method based on field of view openness evaluation. The method surrounds all current robots with a convex hull, selects candidate leaders according to the convex hull points, and obtains the perception field of view of each candidate leader using Gaussian distribution. The openness of the perception field of view of each candidate leader is calculated. At the same time, the unit vector between the position center point of all candidate leaders and the target point is calculated. According to the projection of each candidate leader and the position center point vector on the unit vector, the arrival cost of the destination relative to each candidate leader is obtained. The comprehensive evaluation function value of each candidate leader is obtained by weighted sum of the openness and arrival cost, and the best leader is selected.

[0006] The technical solution of the present application is as follows:

[0007] The application discloses a method for selecting a robot cluster leader based on field of view openness evaluation, and the robot cluster comprises more than two robots.

[0008] Step 1, the robot cluster is grouped according to the position coordinates of the robots, and the point set is sorted; each point in the point set is processed in the sorted order, and the upper and lower envelopes of the convex hull are respectively constructed, and the repeated points in the upper and lower envelopes are removed; the obtained convex hull is a convex polygon surrounding the robot cluster, and the robot at each vertex position of the convex polygon is taken as a candidate leader robot.

[0009] Step 2, the sensor data of the candidate leader robot is segmented, and the interference of the same type of robot is excluded to obtain the field of view angle range of each candidate leader robot; the field of view radius threshold of each candidate leader robot at different field of view angles is calculated through Gaussian distribution.

[0010] Step 3, the field of view angle range of each candidate leader robot is evenly divided into multiple unit sectors by using polar coordinates, and the unit sector field of view angles of all candidate leader robots are the same; the sensor data of each candidate leader robot is matched to each sector, and the field of view openness of each candidate leader robot is obtained according to the occupation condition of each sector under the condition that the sector is not only visible but also passable. When the field of view openness of a candidate leader robot is less than a set field of view openness threshold, the robot is no longer taken as a candidate leader robot.

[0011] Step 4, after the judgment in step 3, the position center point C of all the remaining candidate leader robots is calculated, the target point is a G point, and a unit vector in the direction is obtained; the projection of the vector from each candidate leader robot to the position center point C on the unit vector is recorded as the target point arrival cost of the candidate leader robot.

[0012] Step 5, the comprehensive evaluation function value of each candidate leader robot is calculated , wherein , are the weights of the field of view openness and the target point arrival cost of the current candidate leader robot , respectively, and the candidate leader robot with the maximum comprehensive evaluation function value is selected as the final leader.

[0013] ​​Further, in step 1, the sorting of the point set is specifically: sorting according to the numerical value of the horizontal coordinate or the vertical coordinate of the point set, and when multiple points have the same numerical value of the current sorting coordinate, another coordinate is sorted. Thus, an ordered coordinate point set is obtained, which facilitates the determination of a reasonable starting point and the order of subsequent traversal points, and facilitates the construction of the convex hull from the "outermost" point.

[0014] Further, in step 1, the sorted order is used to process each point in the point set one by one to construct the upper envelope (the right lower part of the left upper bending) of the convex hull, specifically: assuming that the current point being processed is the first point, denoted as , and the point sequence in the upper envelope that has been constructed is , where is represented by coordinates (x, y), denotes the total number of upper envelope points;

[0015] When , the last two points , and the point to be added in the two-dimensional space are calculated.

[0016] (1)

[0017] When the point set is sorted according to the coordinate value from small to large, if the calculation result is greater than 0, it indicates that the point is in the counterclockwise direction relative to the vector , which means that adding can maintain the convexity of the convex hull, so the point is added to the upper envelope, i.e., the upper envelope becomes ; if the calculation result is less than or equal to 0, it indicates that the point is in the clockwise direction relative to the vector or collinear with the two points, and because the inner angle of the convex polygon must be less than or equal to , adding the point in the clockwise direction will make the inner angle greater than , which will destroy the convexity of the upper envelope. Therefore, the last point in the upper envelope is removed, and the upper envelope becomes , and then the relative position of the last two points of the new upper envelope and the point is checked again, and this operation is repeated until the calculation result is greater than 0, and then the point is added to the upper envelope.

[0018] When the point set is sorted according to the coordinate value from large to small, the above situation is reversed, and if the calculation result is less than 0, the point is added to the upper envelope, i.e., the upper envelope becomes ; if the result is greater than or equal to 0, the last point in the upper envelope is removed, and the upper envelope becomes ; then the relative position of the last two points in the new upper envelope and point is checked again, and the operation is repeated until the result is less than 0, and then point is added to the upper envelope.

[0019] The construction of the upper envelope of the convex hull is completed by iterating each sorted point.

[0020] Further, in step 1, each point in the point set is processed in an order opposite to that of constructing the upper envelope of the convex hull, and a lower envelope (the left upper part of the right lower bend) of the convex hull is constructed, specifically: assuming that the current point being processed is the th point, denoted as , and the sequence of points in the lower envelope that has been constructed is , where is represented by coordinates (x, y), and m represents the total number of points in the lower envelope;

[0021] When , the vector cross product of the last two points , in the lower envelope and the point to be added in the two-dimensional space is calculated

[0022] (2)

[0023] When the point set is sorted in descending order of coordinate values, if the result is greater than 0, it indicates that point is in the counterclockwise direction relative to the vector , meaning that adding can maintain the convexity of the convex hull, and then point is added to the lower envelope, i.e., the lower envelope becomes ; if the result is less than or equal to 0, it indicates that point is in the clockwise direction relative to the vector , or collinear with the two points, and because the interior angle of a convex polygon must be less than or equal to , adding in the clockwise direction will make the interior angle greater than , which will destroy the convexity of the lower envelope. Then the last point in the lower envelope is removed, and the lower envelope becomes , and the relative position of the last two points in the new lower envelope and point is checked again, and the operation is repeated until the result is greater than 0, and then point is added to the lower envelope. ​

[0024] When the point set is sorted from small to large according to the coordinate value, contrary to the above situation, if the calculated result is less than 0, the point Add the next package, that is, make the next package become ; If the calculation result is greater than or equal to 0, then the last point in the package Remove, the package becomes , then check the last two points of the new package again Repeat this operation until the result is less than 0, then put the point Add the next package.

[0025] By iteratively processing each sorted point, the construction of the lower part of the convex hull is finally completed.

[0026] Furthermore, the method for constructing the convex hull adopts one of the Graham scanning method, the Jarvis stepping method, and the fast hull method.

[0027] Furthermore, in step 2, the candidate leader robot's sensor is a laser sensor with a 360° field of view. However, part of its field of view is blocked by similar objects, and this blockage is time-varying. Therefore, this part of the field of view is not helpful for the leader to observe the surrounding environment. Therefore, it is necessary to exclude this part of the field of view and only evaluate the openness of the field of view without similar blockage. This part is called the perception field of view.

[0028] The convex hull constructed in step 1 will be used for the perception field of view extraction in this step. As mentioned in step 1, each convex polygon vertex is considered a candidate leader robot. At this point, we only need to determine the exterior angle of each convex polygon. This operation can perfectly eliminate similar interference, thereby focusing on the evaluation of the external environment.

[0029] Furthermore, in step 2, the specific process of calculating the field of view radius threshold of each candidate pilot robot at different field of view angles is as follows: assuming that the vertices of the convex polygon obtained in step 1 are arranged in order to obtain a set of candidate pilot robot positions,

[0030] First calculate a candidate leader robot A Current field of view The difference from the middle angle reflects the degree to which the current angle deviates from the overall center angle.

[0031] (3)

[0032] in, , , and the candidate navigator robot A The adjacent previous candidate leader robot B and the next candidate leader robot C ;

[0033] Considering that different angles of view have different importance, the present application here adopts a threshold difference in the radius of the field of view caused by the angle difference according to the Gaussian distribution, the current angle of view The corresponding weight value is:

[0034] (4)

[0035] Wherein, is a natural constant, is the standard deviation parameter of the Gaussian distribution, which determines the "width" of the Gaussian distribution curve, that is, the sensitivity of the angle difference to the weight;

[0036] The threshold value of the radius of the field of view at the current angle of view is:

[0037] (5)

[0038] Here, and respectively represent the lower limit and upper limit setting values of the threshold value of the radius of the field of view.

[0039] After the above processing, the greater the corresponding threshold value of the part in the middle, and vice versa, showing a bell-shaped curve feature of high in the middle and low on both sides, which helps to focus more attention on the key field of view and avoid misjudgment of the overall field of view due to the interference of the edge field of view.

[0040] Further, the sensor data of the candidate leader robot is segmented, and the ground point cloud and noise are removed.

[0041] Further, step 3, the field of view openness of each candidate leader robot The specific process is:

[0042] Match the sensor data of each candidate leader robot processed in step 2 with its corresponding multiple unit sectors, judge the number of obstacle information in each unit sector, when the number exceeds the set value, the unit sector is an occupied unit sector, otherwise the unit sector is a free unit sector; When the number of continuous free unit sectors on the left and right sides of the free unit sector exceeds the target value, the free unit sector is a passable unit sector; Count the total number of unit sectors in the angle range of the field of view of each candidate leader robot and the number of passable unit sectors , then the field of view openness of each candidate leader robot .

[0043] Further, the specific process of step 4 is:

[0044] After step 3, calculate the position center points of the remaining n candidate leader robots ,Right now

[0045]

[0046] (6)

[0047] in Indicates the The two-dimensional coordinates of the candidate leader robots;

[0048] According to the target point With the center point Find the unit vector , the specific calculation is as follows

[0049]

[0050] (7)

[0051]

[0052] The vector from each candidate leader robot to the position center point C is the unit vector The projection on the target point is recorded as the target point arrival cost of the candidate leader robot ,

[0053] (8)

[0054] in , Indicates the The coordinates of the candidate leader robots.

[0055] The beneficial effect of the present invention is that the cluster leader selection method proposed in the present invention overcomes the problems of traditional methods that are too dependent on the initial navigator and cannot cope with complex dynamic scenes. It can select robots in the cluster with a wide field of view and a lower target point arrival cost as navigators, which can meet the needs of rapid and safe transfer navigation of robot clusters. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a specific flow chart of the present invention.

[0057] Figure 2 It is the effect diagram of generating a convex polygon based on the current position of the robot.

[0058] Figure 3 This is an effect diagram of the perception field of view of the candidate leader robot generated based on convex polygons.

[0059] Figure 4 is the evaluation of the open degree effect map in the range of the perceived visual field.

[0060] Figure 5 is the evaluation of the target point relative to each candidate navigator robot arrival cost effect map.

[0061] Figure 6 is the scene one navigator selection effect diagram.

[0062] Figure 7 is the scene two navigator selection effect diagram. DETAILED DESCRIPTION

[0063] The specific embodiments of the application are described in detail below in combination with the technical solutions and the accompanying drawings.

[0064] The robot cluster navigator selection method based on the visual field open degree evaluation, the robot cluster contains more than two robots, including the following steps:

[0065] Step 1, in the preparation stage, each robot obtains its absolute position based on the world coordinate system through real-time positioning and map construction, and publishes it to the robot operating system local area network through a wireless network. Each robot receives the same position and its current position, and forms a point set according to the position coordinates of each robot in the robot cluster, and sorts the point set according to the horizontal coordinate or vertical coordinate value. When multiple points have the same current sorting coordinate value, another coordinate is used for sorting. Process each point in the point set in order, respectively construct the upper and lower envelopes of the convex hull, and remove the repeated points in the upper and lower envelopes;

[0066] Among them, the upper envelope of the convex hull is constructed as follows: assuming that the current point being processed is the first point, marked as , the point sequence in the upper envelope of the convex hull that has been constructed is , where , and is represented by coordinates (x, y), , and

[0067] When , calculate the vector cross product of the last two points , and the point to be added .

[0068]

[0069] When the point set is sorted in ascending order of coordinate value, if the calculation result is greater than 0, it indicates that the point is in the counterclockwise direction relative to the vector , which means that the point is added If the convexity of the convex hull is maintained, then the point is added to the upper hull, i.e. the upper hull becomes If the result is less than or equal to 0, it indicates that the point is in the clockwise direction relative to the vector or is collinear with the two points, and because the interior angle of a convex polygon must be less than or equal to , adding the point in the clockwise direction will cause the interior angle to be greater than , which will destroy the convexity of the upper hull. The last point in the upper hull is removed, and the upper hull becomes , and the relative position of the last two points of the new upper hull and the point is checked again, and this operation is repeated until the result is greater than 0, and the point is added to the upper hull.

[0070] When the point set is sorted in descending order of coordinate values, the opposite of the above case occurs, and if the result is less than 0, the point is added to the upper hull, i.e. the upper hull becomes If the result is greater than or equal to 0, the last point in the upper hull is removed, and the upper hull becomes , and the relative position of the last two points of the new upper hull and the point is checked again, and this operation is repeated until the result is less than 0, and the point is added to the upper hull.

[0071] By iterating through each sorted point, the construction of the upper hull of the convex hull is completed.

[0072] In which, the points in the point set are processed in the opposite order of constructing the upper hull of the convex hull, and the lower hull of the convex hull is constructed, specifically: assuming that the current point being processed is the th point, marked as , and the sequence of points in the lower hull that has been constructed is , where is represented by coordinates (x, y), and m represents the total number of points in the lower hull;

[0073] When , the vector cross product of the last two points , in the lower hull and the point to be added is calculated

[0074]

[0075] When the point set is sorted in descending order of coordinate values, if the result is greater than 0, it indicates that the point​ Relative to vector In the counterclockwise direction, it means joining Can maintain the convexity of the convex hull, then put the point Add the next package, that is, make the next package become ; If the result is less than or equal to 0, it means that the point Relative to vector In the clockwise direction or collinear with these two points, and because the interior angle of a convex polygon must be less than or equal to , add in clockwise direction Points will make the interior angle greater than , will destroy the convexity of the lower bag. Then put the last point in the lower bag Remove, the package becomes , then check the last two points of the new package again Repeat this operation until the result is greater than 0, then put the point Add to the package;

[0076] When the point set is sorted from small to large according to the coordinate value, contrary to the above situation, if the calculated result is less than 0, the point Add the next package, that is, make the next package become ; If the calculation result is greater than or equal to 0, then the last point in the package Remove, the package becomes , then check the last two points of the new package again Repeat this operation until the result is less than 0, then put the point Add the next package.

[0077] By iteratively processing each sorted point, the construction of the lower part of the convex hull is finally completed.

[0078] The method for constructing the convex hull can be one of the Graham scanning method, Jarvis stepping method, and fast hull method.

[0079] The obtained convex hull is a convex polygon surrounding the robot cluster, and the robots at the vertex positions of the convex polygon are used as candidate leader robots.

[0080] Step 2: The sensor of the candidate navigator robot is a laser sensor with a 360° field of view. The sensor data of the candidate navigator robot is segmented and the ground point cloud and noise are removed. Assume that the vertices of the convex polygon obtained in step 1 are arranged in order to obtain the candidate navigator robot position set.

[0081] First calculate a candidate leader robot A current angle of view The difference between the middle angle and the current angle of view reflects the degree of deviation of the current angle of view from the overall central angle,

[0082]

[0083] wherein, , the candidate leader robot A the adjacent previous candidate leader robot B and the next candidate leader robot C ;

[0084] Considering that the fields of view at different angles have different importance, the present embodiment adopts a threshold difference in the field of view radius caused by the angle difference according to a Gaussian distribution. The current angle of view is The corresponding weight value is:

[0085]

[0086] wherein, is a natural constant, is a standard deviation parameter of the Gaussian distribution, which determines the "width" degree of the Gaussian distribution curve, that is, the sensitivity of the angle difference to the weight;

[0087] The threshold of the field of view radius of the current angle of view is:

[0088]

[0089] Here, and respectively represent the lower limit and upper limit setting values of the threshold of the field of view radius.

[0090] Step 3: uniformly divide the angle range of the field of view of each candidate leader robot into multiple unit sectors by using polar coordinates, and the unit sector field of view angles of all candidate leader robots are the same; match the sensor data of each candidate leader robot processed in step 2 with the corresponding multiple unit sectors, and judge the number of obstacle information in each unit sector. When the number exceeds a certain value, the unit sector is an occupied unit sector, otherwise the unit sector is a free unit sector; when the number of continuous free unit sectors on the left and right sides of the free unit sector exceeds a target value, the free unit sector is a passable unit sector. This adds continuous unit sector detection when calculating the openness, and only when the surrounding free sectors appear continuously, the sector is considered to be open, thereby avoiding the problem of "false openness". Count the total number of unit sectors in the angle range of the field of view of each candidate leader robot and the number of passable unit sectors , then the openness of the field of view of each candidate leader robot ; When the field of view of a candidate leader robot is less than the set field of view threshold, the robot will no longer be a candidate leader robot.

[0091] Step 4: After step 3, calculate the position center points of the remaining n candidate leader robots. ,Right now

[0092]

[0093]

[0094] in Indicates the The two-dimensional coordinates of the candidate leader robots;

[0095] According to the target point With the center point Find the unit vector , the specific calculation is as follows

[0096]

[0097]

[0098]

[0099] The vector from each candidate leader robot to the position center point C is the unit vector The projection on the target point is recorded as the target point arrival cost of the candidate leader robot ,

[0100]

[0101] in , Indicates the The coordinates of the candidate leader robots.

[0102] Step 5: Calculate the comprehensive evaluation function value of each candidate leader robot ,in , They are the current candidate navigator robots The weight of the field of view and the cost of reaching the target point is used to select the candidate navigator robot with the largest comprehensive evaluation function value as the final navigator. The effects are as follows: Figure 6 and Figure 7 As shown, after comprehensive evaluation based on the field of view width evaluation and the target point arrival cost, Figure 6 The navigator in the game is robot No. 2. Figure 7The leader in the figure is the first robot (the rhombus in the figure is the leader, the circle is the follower, and the pentagram is the target point).

[0103] At this point, the leader selection task is completed, and the detection is performed in a cycle at a frequency of 1HZ, to ensure the optimality and timeliness of the leader selection.

[0104] The above describes the preferred embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacements or changes within the technical range disclosed by the present application and according to the technical scheme and the inventive concept of the present application, which should be covered within the protection scope of the present application.

Claims

1. A robot swarm leader selection method based on field of view evaluation, characterized by: The robot cluster includes more than two robots, including the following steps: Step 1: The robot cluster is organized into a point set based on the position coordinates of each robot, and the point set is sorted. Each point in the point set is processed one by one in the sorted order, and the upper and lower hulls of the convex hull are constructed respectively. Duplicate points in the upper and lower hulls are removed. The resulting convex hull is a convex polygon surrounding the robot cluster. The robot at each vertex position of the convex polygon is selected as a candidate leader robot. Step 2: Segment the sensor data of the candidate pilot robot, eliminate interference from similar robots, and obtain the field of view angle range of each candidate pilot robot; calculate the field of view radius threshold of each candidate pilot robot at different field of view angles using Gaussian distribution; Step 3: Use polar coordinates to evenly divide the field of view of each candidate navigator robot into multiple unit sectors. The field of view of all candidate navigator robots is the same. Match the sensor data of each candidate navigator robot to each sector. According to the occupancy of each sector, under the condition that the sector is not only visible but also passable, obtain the field of view of each candidate navigator robot. ; When the field of view of a candidate leader robot is less than the set field of view threshold, the robot will no longer be a candidate leader robot; Step 4: After step 3, calculate the position center point C of all remaining candidate leader robots. The target point is point G. The unit vector of the direction, the projection of the vector from each candidate leader robot to the position center point C on the unit vector, is recorded as the target point arrival cost of the candidate leader robot ; Step 5: Calculate the comprehensive evaluation function value of each candidate leader robot ,in , They are the current candidate navigator robots The candidate leader robot with the largest comprehensive evaluation function value is selected as the final leader based on the weights of the field of view and the cost of reaching the target point.

2. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: In step 1, the point set is sorted specifically by sorting the point set according to the magnitude of the horizontal coordinate or vertical coordinate value. When multiple points have the same current sorting coordinate value, the other coordinate is used for sorting.

3. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: In step 1, process each point in the set in the arranged order and construct the upper hull of the convex hull. Specifically: Assume that the point is currently being processed. points, marked as , the point sequence in the constructed upper package is ,in Represented by coordinates (x, y), Indicates the total number of packages; when , calculate the last two points in the package , And the point to be added The vector cross product of （1） When the point set is sorted from small to large according to the coordinate value, if the calculated result is greater than 0, the point Adding the upper package will make the upper package become ; If the calculation result is less than or equal to 0, then the last point in the package Remove, the package becomes , then check the last two points of the new package again Repeat this operation until the result is greater than 0, then put the point Add to the package; When the point set is sorted from large to small according to the coordinate value, if the calculated result is less than 0, the point Adding the upper package will make the upper package become ; If the calculation result is greater than or equal to 0, then the last point in the package Remove, the package becomes , then check the last two points of the new package again Repeat this operation until the result is less than 0, then put the point Add to the package.

4. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: In step 1, the points in the point set are processed in the opposite order to the order of constructing the upper convex hull to construct the lower convex hull. Specifically, suppose that the point is currently being processed. points, marked as , the point sequence in the constructed lower package is ,in It is represented by the coordinates (x, y), and m represents the total number of places where the package is placed; when Calculate the last two points in the packet , And the point to be added The vector cross product of （2） When the point set is sorted from large to small according to the coordinate value, if the calculated result is greater than 0, the point Add the next package, that is, make the next package become ; If the calculated result is less than or equal to 0, then the last point in the package Remove, the package becomes Then check the last two points of the new package again Repeat this operation until the result is greater than 0, then put the point Add to the package; When the point set is sorted from small to large according to the coordinate value, if the calculated result is less than 0, the point Add the next package, that is, make the next package become ; If the calculation result is greater than or equal to 0, then the last point in the package Remove, the package becomes Then check the last two points of the new package again Repeat this operation until the result is less than 0, then put the point Add the next package.

5. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: In step 2, the sensor of the candidate leader robot is a laser sensor with a 360° field of view.

6. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: In step 2, the specific process of calculating the field of view radius threshold of each candidate navigator robot at different field of view angles is as follows: suppose the vertices of the convex polygon obtained in step 1 are arranged in order to obtain a set of candidate navigator robot positions, First calculate a candidate leader robot A Current field of view The difference from the middle angle, （3） in, , , and the candidate navigator robot A The adjacent previous candidate leader robot B and the next candidate pilot robot C ; Current field of view The corresponding weight values ​​are: （4） in, is a natural constant, is the standard deviation parameter of the Gaussian distribution; The field of view radius threshold for the current field of view angle is: （5） Here and Respectively represent the lower and upper limit settings of the field of view radius threshold.

7. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: Step 3: Obtain the field of view of each candidate leader robot The specific process is: Match the sensor data of each candidate navigator robot processed in step 2 with its corresponding multiple unit sectors, and determine the number of obstacle information in each unit sector. When the number exceeds the set value, the unit sector is occupied, otherwise the unit sector is a free unit sector; when the number of consecutive free unit sectors on the left and right sides of the free unit sector exceeds the target value, the free unit sector is a passable unit sector; count the total number of unit sectors within the field of view of each candidate navigator robot and the number of accessible unit sectors , then the field of view of each candidate leader robot is .

8. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: The specific process of step 4 is as follows: After step 3, calculate the position center points of the remaining n candidate leader robots ,Right now （6） in Indicates the The two-dimensional coordinates of the candidate leader robots; According to the target point With the center point Find the unit vector , the specific calculation is as follows （7） The vector from each candidate leader robot to the position center point C is the unit vector The projection on the target point is recorded as the target point arrival cost of the candidate leader robot , （8） in , Indicates the The coordinates of the candidate leader robots.

9. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: Step 2: Segment the sensor data of the candidate leader robot and remove the ground point cloud and noise.

10. The method for selecting a robot swarm leader based on field of view evaluation according to claim 1, characterized in that: Step 1: constructing a convex hull by using one of the Graham scanning method, Jarvis stepping method, and quick hull method.

Images