Disclosure of Invention
The invention aims to solve the technical problem that the invention provides a group robot multi-target searching method based on a triangular pyramid, which improves each capability of a robot and simultaneously considers the necessity of turning directions, thereby reducing time consumption, energy consumption and mechanical abrasion caused by unnecessary turning processes.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
A multi-target searching method for group robots based on triangular cones comprises the steps of initializing a task environment before task searching, wherein initialization parameters at least comprise group robot system parameters, obstacle parameters and target parameters, the group robot system parameters comprise population quantity, robot position and speed information, the robots are in roaming searching states when no target information is detected, the robots develop global environments based on roaming direction cones, the robots are in collaborative searching states after target signals are detected, local searching is conducted by adopting a target search cone model, the robots are provided with avoidance strategies, the priority of the avoidance strategies is highest, the robots sense threat objects in a detection range and conduct prediction avoidance by utilizing the avoidance strategies, and when the target signals detected by the robots reach a preset threshold, the robots are regarded as successful searching, and when all targets are successful searching, the searching is completed.
The technical scheme is further improved as follows:
Preferably, the multi-objective searching method specifically includes the following steps:
Step S1, modeling
The module for constructing the group robot to perform multi-target task search in the two-dimensional task environment comprises a task environment model, a target excitation model and a task distribution model, wherein the task environment model constructs a task environment required by the multi-target task search of the group robot; the task allocation model is used for constructing resource allocation and sub-alliance formation processes of the group robots when performing target search;
step S2, group robot multi-target searching strategy based on triangular pyramid
The method comprises the steps of designing an expansion triangular pyramid model suitable for a non-convex obstacle, a roaming direction cone searching method and a target search cone searching method for searching, wherein the roaming direction cone searching method is used for developing an unknown global environment of a robot in a roaming state, the target search cone searching method is used for locally searching the robot in a cooperative state, predicting and avoiding based on collision cones and expansion triangular pyramids in combination with design and avoidance strategies, and searching by the robot according to a searching strategy to finally finish all searching targets.
Preferably, in the step S1, when the group robot performs multi-objective search in a non-convex obstacle environment, the task environment is initialized, wherein the obstacle includes a static obstacle and a dynamic non-convex obstacle, when the robot performs task search in an unknown dynamic environment, the robot selects a next moving direction according to the searched information and avoids the obstacle and other mobile robots in the searching process, when the robot searches in the unknown task environment, the robot performs objective search according to an objective excitation model, when the group robot performs task search in the dynamic environment, information exchange is performed in a communication range, and after a sub-alliance is formed, the robot cooperatively searches for the same objective.
Preferably, in the step S2, the process of designing the expansion triangular pyramid model suitable for the non-convex obstacle is to process the obstacle information collected in the detection range of the robot, set a triangular pyramid capable of completely surrounding the obstacle in the detection range, form a triangle with the robot particle and two intersection points of the triangular pyramid and the obstacle, obtain parameter information by obtaining the information of triangle inscribed circle, expand the triangular pyramid in combination with the size of the robot, and obtain the expansion triangular pyramid.
Preferably, the design is suitable for an expansion triangular pyramid model of a non-convex obstacle, comprising the steps of:
(1) Discretizing threat objects in the detection range, and obtaining corresponding coordinates:
the robot obtains its own position by the sensor, and the position of the robot at a certain moment is set asCalculation ofTo the vector of each coordinate, obtaining an included angle set between every two vectors:
Finding the maximum included angle and obtaining two corresponding coordinates: let the position of the robot at this moment beThe triangle inscribed circle formed by three points is PO＝[xo,yo, and the triangular cone threat radius is rTC;
Calculating half apex angle beta1 of triangular cone and the included angle between the relative speed and the center of inscribed circle
(2) According to the size of the robot, the triangular cone is expanded, and the part needing to be expanded is beta2:
wherein, the radius of the dimension of the robot is Rr,The Euclidean distance between the current robot Ri and the obstacle Os1、Os2 is respectively the half apex angle beta=beta1+2*β2 of the expansion triangular cone;
(3) An avoidance strategy based on collision cones and expansion triangular cones is designed, namely collision avoidance between robots uses a collision cone model, and static and dynamic non-convex barriers use an expansion triangular cone model.
Preferably, the collision cone model is constructed by taking a robot as a particle, expanding the obstacle into a threat circle with a radius r according to the threat radius of the obstacle and the size of the robot, setting the intersection point of one ray l and the threat circle at the position of the robot as M, wherein the collection of all rays with the intersection point of the ray led out from the position of the robot and the threat circle can be called as a collision cone, and the relative speed vector of the robot and the obstacle falls outside the collision cone area, so that the robot can avoid collision with the obstacle.
Preferably, the roaming direction cone searching method adopts a roaming searching strategy based on a roaming direction cone, wherein the roaming searching strategy is to increase the step length of each step of a robot and avoid the position direction before the robot, the roaming direction cone adopts all robots which avoid the maximum distance range in which the robot can detect a target, the necessity of judging the rotation direction through the roaming direction cone before the robot decides the next moving direction, and if the speed direction at the last moment falls in the feasible range of the roaming direction cone, the rotation direction is not needed.
Preferably, the target search cone searching method adopts a collaborative search strategy based on a target search cone, and comprises the following three cases:
(1) When the current position of the robot is the optimal individual or the optimal population, the robot predicts the position of the robot at the next moment according to the current direction, the position of the robot at the next moment is used as the circle center of the target search cone, and the radius of the robot is used as the radius of the target search cone;
(2) When the current position of the robot is neither the optimal individual nor the optimal population, and the optimal individual position is different from the optimal population position, the robot takes the direction of an included angle between the optimal individual and the optimal population as the range of the next searching direction;
(3) When the current position of the robot is neither individual nor population optimal but the individual optimal position is equal to the population optimal position, the robot searches for the cone radius with the individual optimal position as the center and the maximum speed of the robot as the target, so as to expand the search range.
Compared with the prior art, the group robot multi-target searching method based on the triangular pyramid has the following advantages:
(1) The group robot multi-target searching method based on the triangular pyramid aims at solving the obstacle avoidance capability problem, the global development capability problem and the local searching capability problem under the unknown dynamic non-convex obstacle environment, improves various capabilities, and simultaneously mainly considers the necessity of turning directions, thereby reducing time consumption, energy consumption and mechanical abrasion caused by unnecessary turning processes. The MSTC strategy is put forward in three aspects, for obstacle avoidance, the ETC obstacle avoidance model can better solve the obstacle avoidance problem in the complex dynamic non-convex obstacle environment, the RDC strategy effectively improves the capability of a robot for developing an unknown environment so as to improve the task search completion efficiency, and for the problem of local search capability, the TSC strategy can effectively improve the local search capability while avoiding the robot from falling into local optimum, and finally improves the target search completion efficiency. In addition, the three methods can judge the necessity of the rotation direction before the next movement, thereby reducing the rotation frequency.
(2) The multi-target searching method of the group robot based on the triangular pyramid aims at solving the collision prevention problem and the multi-target searching problem under the unknown dynamic non-convex obstacle environment, and mainly considers the necessity of direction change, so that unnecessary corner processes of the group robot in the task searching process are reduced, namely, time consumption, energy consumption and mechanical abrasion caused by the corner processes are reduced. For the problem of obstacle avoidance, the improved collision cone model provides a collision avoidance model suitable for complex dynamic non-convex obstacles, namely an Expansion Triangular Cone (ETC), and the method can not only effectively avoid the obstacles, but also predict collision conflicts to reduce unnecessary collision avoidance processes.
Detailed Description
The following describes specific embodiments of the present invention in detail. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
As shown in fig. 1 to 4, the group robot multi-target searching method based on the triangular pyramid of the present invention comprises the following steps:
Step S1, modeling
The group robot performs multi-target task search in a two-dimensional task environment and needs construction of a plurality of modules, and is mainly divided into a task environment model, a target excitation model and a task allocation model. The task environment model is mainly used for constructing task environments required by multi-target task searching of the group robots and comprises robots, targets, barriers and the like, the target excitation model is mainly used for constructing a signal model sent by a target perceived by the robots, and the task allocation model is used for resource allocation and sub-alliance formation processes of the group robots during target searching.
Step S1-1, task environment model
When the group robot performs multi-target search in a two-dimensional dynamic non-convex obstacle environment, the task environment mainly comprises three parts, namely a robot Ri{i＝1,2,…,NR, a target Tj{j＝1,2,…,NT and an obstacle Os{s＝1,2,…,NO. Wherein the obstacles comprise static and dynamic non-convex obstacles, and the task environment diagram of the group robot multi-target search is illustrated in fig. 1. The task environment is 1000×1000 units, the dots ' + ' represent the robot, the five-pointed star ' represent the target to be searched, and the gray irregular object is the obstacle to be avoided by the robot. When the robot searches tasks in an unknown dynamic environment, the next moving direction is selected automatically according to the searched information, and static and dynamic non-convex barriers and other moving robots are avoided in the exploration process. The targets are unknown, and the robot searches the targets according to the target excitation model when exploring in the task environment. The pose information of the dynamic obstacle at a certain moment follows the following movement model:
Wherein,Is the position vector of the obstacle Os at the moment t,For the coordinate position of the obstacle in a two-dimensional cartesian coordinate system,Is the velocity vector of obstacle Os at time t,For the angle between the obstacle Os and the positive direction of the x axis at time t, Vom is the maximum speed of the obstacle moving on the map.
When the group robot searches for tasks, the group robot can sense the surrounding environment according to the sensors carried by the group robot and make a response trend according to the surrounding environment. When the sensed distance is smaller than a certain threshold value, the robot selects an obstacle avoidance mechanism, and judges the next action according to the obstacle avoidance mechanism. In addition, the group robots can communicate with each other within the communication range dcom during the moving process, so that information can be exchanged to cooperatively search for the target.
When the group robot searches tasks in a dynamic environment, task targets are unknown, but the robot can sense signals sent by the targets according to sensors carried by the robot, the condition that the strength of the signals of the targets sensed by the robot is positively correlated with the distance between the robot and the targets is assumed, when the strength of the signals of the targets sensed by the robot reaches a certain threshold Il, the targets are successfully searched, and when all the targets in the task environment are successfully searched, the group robot finishes the multi-target searching tasks, and the robot stops the task searching.
Step S1-2, target excitation model
When the group robot searches tasks in a dynamic environment, the targets are unknown in advance, but signals sent by the targets can be perceived according to sensors carried by the group robot, and a detection excitation model for the signals sent by the targets is set as follows:
Wherein Iij represents a response function of the robot Ri to the target Tj, which is related to the distance d from the robot to the target and inversely proportional to the square of the distance, m represents an attenuation coefficient of the target signal when the target signal propagates in a dynamic environment, Q is constant power sent by a sensor carried by the target, rand () represents random disturbance of the target signal when the target signal propagates in the dynamic environment, and dm represents the maximum sensing distance of the target signal which can be detected by the robot.
Step S1-3, task allocation model
When the group robot searches for tasks in a dynamic environment, the group robot can exchange information in the communication range and cooperate together to quickly search for targets. Therefore, the state of the group robot when performing task search can be divided into three states, namely a roaming search state (roaming search status, RSS), a collaborative search state (collaborative search status, CSS) and a declaration state (declaration status, DC), and the migration relationship of the three states of the robot when performing task search is shown in FIG. 2. The robot can form a sub-alliance to search the same target together in the collaborative search state. If the robot has a plurality of alternative targets at the same time, the same target in the last step is preferably selected, and if the same target is not available, the current unique target is selected according to a roulette method, wherein the roulette formula is as follows:
Where ki represents the number of targets for which robot Ri is a candidate, Pij represents the probability that candidate j is selected, P (i, j) refers to the cumulative probability of candidate j, and when P (i, j-1) < Ra < P (i, j), robot Ri selects target Tj as the target for collaborative search, where Ra is a uniform random number within interval (0, 1).
Table 1 is a sub-alliance sub9 ordering mechanism at t=60, wherein class I refers to robots that can directly detect target information, class II refers to robots that can only obtain target information by communicating with class I, and the priority of robots is selected according to the sub-alliance mechanism, which is known as ：R19>R33>R6>R28>R18>R39>R34>R11>R8.
Watch 1 Subgroup sub9 alliance ranking mechanism (T=60)
Setting that when the sub-union searches the same target, if the number of sub-union members is too large, the searching area is crowded, and at this time, the robots are preferentially prevented from collision and then select to approach the target, so that the searching time can be excessively consumed due to the too large number of members, and setting that the number Nm of the sub-union members follows the formula (5):
Wherein NⅠ represents the number of class I robots, when the number of class I robots is too small, alliance members are required to be recruited in an emergency to ensure the diversity of the alliance, when the number of class I robots reaches a certain value Nm, the number of the alliance members is not required to be too large so as not to be blocked.
Step S2, group robot multi-target searching strategy based on triangular pyramid
In order to find all targets as soon as possible, the group robot performs task search in an unknown non-convex complex environment, and the number of iterations is usually reduced, so that the problems of time, energy, mechanical wear and the like consumed by each rotation direction of the robot are ignored. In the embodiment, the ETC model suitable for the dynamic non-convex obstacle, the roaming direction cone and the target search cone method for searching are provided, so that not only can static and dynamic non-convex obstacles be safely and effectively avoided, but also unnecessary collision prevention processes can be avoided in advance, and the ability of the robot to develop an unknown global environment and explore a local environment is respectively enhanced by the roaming direction cone and the target search cone, so that the three methods can improve the searching efficiency, but also effectively reduce unnecessary corner processes of the robot in moving, and further reduce time consumption, energy consumption and mechanical abrasion caused by the corner processes.
S2-1, constructing a group robot system motion control model
The group robot movement in a dynamic non-convex obstacle environment follows the following motion model:
Wherein,The position vector of the robot Ri at the moment t,For the coordinate position of the robot in a two-dimensional cartesian coordinate system,Is the velocity vector of the robot Ri at time t,And Vrm is the maximum speed of the robot moving on the map for the positive included angle between the robot Ri and the x-axis at the time t.
Step S2-2, avoiding strategy of group robot in task environment
In the research of the existing group robot multi-target searching method, in order to facilitate analysis results, most of the arranged obstacles are simple in structure, single in shape and static. However, in the actual search environment, the obstacle is mostly an uncertain factor, namely a complex non-convex and dynamic environment. The collision cone can predict the movement direction of the obstacle and can be widely applied to the collision prevention of the dynamic obstacle through simple geometric relationship judgment, but the obstacle which is prevented from being collided by the collision cone is mostly a simple and single obstacle such as a circle or a sphere.
In this embodiment, modeling is performed for a complex dynamic non-convex obstacle, and a triangular pyramid obstacle avoidance model is applied to predict the action track of a robot and the obstacle in real time so as to judge whether the robot needs to avoid the obstacle, and then the next movement is decided according to the judgment result. Combining a triangular cone model for collision avoidance of non-convex obstacles with a collision cone model for collision between robots to establish an obstacle avoidance strategy of the group robot system in a task environment. In this embodiment, in order to explain that the non-convex static obstacle, the moving obstacle and other robots detected in the obstacle avoidance range of the robot are collectively called as threats, the design of the avoidance strategy includes the following steps:
s2-2-1, constructing a collision cone model
The robot is taken as a particle, an obstacle collides into a threat circle with the radius r according to the threat radius of the obstacle and the size of the robot, the intersection point of the ray l and the threat circle is set as M, namely, the collection of all rays with the intersection point of the ray led out from the position of the robot and the threat circle is called as a collision cone, the relative speed vector of the robot and the obstacle falls outside the collision cone area, and the robot can avoid collision with the obstacle.
FIG. 3 shows a collision cone model in whichFor the angle of the relative velocity vector VRE with the position vector RiEs, RiEs represents the vector of the robot Ri to the obstacle Es,Values of RiEs projected on the x and y axes are respectively represented; Is the angle between the relative velocity vector VRE and the x-axis, wherein the relative velocity vector refers to the velocity vector of the robot RiVelocity vector with obstacle EsBeta is the half apex angle of CC, r is the threat radius of obstacle Es,Is the angle between RiEs and the x-axis. When (when)When the obstacle Es is threatening, the robot Ri needs to avoid collision with Es, whenWhen Es is not threatening, robot Ri does not need to be collision-protected. Solving the relevant parameters can be obtained:
step S2-2-2, ETC-based dynamic non-convex obstacle avoidance model
In the moving process of the robot in the complex dynamic non-convex obstacle environment, the specific position and shape of the obstacle are unknown in advance, whether the obstacle exists in the detection range or not can be judged only through a sensor carried by the robot, the detected obstacle information is returned to the robot system, then the robot system judges whether the obstacle avoidance is needed or not according to the algorithm of the robot system (the algorithm is an algorithm which is set in the running process of the robot and is a conventional setting means of the robot and is not described too much), and trend reaction is carried out according to the judgment result.
Most of the traditional collision cone obstacle avoidance methods convert threat objects into circles or ellipses with certain threat radius, but in an actual search environment, most of the obstacles are irregular patterns, and if all the obstacles are converted into the special shape of the circles or the ellipses, search resource environments are wasted excessively, so that an unnecessary collision avoidance process is caused. In this embodiment, the method of expanding the triangular pyramid is adopted, so that the problem of avoiding complex non-convex obstacles is solved, meanwhile, unnecessary avoidance processes can be effectively reduced, and fig. 4 shows an obstacle avoidance model when the robot detects obstacle information in a dynamic environment. The detection radius of the robot is dr, obstacle information collected in the detection range is processed, and a triangular cone which enables all obstacles in the detection range to be surrounded is found, so that a triangle can be formed by robot particles and two intersection points of the triangular cone and the obstacle, and related parameter information similar to a collision cone can be obtained by obtaining information of triangle inscribed circles. Considering that the robot has a certain size, combining the size of the robot to expand the triangular pyramid, and obtaining the expanded triangular pyramid.
The method specifically comprises the following steps:
step S2-2-2-1, discretizing threat objects in the detection range, and obtaining corresponding coordinates:
the robot obtains its own position by the sensor, and the position of the robot at a certain moment is set asCalculation ofVector to each coordinate, get the included angle set between every two vectors;
Finding the maximum included angle and obtaining two corresponding coordinates: the position of the robot itself is obtained by a sensor. Let the position of the robot at this moment beThe triangle inscribed circle PO＝[xo,yo formed by three points and the triangle cone threat radius rTC can be obtained.
Obtaining the half apex angle beta1 of the triangular cone, and the included angle between the relative speed and the center of the inscribed circle
In the step S2-2-2-2, considering that the robot has a size and a dimension, in order to enable the robot to safely move, the robot should expand according to the dimension of the robot when calculating the triangular pyramid, and then the parts needing to expand are:
wherein, the radius of the dimension of the robot is Rr,The euclidean distance between the current robot Ri and the obstacle Os1、Os2, respectively. Thus, the half apex angle β=β1+2*β2 of the expanded triangular cone.
S2-2-2-3, designing an avoidance strategy based on collision cone and expansion triangular cone
The precondition that the group robot completes the task is to ensure the safe movement of the group robot, so that the safe and effective robot avoidance is the primary consideration. The avoidance not only needs to avoid static and dynamic non-convex barriers, but also ensures that the robots can avoid each other safely and effectively.
The robots in this embodiment are robots with radius Rr, collision between robots can use collision cone model, and static and dynamic non-convex obstacle uses expansion triangular cone obstacle avoidance model. Fig. 5 shows an avoidance model of a robot within a detection range at a certain moment, according to the detection of surrounding threats by a sensor of the robot, calculating a next feasible range of the robot according to an expansion triangular cone or collision cone model, and randomly selecting a direction as a next speed direction within a feasible region range, wherein the radius of a CC threat for robot collision avoidance is r=2×rr.
Step S2-2-3, roaming search strategy based on RDC
The biggest reason that the group robots have lower searching efficiency in the process of moving searching in a task environment is that repeated searching is generated, and the repeated searching mainly comprises two sources, namely 1) repeated searching generated by the robots of the group robots and 2) repeated searching generated by other robots.
In this embodiment, in order to avoid repeated searches generated by the robot, that is, when the robot performs task searches in a dynamic environment, there is a situation that the same robot repeatedly performs searches in the same area, and in this regard, the roaming direction cone is solved by increasing the step length of each step of the robot and avoiding the position direction before the robot, so that during the roaming search, the robot will move with the maximum step length and avoid the position at the previous moment.
In this embodiment, in order to avoid repeated searches generated by other robots, that is, the area that the robot has explored may be an area that the other robot has explored, the roaming direction cone is set to avoid all robots within the range of dm, where dm is the maximum distance that the robot can detect the target, and avoid the robot within the range of dm can prevent from missing the target, and also can make the robots uniformly distributed in the developed area. The range of avoidance angles of the robots is related to the distance between the robots, i.e. the smaller the distance the greater the range of avoidance, where d0＝Vrm+Rr. Before the robot decides the next moving direction, the necessity of the rotating direction can be judged in advance through the roaming direction cone, and if the speed direction at the last moment falls in the feasible range of the roaming direction cone, the rotating direction is not needed.
As shown in fig. 6, the roaming direction cone of the robot Ri at time t is the roaming direction cone center of the robot at time t-1, the robot at the maximum speed Vrm can avoid surrounding robots, the measure is that the robot takes the position at time t-1 as the roaming direction cone center, the half apex angle is 90 degrees, the positions of the robot Rj1、Rj2、Rj3 at time t are respectively taken as the center, and the half apex angle beta* follows the formula (16).
The robot speed update formula in roaming state is:
where RDC represents the roaming direction cone.
Step S2-2-4, collaborative search strategy based on target search cone
In the conventional robot, a particle swarm algorithm is mostly adopted when multi-target searching is performed, however, the particle swarm is easy to sink into local optimum, the convergence speed is not fast enough, and other factors. In this embodiment, based on a multi-target search strategy of a target search cone, when a robot performs task search in a dynamic non-convex obstacle environment, the state of the robot is changed from a roaming state to a collaborative search state when a signal sent by a target is detected.
The target search cone is based on the area contained between the optimal direction of the group and the optimal direction of the individual as the search range of the next step, so that the group robot can quickly approach the target and simultaneously maintain the diversity of the group. The method is mainly divided into two aspects, namely a direction and a step size. In the direction, the target direction range is predicted and the robot turns to the target direction range, the predicted target direction range does not point to a specific factor, in the collaborative searching process, in order to ensure the diversity of the population, the members in the alliance generally comprehensively consider the inertia factors, the cognitive factors and the social factors, and the target search cone also respectively considers the factors according to different conditions in the group robot searching process. The method comprises the following three cases:
(1) When the current position of the robot is the optimal individual or the optimal population, the inertial factors of the robot are considered more, so the robot firstly predicts the position of the robot at the next moment according to the current direction, takes the position as the center of a target search cone, and takes the radius of the robot as the radius of the target search cone, as shown in fig. 7 a).
(2) When the current position of the robot is neither the optimal individual nor the optimal population, and the optimal individual position is different from the optimal population position, the robot comprehensively considers the cognitive factors and the social factors, namely, the direction of an included angle between the optimal individual and the optimal population is taken as the range of the next searching direction, as shown in fig. 7 b).
(3) When the current position of the robot is neither optimal for the individuals nor optimal for the population but the optimal position of the individuals is equal to the optimal position of the population, as shown in fig. 7 c), the cognitive factors and the social factors are the same at the moment, and only one of the cognitive factors and the social factors needs to be considered. In order to expand population diversity, the robot searches for the cone radius by taking the optimal position of an individual as the center of a circle and taking Vrm as the target so as to expand the searching range.
According to the method, the speed direction of the robot falls into the range of the target search cone and approaches the target, and the direction is randomly selected in a certain range instead of being singly approaching one direction each time, so that the robot can avoid sinking into local optimum and increase population diversity. In addition, the target search cone can be used for judging whether the rotation direction is needed or not so as to reduce unnecessary rotation angle process.
Further, in the multi-target searching process of the group robot, too long step length may result in missing targets, causing unnecessary repeated searching and increasing energy consumption, while too short step length may result in lengthening the searching time consumption. In this embodiment, a nonlinear decreasing step function based on a target response signal is adopted, and the step is related to the detected target signal, when the detected target signal is smaller, the step should be increased, and when the detected target signal is larger, the step should be decreased, so that the detected target is avoided. The nonlinear decremental step function based on the target response signal is:
Wherein Il is a set target signal threshold, Iij (T) is the signal intensity of the robot Ri and the target Tj in the collaborative search state at the moment T, and the range of l is controlled at (lmin,lmax).
The robot speed update formula in roaming state is:
Where TSC represents the target search cone.
The displacement update formula of the group robot in the searching process in the two-dimensional environment:
step S2-3, multi-target search control strategy of group robot
As shown in fig. 8, a multi-objective searching flow chart of the group robot in the two-dimensional unknown complex dynamic non-convex obstacle environment is shown. Before task searching, the task environment is initialized, including the parameters of a group robot system, obstacles, targets and the like, wherein the parameters of the group robot system comprise population quantity, robot position and speed information, the group robot keeps a roaming searching state before detecting no target information and enables the robot to develop a global environment quickly based on RDC, after detecting target signals, the robot is turned into a collaborative searching state and performs local searching by using a TSC model, an obstacle avoidance process penetrates through the whole process and has the highest priority, namely, an avoidance strategy combining collision cones and expansion triangular pyramid is adopted to predict avoidance whenever the robot detects threat objects within the range of dm, when the target signals detected by the robot reach a preset threshold, the targets are regarded as successful searching, and when all targets are successful searching, the searching is completed.
Method verification
The search method in this embodiment is verified, and includes the following steps:
step 1, establishing a system parameter setting and a group robot task search performance evaluation index model, wherein the task environment parameter setting is shown in table 2.
Table 2 task environment parameter settings
| Sign symbol | Meaning of the following description | Setting value |
| NR | Number of group robots | 30~100 |
| NT | Target quantity | 10 |
| NO | Number of obstacles | 7 |
| S | Search area (unit) | 1000×1000 |
| Rr | Radius of robot | 2 |
| Vrm | Maximum speed of robot operation | 10 |
| Vom | Dynamic obstacle running speed | 1 |
| dm | Maximum radius of induction of target and threat | 30 |
| dcom | Maximum communication distance of robot | 300 |
| Q | Target signal energy | 105 |
| m | Target signal attenuation coefficient | 0.1 |
| Il | Target signal threshold | 400 |
| nm | Class I robot count threshold | 4 |
| lmax | Maximum value of step length | 10 |
| lmin | Step minimum | 5 |
The performance of task search of the group robot is evaluated from 2 aspects, namely, the system time consumption Tr and the rotation angle number Na, wherein the system time consumption is the iteration number when the task is completed, and the rotation angle number is the number of times that the robot changes the direction at the last moment in the searching process. Let Tr1、Tr2、Tr3,Na1t、Na2t、Na3t,NR1t、NR2t、NR3t be the time consumption, the number of corners and the number of robots when the robot is at the roaming time, the collaborative search time and the declaration time at time T, respectively, wherein NR|t＝NR1|t+NR2|t+NR3|t,tk is the time consumption of successful search of a single target Tk, and then Tk, Tr,Na and Na1|t、Na2|t、Na3|t have the following relationship:
NRk is the number of sub-alliance members formed by the search target Tk, and since the robot in the state of statement does not participate in other search tasks, the speed is 0, and the position remains unchanged, the robot in the state of statement does not affect the rotation angle number Na of the subsequent task search, namely Na3 =0.
Step 2, verification test
In the searching process of the two-dimensional unknown dynamic environment, complex static and dynamic non-convex barriers are considered, and NR =50 is taken as an example, and the multi-target searching demonstration experiment process of the group robot based on the multi-target searching strategy (MSTC) is specifically described, as shown in fig. 9.
Step 2.1, simulation experiment process based on expansion triangular pyramid (ETC)
Fig. 9 (c), 9 (d) demonstrate the ETC-based evasion process. In fig. 9 (c), the robot R12 detects the obstacle information in the detection range thereof, and determines that the obstacle is encountered if the robot continues to travel in the current direction according to ETC, so that R12 calculates the feasible range thereof according to ETC and performs obstacle avoidance. In fig. 9 (d), the robot R6 detects obstacle information, but it is judged that no collision occurs according to the present time according to ETC, so that the robot R6 does not need to change direction, whereas R36、R44 is judged that no collision occurs according to the present time direction according to ETC when obstacle information is detected, so that R36、R44 calculates a feasible range according to ETC and randomly selects one direction to move in the feasible range, and similarly, R45 is judged that no collision occurs according to ETC when obstacle information is detected, so that the robot R45 changes direction according to ETC, and R49 is judged that no collision occurs, so that no direction change is required.
Step 2.2, simulation experiment procedure based on Roaming Direction Cone (RDC)
Initially, the robots are at 100 x 100 units, their locations are random, and the current state is the roaming search state, so they will be spread around according to the RDC method to develop an environment, as shown in fig. 9 a). Fig. 9 (b) is a process of global environment development by the robot in a roaming state at time t=30-40 according to RDC, when the robot is rapidly developing an unknown environment, and the target 9 has been successfully searched by the robot R23. Fig. 9 (c) shows a process of developing a global environment according to RDC by a robot in a roaming state during t=60-70, and it can be seen that the robot is performing development and exploration to an unknown environment. The strategy can traverse the search space at the maximum speed without acquiring prior information, and has strong unknown environment development capability.
Step 2.3, simulation experiment procedure based on Target Search Cone (TSC)
Fig. 9 (d) shows a process of searching for the target 10 based on TSC with t=70 to 89. When t=70, the robot R38、R16 detects the signal sent by the target 10 as a class I robot and rapidly forms a sub-union sub10, and at this time, the number of sub-union members is smaller than Nm, so that the sub10 emergently issues recruitment information to the surroundings, and the robot R8、R15 receives the information and joins the sub-union sub10 as a class II robot and searches the target 10 together according to the TSC policy. When t=89, the robot R15 in the sub-alliance sub10 successfully searches the target 10, the sub10 is broken down, and the rest members are converted into a roaming search state to continue searching other robots.
Fig. 9 (e) shows a search process of sub-union sub7 of t=140 to 167 according to the TSC policy. When t=140, the robot R34 detects a signal sent by the target 7 and interacts with the robot in the communication range, R6、R11、R44 forms a sub-alliance sub7 with R34 through interaction, and searches the target 7 together by using the TSC policy, and when t=167, the robot R34 successfully searches the target 7, and the sub-alliance sub7 is dismissed.
At t=229, the robot R11 detects the signal sent by the target 4, and R6、R44 receives the information sent by the class I robot R11 and forms a union sub4 together with it to search for the target 4, as shown in fig. 9 (f). When t=251, the target response signal of the robot R44 reaches the threshold, which is regarded as that the target 4 search is successful, as shown in fig. 9 (g), and all the target searches are completed at this time, which marks that the search task is completed.
Step 2.4, simulation experiment process of group robot multi-target search strategy (MSTC strategy) based on triangular pyramid
Fig. 9 (h) shows the whole path of the robot for finding the target position in the whole searching process, and the static and dynamic obstacles in the figure are the final states when the task is completed, and although the robot part track shown in the figure is overlapped with the obstacle, the robot does not touch the obstacle in the actual running process of the simulation experiment, and the robot collision preventing process in fig. 9 (c) and (d) can be seen specifically. In fig. 9 (h), the robot performs global environment development based on RDC in roaming state, through RDC, it can predict whether there is a need to change direction in current direction before deciding next moving direction, when detecting the target, it judges feasibility of current direction based on TSC, if current direction falls within TSC range, it does not need to change direction, so that the robot searches the target more quickly. For example, the robot R23、R49 that searched for the target 9 and the target 8 quickly searched for the target without excessive rotational directions based on the TSC judgment criteria. And the rest of the robots select the non-rotation direction to quickly approach the target in the case that the current direction is judged to be within the range of the feasible region. In the moving process, besides the threat of the obstacle, collision threats from other robots are also included, and the priority of avoiding the threat object is highest, namely, when the threat object is encountered, whether collision avoidance is needed is judged by using ETC first, and then the next moving direction is decided, so that the safe movement of the robot is ensured. Therefore, other robots in fig. 9 (h) also have an unavoidable rotational direction process.
Step 2.5, simulation contrast experiment
In this embodiment, six sets of comparative experiments were set to analyze two performance indicators of multi-objective search in a dynamic non-convex obstacle environment, as shown in table 3. Wherein SVF, LSDN, KCPSO is a simplified virtual stress obstacle avoidance method, a straight line search Mode different from that of a neighbor, and a particle swarm optimization algorithm with kinematic constraint, ETC, RDC, TSC is an obstacle avoidance method, a roaming search strategy and a collaborative search strategy, respectively, wherein Mode6 is a multi-target search strategy (MSTC strategy) formed by combining the three methods in the embodiment.
Table 3 Multi-target search mode for group robot in two-dimensional unknown dynamic environment
| Mode | Obstacle avoidance strategy | Roaming search strategy | Collaborative search strategy |
| Mode1 | SVF | LSDN | KCPSO |
| Mode2 | SVF | RDC | KCPSO |
| Mode3 | SVF | RDC | TSC |
| Mode4 | ETC | LSDN | KCPSO |
| Mode5 | ETC | RDC | KCPSO |
| Mode6 | ETC | RDC | TSC |
According to the number of robots 20, 40, 60, 80 and 100, simulation experiments are independently carried out for 30 times in the six groups of modes, the searching time and the rotation angle times of the robots after each running are recorded, and the average value, the maximum value and the minimum value are obtained, and the obtained data are shown in table 4.
Table 4 comparison of performance of different scale group robot systems in 6 modes
With the population size NR as the horizontal axis and Tr、Na as the vertical axis, as shown in fig. 10, as the population number NR increases, the search time Tr decreases gradually, and the number of turns is related to the number of iterations in addition to the population size.
Step 2.6, analysis of simulation results
As can be seen from fig. 10, the number of rotation angles increases with the time consumption, and thus the number of rotation angles is divided by the time consumption in the subsequent data comparison.
Step 2.6.1 ETC comparison result analysis
Three sets of comparison experiments are established, namely Mode1 and Mode4, mode2 and Mode5, and Mode3 and Mode6, and ETC and SVF are respectively compared under the condition that the roaming search strategy and the collaborative search strategy are the same. For Mode1 and Mode4, when the roaming search strategy is MSTC and the collaborative search strategy is KCPSO, the number of corners of SVF is reduced by 0.04% -5.64% compared with ETC, the time consumption is reduced by 0.73% -6.13%, the number of corners of Mode2 and Mode5 is respectively reduced by 0.19% -10.09%, 0.81% -6.13%, and the number of corners of Mode3 and Mode6 is respectively reduced by 1.68% -8.68% and 3.50% -13.85%.
When the group robots search tasks, the optimal route originally planned can be abandoned due to threat object avoidance, and the ETC can not only safely avoid the threat object, but also predict the avoidance necessity, reduce the unnecessary collision avoidance process, and reduce the time consumption and the corner frequency to a certain extent.
Step 2.6.2, analysis of RDC comparative results
In order to compare the characteristics of RDC in search performance, two sets of comparison experiments, namely Mode1 and Mode2, and Mode4 and Mode5, are established. These two sets of comparison experiments were performed with the same collision avoidance strategy as the collaborative search strategy and a different roaming strategy. As can be seen from the Mode1 and the Mode2 in the table 4, when the obstacle avoidance strategy and the collaborative search strategy are SVF and KCPSO and the roaming strategy is changed from LSDN to RDC, the search efficiency is improved by 1.02% -13.32%, and the rotation frequency is reduced by 24.96% -52.45%. For Mode4 and Mode5, when the obstacle avoidance strategy and the collaborative search strategy are ETC and KCPSO and the roaming strategy is changed from LSDN to RDC, the search efficiency is improved by 1.4-13.46%, and the rotation frequency is reduced by 21.13-54.52%.
The roaming searching strategy based on RDC can uniformly explore undeveloped environments under the condition of ensuring that targets are not missed, so that the searching capability of the global environment is improved to a certain extent, and the efficiency of completing the whole task searching is further improved. Because the RDC can judge whether the current direction falls in a feasible area in advance and then decide the next movement, the number of rotation angles is reduced to a great extent, and further the time consumption, the energy consumption and the mechanical abrasion caused by the rotation angle process are reduced.
Step 2.6.3 TSC contrast results analysis
To compare the features of TSC in search performance, two sets of comparative experimental analyses were performed on TSC and KCPSO. For Mode2 and 3, when the obstacle avoidance strategy and the roaming strategy are SVF and RDC, and the collaborative search strategy is changed from KCPSO to TSC, the efficiency is improved by 2.6 to 11.18 percent, and the rotation frequency is reduced by 8.89 to 18.60 percent. For Mode5 and Mode 6, when the obstacle avoidance strategy and the roaming strategy are ETC and RDC, and the collaborative search strategy is changed from KCPSO to TSC, the efficiency is improved by 10.61-22.9%, and the rotation frequency is reduced by 3.68-22.91%;
The collaborative search strategy based on TSC can ensure diversity of population and simultaneously rapidly approach to the target, and effectively improves the local search capability of the robot. The TSC direction strategy enables the robot to randomly select any direction within a feasible range before deciding the next movement, so that population trapping is avoided to be locally optimal, and the TSC step size strategy effectively controls the step size of each step of the robot according to a nonlinear decreasing step size function based on a target response signal, so that the robot cannot miss a target due to too large step size or increase time consumption due to too short step size in the movement process, and therefore the robot can find the target faster. In addition, the TSC can judge the necessity of the rotation direction in advance, so that the optimum direction of the searching target can not be missed, and the rotation frequency can be reduced.
Step 2.6.4, MSTC analysis of comparative results
Compared with the Mode1, the Mode6 has the advantages that the searching efficiency is improved by 15.93-28.31%, and the rotation frequency is reduced by 34.51-61.18%. Because ETC can safely and effectively avoid complex non-convex barriers, RDC is more beneficial to developing an unknown environment, the local searching capability of TSC is more outstanding, and ETC, RDC, TSC can judge the superiority and inferiority of the current direction before deciding the next movement, so that the problems of missing the original optimal direction due to the rotation direction, time loss, energy loss, mechanical abrasion and the like caused by the rotation angle process of the robot can be avoided to a certain extent. As is apparent from fig. 10, mode6 is superior to Mode1 in terms of improving the search efficiency and reducing the number of corners. Therefore, the MSTC strategy of the invention is more beneficial to multi-target searching of the group robot than the other three strategies.
The above embodiments are merely preferred embodiments of the present invention, and are not intended to limit the present invention in any way. While the invention has been described with reference to preferred embodiments, it is not intended to be limiting. Therefore, any simple modification, equivalent variation and modification of the above embodiments according to the technical substance of the present invention shall fall within the scope of the technical solution of the present invention.