CN112405547A - Swarm robot multi-target searching method under unknown environment - Google Patents

Abstract

The invention discloses a swarm robot multi-target searching method under an unknown environment, which comprises the following steps: constructing an unknown environment model; the robot detects a target signal, performs dynamic division based on a target response function, and sets where robots completing the same subtask are form a sub-alliance; introducing closed-loop regulation, evaluating resource allocation of each subtask, and forming a new sub-alliance; the robots which do not form the subunion carry out roaming search; and the robots forming the sub-alliance search the target based on the particle swarm algorithm of position estimation and the obstacle avoidance strategy of boundary scanning. The obstacle avoidance strategy of boundary scanning of the invention utilizes the distance and angle relationship between the nearest two points and the boundary point and the robot to avoid the obstacle; the particle swarm algorithm for target position estimation utilizes the available target signals to deduce the approximate position of the target point, and the particle swarm algorithm is matched to quickly reach the target position, so that the path and the searching time during target searching are reduced.

Description

Swarm robot multi-target searching method under unknown environment

Technical Field

The invention relates to a robot target searching method, in particular to a swarm robot multi-target searching method in an unknown environment.

Background

The group intelligence is inspired by the group behaviors of social organisms in the nature, is the capability of solving problems in interaction of simple information processing units, is independent from one another, and is a distributed method. The swarm robot system is a swarm intelligent system with simple robots as information processing units, and has the characteristic of typical distribution, and the robot individuals have the characteristics of isomorphism and simple structure. Simple interaction exists between the robots and the environment, each robot carries out distributed motion according to information obtained through interaction, intelligent behaviors are developed on the whole robot system, and certain tasks can be completed. The swarm robot technology can be widely applied to tasks such as post-disaster search and rescue, material handling, environment detection, target capture, fire monitoring and the like, and can be widely applied to the fields such as military use, industry, civilian use and the like.

The target search is an application direction of swarm robots and can be applied to search and rescue in reality, search of certain signals, exploration of natural resources, detection of enemy positions and the like. The environmental information of multi-target search of swarm robots in unknown environments cannot be obtained in advance, the robots cannot know a global map, only obstacles, targets and the like in the detection range of the sensor can be known, and the robots can exchange respective information only in the range allowed by communication. In this field, how to improve the search efficiency by using limited information is a difficult problem to overcome.

Disclosure of Invention

In order to solve the technical problems, the invention provides the swarm robot multi-target searching method under the unknown environment, which is simple in algorithm and high in searching efficiency.

The technical scheme for solving the problems is as follows: a swarm robot multi-target searching method under unknown environment comprises the following steps:

the method comprises the following steps: constructing an unknown environment model and a target response function;

step two: the robot detects a target signal, dynamically divides work based on a target response function, searches a certain target to be regarded as a subtask, and completes the set of the robots of the same subtask to form a sub-alliance;

step three: introducing closed-loop regulation, evaluating resource allocation of each subtask, and forming a new sub-alliance;

step four: the robots which do not form the subunion carry out roaming search; and the robots forming the sub-alliance search the target based on the particle swarm algorithm of position estimation and the obstacle avoidance strategy of boundary scanning.

In the above multi-target searching method for swarm robots in unknown environment, in the first step, the process of constructing the model of unknown environment is as follows:

limited two-dimensional space R²In which there is a corpus U ═ { R ═ T ═ B }, where the swarm robots R are present_i|i＝1,2,...,i_max,i_max≥10}，R_iFor the ith robot, i_maxThe total number of the robots; target T ═ { T ═ T_j|j＝1,2,...,j_max,j_max>1}，T_jIs the jth target, j_maxIs the target total number; obstacle B ═ B_s|s＝1,2,...,s_max,s_max≥1}，B_sIs the s-th obstacle, s_maxIs the total number of obstacles; r_i、T_j、B_sThe positions are respectively represented as X_i(t)、T(x_j,y_j)、B(x_s,y_s) (ii) a In this environment, the swarm robots R are the subject of the search, with the aim of searching all the targets T to be searched, i.e., there is one robot whose distance to a certain target is less than the target arrival threshold, while avoiding collisions with obstacles and other robots.

In the above multi-target searching method for swarm robots in unknown environment, in the first step, when the swarm robots search for a target, the swarm robots detect the signal of the target in the environment uninterruptedly through their own sensors, and the signal strength is expressed by the target response function as follows:

in the formula I_ijIs a robot R_iDetecting a target T_jThe response function value of (a); m is the attenuation coefficient of the environment, and m is more than 0 and less than 1; q is the constant signal power of the target; η () represents white gaussian noise and satisfies the standard normal distribution; t is_distanceIs the Euclidean distance, L, of the robot from the target_TRepresenting the maximum target perceived radius.

In the above multi-target searching method for swarm robots under unknown environment, in the second step, when the swarm robots search targets, the total task is to search all targets, and a certain target is searched as a subtask, the set where the robots completing the same subtask are located forms a subunion, and the robots form a subunion according to the following mode:

the robots which acquire target information are divided into two types, wherein the robots which detect target signals by self sensors are marked as A-type robots, and the robots which acquire the target signals by communication among the robots are marked as B-type robots; setting the target response threshold to I_minIf the A-type robot detects signals of a plurality of targets, firstly selecting I_ij＞I_minThe target is used as an alternative target, and then a current unique subtask searching target is selected from the alternative targets according to a roulette method; roulette is represented by the following two equations:

wherein, t_iIndicating a robot R_iThe number of alternative targets; when P (i, j-1) < Rand < P (i, j), Rand is a random number which satisfies 0-1 uniform distribution, and the robot R_iSelecting a target T_jAs a search target for the subtask; other A-type robots select a target search subtask according to the same principle, and robots with the same subtask form a subunion; and the B-type robot directly joins the sub-alliance where the A-type robot with the minimum distance and already forming the sub-alliance is located in the robot communicating with the B-type robot.

In the third step, closed-loop regulation is introduced after the subtasks are allocated, namely after the initial subtask allocation is finished, the resource allocation of each subtask is evaluated; judging whether the number G of robots of each sub-alliance is larger than the maximum allowable number G or not_maxWhen G > G_maxSelecting G by the sub-task in the principle of preferred selection_maxThe individual robot is taken as a union individual of the subtasks, and the preferred principle is as follows: the A-class robot is preferentially selected, the one with strong target signal in the A-class robot is preferentially selected, and the one close to the communication robot is selected in the B-class robot.

In the fourth step, the process of roaming search by the robots not forming the subunion is as follows:

firstly, a robot motion model is constructed, target search in a two-dimensional environment is considered, and a kinematic equation of the robot is expressed as

In the formula V_i(t) is robot R_iSpeed of movement at time t, | V_i(t) | | is robot R_iMagnitude of speed of movement, theta, at time t_i(t) is robot R_iDirection of speed of movement at time t, x_i(t) and y_i(t) is the horizontal and vertical axis coordinate of velocity, i.e. V_i(t)＝(x_i(t),y_i(t)), while | | V_i(t)||≤V_max，V_maxIs the maximum value of the speed;

when the swarm robots search for the target, the search objects comprise other robots and obstacles, the search objects have virtual force on the robots and meet the virtual force application function, and the formula (5)

Wherein f is_virtualIs a virtual force, is inversely proportional to the square of the distance, d is the Euclidean distance between the robot and the object, d is the robot time₀＝2L_TWhen the object is an obstacle d₀＝L_B，L_BIs the maximum obstacle sensing radius, c is the virtual forcing function coefficient, l_kTo enhance the obstacle avoidance distance;

repulsion exists between the robot and the barrier, and attraction exists between the barrier boundary and the robot; robot R_iIn roaming state, distance R_iThe two nearest robots have repulsive force to the robots, when the obstacles are avoided, the boundary points of the obstacles attract the robots, a simplified virtual stress model is constructed, and f in the simplified virtual stress model_cirtual1And f_virtual2Is off robot R_iTwo nearest robots R₁And R₂To robot R_iVirtual force applied, f_virtual1And f_virtual2Of (a) is | | f_virtual1I and F_virtual2The | | satisfies the formula (5) and forms an angle with the horizontal axis x<f_virtual1>＝θ₁、<f_virtual2>＝θ₂；f_iIs a robot R_iThe resultant force is the vector sum of two virtual forces, and the magnitude of the resultant force is | | f_i||＝||f_virtual1+f_virtual2Angle | |, angle<f_i>＝<f_virtual1+f_virtual2>I.e. direction of desired speed during roaming, at which the robot moves at maximum speed, V_if(t+1)＝V_max[cos(<f_i>),sin(<f_i>)]I.e. the magnitude of velocity V_if(t+1)||＝V_maxDirection of rotation<V_if(t+1)>＝<f_i>，V_if(t +1) represents the expected speed of the robot at time t +1 when obstacle avoidance is not considered.

In the fourth step, the position estimation particle swarm algorithm comprises the following steps:

the target signal is inversely proportional to the square of the distance as shown in the formula (1), and the robot R_iThe distance to the target obtained by the sensor is d_iWhen there isThree robots detect a target T_iWhen the signals are received, the positions of the three robots are obtained

To the target point T_estDistance of (x, y)

Wherein i₁,i₂,i₃Represents the robot number and takes the value of the interval [10, i_max]An integer of (d); under the condition of no environmental disturbance, the target points are respectively three points (x)_i,y_i)，i＝i₁,i₂,i₃As a center of a circle, d_i，i＝i₁,i₂,i₃Is the intersection of circles of radii;

in the formula (6) (x)_oni,y_oni) Is represented by (x)_i,y_i) As a center of a circle, d_iIs a point on a circle of radius, d_iSatisfies the formula (7), i ═ i₁,i₂,i₃(ii) a The estimated point (x, y) of the target satisfies equation (8):

the common intersection point of the three circles is at most one, so the solution of equation (8) is unique; however, there is white gaussian noise in the environment, so equation (8) has no solution, and the residue norm in equation (9) is obtained by opening root numbers on both sides of equation (8) in consideration of the least square solution:

namely:

when f (x, y) takes the minimum value, namely the residual norm takes the minimum value, and then (x, y) is the result; solving for the minimum of f (x, y) and its T_estThe value-taking method of (x, y) adopts a particle swarm algorithm based on hybridization, and the step of iteration is 100 by taking 100 as the population size, thus obtaining the approximate solution T_est(x,y)；

Mapping a subtask searching process of the swarm robot with an iterative process of a particle swarm algorithm, wherein the displacement of each step of iteration is represented by the speed in the particle swarm algorithm; when the swarm robots search for the targets, the global information cannot be obtained, only the target information of the subtasks is obtained, so the global optimum is the optimum of a single sub-alliance, namely the semi-global optimum, and the target position estimation obtained by the formula (10) is added to obtain the particle swarm algorithm of the target position estimation, as shown in the formula (11):

wherein V_ipso(t +1) calculating the speed at the time of t +1, namely the speed of the next step, by using a particle swarm algorithm; v_i(t +1) is the next desired velocity after considering the inertia of the robot motion, X_i(t +1) is the next desired position, V_i(t)、X_i(t) the speed and position of the current step; c. C₁、c₂、c₃Respectively a cognitive coefficient, a social coefficient and a target position estimation coefficient; r is₁、r₂、r₃To satisfy random numbers uniformly distributed in the interval (0,1), ω is the inertial weight, α is the inertial link considering kinematics, pb_iFor individual historical optimality, sgb is semi-global optimality, T_estIs the target position estimate obtained by equation (10).

In the fourth step, the process of the obstacle avoidance strategy of boundary scanning is as follows:

let the expression of the function F (θ) be as follows:

F(θ)＝θ-2π·sgn(θ)·δ(|θ|-π)-2π＜θ＜2π (12)

wherein:

the obstacle avoidance directions are divided into two types of anticlockwise obstacle avoidance or anticlockwise obstacle avoidance, and the speed V obtained after obstacle avoidance is considered_i(t +1) when F (< V) is set_i(t+1)>-<V_if(t+1)>) Greater than 0 is clockwise obstacle avoidance, F: (<V_i(t+1)〉-〈V_if(t+1)>) If less than 0, avoiding the obstacle anticlockwise;

the method is characterized in that a target is searched in an environment with obstacles, and two situations without obstacle avoidance are provided: the first is that no obstacle is detected in the sensor scanning area of the robot, the second is that there is an obstacle in the sensor scanning area of the robot, but the next desired speed V_ifNo obstacle in the (t +1) direction;

there are two situations in which obstacle avoidance is required: the first is that a continuous section of obstacles is detected in a scanning area of a robot sensor; the second is that the robot detects a discontinuous obstacle; after the obstacle is detected, the obstacle is projected to the maximum obstacle perception radius L_BIs on the arc of radius to obtain the simplified diagram of the barrier; in the case of obstacle avoidance, the speed of the next step is finally expressed as

V_i(t+1)＝V_max(cos(a),sin(a)) (15)

Suppose P_j(x_j,y_j)、Q₁(x₁,y₁)、Q₂(x₂,y₂) Respectively represent robots R_iDetecting a boundary point, a closest point and a second closest point of the obstacle in the range; k represents the number of boundary points; l_mRepresenting a forced obstacle avoidance distance; sign represents a collision avoidance symbol, Sign is 0,1 and 2, and collision avoidance symbol 0,1 and 2 represent no collision, clockwise collision avoidance and anticlockwise collision avoidance in the previous step respectively;

the angle a is obtained by the following method:

if Q | |₁-X_i(t)||≤l_mAnd then:

b＝F(<V_if(t+1)>-<Q₁-Q₂>) (16)

wherein V_if(t+1)＝(x_i,y_i) (17)

If Q | |₁-X_i(t)||>l_mAnd then:

b₁＝<P_j1-X_i(t)> (18)

wherein P is_j1Represents the boundary point j₁Coordinate of (2), boundary point j₁Satisfies the following conditions:

when sign is 0, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝min|F|；

When sign is 1, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝max(F<0)；

When sign is 2, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝min(F>0)

(1) k 2, i.e. in the first obstacle avoidance situation described above

a＝b₁ (19)

(2) k >2, i.e. in the second obstacle avoidance situation described above

Wherein b, b₁、b₂Are all intermediate variables, boundary points j₁And j₂Adjacent without obstacles in between;

a＝<[||f₁||cos(b₁)+||f₂||cos(b₂),||f₁||sin(b₁)+||f₂||sin(b₂)]> (21)

wherein f is₁，f₂Is a boundary point j₁，j₂To robot R_iSatisfies the formula (5); at each iteration, V is obtained_iAfter (t +1), updating the obstacle avoidance flag sign according to the formula (22):

sign＝1×(F(<V_i(t+1)-V_if(t+1)>)>0)+2×(F(<V_i(t+1)-V_if(t+1)>)<0) (22)

the robot moves position through a plurality of position iterations when a certain robot and a certain target T exist_distance≤L₀When L is₀And when at least one robot of all the targets reaches the condition, the robot is considered to complete the search task.

The invention has the beneficial effects that: the invention provides an obstacle avoidance strategy of boundary scanning and a particle swarm algorithm for target position estimation, wherein the obstacle avoidance strategy of the boundary scanning is divided into two conditions, and the obstacle avoidance is carried out by utilizing the distance and angle relationship between the nearest two points and the boundary point and a robot; the particle swarm algorithm for target position estimation utilizes the available target signals to deduce the approximate position of the target point, and the target position is quickly reached by matching with the particle swarm algorithm. The method reduces the path and the searching time of the swarm robots in searching the target, and obviously improves the efficiency of the swarm robot multi-target search in the unknown environment compared with the obstacle following and avoiding method (SRSMT-SVF) based on the simplified virtual stress analysis model.

Drawings

FIG. 1 is a flow chart of the present invention.

Fig. 2 is a transition relationship diagram of the robot state.

FIG. 3 is a simplified diagram of a virtual stress model.

Fig. 4 is a schematic diagram showing the robot having an obstacle in the sensor scanning area when searching for a target, but having no obstacle in the next expected speed direction.

Fig. 5 is a schematic diagram of the robot detecting continuous convex obstacles in the sensor scanning area when searching for a target.

Fig. 6 is a schematic diagram of the robot detecting continuous non-convex obstacles in the sensor scanning area when searching for a target.

Fig. 7 is a schematic diagram of the robot detecting discontinuous convex obstacles in the sensor scanning area when searching for a target.

Fig. 8 is a simplified schematic diagram of a continuous barrier.

Fig. 9 is a simplified schematic of a discontinuous barrier.

Fig. 10 is a schematic diagram of the initialized robot and the obstacle corresponding to the boundary scan obstacle avoidance strategy during simulation.

Fig. 11 is a schematic diagram of the robot and the obstacle after initialization of the particle swarm algorithm corresponding to position estimation in simulation.

Fig. 12 is a schematic diagram of the initialized robot and the obstacle after the integration of the boundary scanning obstacle avoidance strategy and the particle swarm optimization for position estimation in simulation.

Fig. 13 is a trajectory diagram of a robot for obstacle avoidance by a boundary scan strategy, where fig. 13(a) and 13(b) are diagrams of obstacle avoidance trajectories for non-convex obstacles, and fig. 13(c) is a diagram of obstacle avoidance trajectories for convex obstacles.

Fig. 14 is an obstacle avoidance trajectory diagram of a conventional obstacle following strategy, wherein fig. 14(a) and 14(b) are obstacle avoidance trajectory diagrams for a non-convex obstacle, corresponding to fig. 13(a) and 13 (b); fig. 14(c) is an obstacle avoidance trajectory diagram for a convex obstacle, corresponding to fig. 13 (c).

Fig. 15 is a schematic diagram of the proportion of the detectable area of swarm robots in the total area in the boundary scanning strategy obstacle avoidance and the conventional obstacle following strategy. Hum and hum

Fig. 16 is a target search trajectory diagram of the particle swarm optimization using target position estimation.

Fig. 17 is a target search trajectory diagram of a particle swarm optimization with dynamics taken into account.

Fig. 18 is a diagram showing the convergence rate of the particle swarm algorithm using target position estimation.

Fig. 19 is a graph showing the convergence rate of the particle swarm algorithm in consideration of dynamics.

FIG. 20 is a comparison graph of the time spent in searching swarm robot targets combined with the particle swarm optimization using the obstacle following strategy and kinematics.

Detailed Description

The invention is further described below with reference to the accompanying drawings and examples.

As shown in fig. 1, a swarm robot multi-target search method under unknown environment includes the following steps:

the method comprises the following steps: and constructing an unknown environment model and an objective response function.

The process of constructing the unknown environment model comprises the following steps:

limited two-dimensional space R²In which there is a corpus U ═ { R ═ T ═ B }, where the swarm robots R are present_i|i＝1,2,...,i_max,i_max≥10}，R_iFor the ith robot, i_maxThe total number of the robots; target T ═ { T ═ T_j|j＝1,2,...,j_max,j_max>1}，T_jIs the jth target, j_maxIs the target total number; obstacle B ═ B_s|s＝1,2,...,s_max,s_max≥1}，B_sIs the s-th obstacle, s_maxIs the total number of obstacles; r_i、T_j、B_sThe positions are respectively represented as X_i(t)、T(x_j,y_j)、B(x_s,y_s) (ii) a In this environment, the swarm robots R are the subject of the search, with the aim of searching all the targets T to be searched, i.e., there is one robot whose distance to a certain target is less than the target arrival threshold, while avoiding collisions with obstacles and other robots. Setting parameters, wherein Euclidean distances between the robot and other robots, targets and obstacles are respectively R_distance、T_distance、B_distanceMaximum communication radius L_comaxMaximum obstacle sensing radius L_BMaximum target perception radius L_TTarget arrival threshold L₀. At R²In addition, communication can be realized between the robots under certain conditions, and the robots can detect target signalsAnd the position of the obstacle under the conditions: r_distance≤L_comax、T_distance≤L_T、B_distance≤L_B. Robot R_iPosition X of_i(t) position of the robot at iteration step number t, V_i(t) denotes a robot R_iThe iteration speed, i.e. the kinematic displacement, within the iteration step t is both represented in vector form.

When the swarm robots search for the target, the self sensors are used for uninterruptedly detecting the signal of the target in the environment, and the signal intensity is expressed by a target response function as follows:

in the formula I_ijIs a robot R_iDetecting a target T_jThe response function value of (a); m is the attenuation coefficient of the environment, and m is more than 0 and less than 1; q is the constant signal power of the target; η () represents white gaussian noise and satisfies the standard normal distribution; t is_distanceIs the Euclidean distance, L, of the robot from the target_TRepresenting the maximum target perceived radius.

Step two: the robot detects a target signal, dynamically divides work based on a target response function, searches a certain target to be regarded as a subtask, and completes the set of the robots of the same subtask to form a sub-alliance.

When the swarm robots search targets, the total task is to search all targets, a certain target is searched and is regarded as a subtask, the set where the robots completing the same subtask are located forms a subunion, and the robots form the subunion according to the following modes:

the robots capable of acquiring target information are divided into two types, wherein the robots which detect target signals by sensors are marked as A-type robots, and the robots which acquire the target signals by communication among the robots are marked as B-type robots; setting the target response threshold to I_minIf the A-type robot detects signals of a plurality of targets, firstly selecting I_ij＞I_minAs a candidate, and then bet on the roulette wheelSelecting a current unique subtask search target from the alternative targets; roulette is represented by the following two equations:

wherein, t_iIndicating a robot R_iThe number of alternative targets; when P (i, j-1) < Rand < P (i, j), Rand is a random number which satisfies 0-1 uniform distribution, and the robot R_iSelecting a target T_jAs a search target for the subtask; other A-type robots select a target search subtask according to the same principle, and robots with the same subtask form a subunion; and the B-type robot directly joins the sub-alliance where the A-type robot with the minimum distance and already forming the sub-alliance is located in the robot communicating with the B-type robot.

Step three: and introducing closed-loop regulation, evaluating resource allocation of each subtask, and forming a new sub-alliance.

After direct subtask allocation, the number of robots in the same sub-federation may be excessive, thereby wasting resources of the robots. Therefore, closed-loop regulation is introduced after the subtask allocation, namely after the preliminary subtask allocation is finished, the resource allocation of each subtask is evaluated; judging whether the number G of robots of each sub-alliance is larger than the maximum allowable number G or not_maxWhen G > G_maxSelecting G by the sub-task in the principle of preferred selection_maxThe individual robot is taken as a union individual of the subtasks, and the preferred principle is as follows: the A-class robot is preferentially selected, the one with strong target signal in the A-class robot is preferentially selected, and the one close to the communication robot is selected in the B-class robot.

For details, see Table 1, if G is taken_max6, in Table 1, the final robot R is ranked according to a preferred principle₇₉、R₈、R₈₁、R₅₉、R₉₂、R₁₅Six robots keeping coupletAlly complete subtasks, robot R₆₁、R₉₆And exiting the subtask union.

TABLE 1 subtask robot sequencing

Step four: the robots which do not form the subunion carry out roaming search and roaming search, and the diffusion search is carried out in the direction of the resultant force of the two nearest robots to the virtual repulsive force thereof according to the simplified virtual stress model, and in order to accelerate the diffusion speed of the robots | | | V_i(t)||＝V_maxI.e. take the maximum speed. And the robots forming the sub-alliance search the target based on the particle swarm algorithm of position estimation and the obstacle avoidance strategy of boundary scanning. And when the actual distance T between the robot and the target_distance≤L₀At this time, whatever the state of the robot is, the target is regarded as searched, and the state declaration is completed. The state of the robot can be divided into three types: and forming subtask search of the sub-alliance, roaming search according to simplified virtual stress, and searching the state after the target is obtained. The transition of the three states can be represented by fig. 2.

The swarm robot searches the target by reducing the length of the motion trail and the searching time as much as possible under the same condition, and the searching is carried out according to a position iteration mode when the target is searched, so the searching time is replaced by iteration steps. The invention provides an obstacle avoidance strategy of boundary scanning and a particle swarm algorithm of target position estimation so as to reduce the length of a motion track and the number of iteration steps.

The specific process of roaming search for robots not forming a subunion is as follows:

firstly, a robot motion model is constructed, target search in a two-dimensional environment is considered, and a kinematic equation of the robot is expressed as

In the formula V_i(t) is robot R_iSpeed of movement at time t, | V_i(t) | | is robot R_iMagnitude of speed of movement, theta, at time t_i(t) is robot R_iDirection of speed of movement at time t, x_i(t) and y_i(t) is the horizontal and vertical axis coordinate of velocity, i.e. V_i(t)＝(x_i(t),y_i(t)), while | | V_i(t)||≤V_max，V_maxIs the maximum value of the speed;

when the swarm robots search for the target, the search objects comprise other robots and obstacles, the search objects have virtual force on the robots and meet the virtual force application function, and the formula (5)

Wherein f is_virtualIs a virtual force, is inversely proportional to the square of the distance, d is the Euclidean distance between the robot and the object, d is the robot time₀＝2L_TWhen the object is an obstacle d₀＝L_B，L_BIs the maximum obstacle sensing radius, c is the virtual forcing function coefficient, l_kTo enhance the obstacle avoidance distance;

repulsion exists between the robot and the barrier, and attraction exists between the barrier boundary and the robot; robot R_iIn roaming state, distance R_iTwo nearest robots have repulsive force to the robots, when avoiding obstacles, the boundary points of the obstacles attract the robots, and a simplified virtual stress model is constructed as shown in fig. 3, wherein f in fig. 3_virtual1And f_virtual2Is off robot R_iTwo nearest robots R₁And R₂To robot R_iVirtual force applied, f_virtual1And f_virtual2Of (a) is | | f_virtual1I and F_virtual2The | | satisfies the formula (5) and forms an angle with the horizontal axis x<f_virtual1>＝θ₁、<f_virtual2>＝θ₂；f_iIs a robot R_iThe resultant force is the vector sum of two virtual forces, and the magnitude of the resultant force is | | f_i||＝||f_virtual1+f_virtual2Angle | |, angle<f_i>＝<f_virtual1+f_virtual2I.e. direction of desired speed during roaming, at which the robot moves at maximum speed, V_if(t+1)＝V_max[cos(<f_i>),sin(<f_i>)]I.e. the magnitude of velocity V_if(t+1)||＝V_maxDirection of rotation<V_if(t+1)>＝<f_i>。V_if(t +1) represents a desired speed of the robot at time t +1 when obstacle avoidance is not considered, and is not necessarily a moving speed at that time, V_i(t) represents the integration of the desired speed V at time t_if(t) and the moving speed of the robot after obstacle avoidance, namely the moving speed at the moment.

The procedure of the particle swarm algorithm (BSS-TPEPSO) for position estimation is as follows:

the target signal is inversely proportional to the square of the distance as shown in the formula (1), and the robot R_iThe distance to the target obtained by the sensor is d_iWhen three robots detect the target T_iWhen the signals are received, the positions of the three robots are obtained

To the target point T_estDistance of (x, y)

Under the condition of no environmental disturbance, the target points are respectively three points (x)_i,y_i)，i＝i₁,i₂,i₃As a center of a circle, d_i，i＝i₁,i₂,i₃Is the intersection of circles of radii;

in the formula (6) (x)_oni,y_oni) Is represented by (x)_i,y_i) As a center of a circle, d_iIs a point on a circle of radius, d_iSatisfies formula (7); the estimated point (x, y) of the target satisfies equation (8):

the common intersection point of the three circles is at most one, so the solution of equation (8) is unique; however, there is white gaussian noise in the environment, so equation (8) has no solution, and the residue norm in equation (9) is obtained by opening root numbers on both sides of equation (8) in consideration of the least square solution:

namely:

when f (x, y) takes the minimum value, namely the residual norm takes the minimum value, and then (x, y) is the result; solving for the minimum of f (x, y) and its T_estThe value-taking method of (x, y) adopts a particle swarm algorithm based on hybridization, and the step of iteration is 100 by taking 100 as the population size, thus obtaining the approximate solution T_est(x,y)。

Mapping a subtask searching process of the swarm robot with an iterative process of a particle swarm algorithm, wherein the displacement of each step of iteration is represented by the speed in the particle swarm algorithm; when the swarm robots search for the targets, the global information cannot be obtained, only the target information of the subtasks is obtained, so the global optimum is the optimum of a single sub-alliance, namely the semi-global optimum, and the target position estimation obtained by the formula (10) is added to obtain the particle swarm algorithm of the target position estimation, as shown in the formula (11):

wherein V_ipso(t +1) calculating the speed at the time of t +1, namely the speed of the next step, by using a particle swarm algorithm; v_i(t +1) is the next desired velocity after considering the inertia of the robot motion, X_i(t +1) is the next desired position, V_i(t)、X_i(t) the speed and position of the current step; c. C₁、c₂、c₃Respectively a cognitive coefficient, a social coefficient and a target position estimation coefficient; r is₁、r₂、r₃To satisfy random numbers uniformly distributed in the interval (0,1), ω is the inertial weight, α is the inertial link considering kinematics, pb_iFor individual historical optimality, sgb is semi-global optimality, T_estIs the target position estimate obtained by equation (10).

In multi-target search of swarm robots under unknown environment, no matter in roaming search state or subtask search, many different types of obstacles may be encountered, and there are convex or non-convex obstacles, and the set of convex obstacles, i.e. obstacle points, is a convex set, and obstacle avoidance by the convex obstacles is simple, but the non-convex obstacles are easy to fall into the problem of local optimum when avoiding obstacles. The traditional processing method considers the left or right obstacle following movement formed by the two nearest obstacle points, although the method processes the problem of obstacle avoidance of non-convex obstacles, the travel distance is long, the method classifies the situations that the robot detects the obstacles and simplifies the situations, and provides an obstacle avoidance strategy of boundary point scanning. The process of the obstacle avoidance strategy of boundary scanning is as follows:

let the expression of the function F (θ) be as follows:

F(θ)＝θ-2π·sgn(θ)·δ(|θ|-π)-2π＜θ＜2π (12)

wherein:

the obstacle avoidance directions are divided into two types of anticlockwise obstacle avoidance or anticlockwise obstacle avoidance, and the speed V obtained after obstacle avoidance is considered_i(t +1) when F (< V) is set_i(t+1)>-<V_if(t+1)>) Greater than 0 is clockwise obstacle avoidance, F: (<V_i(t+1)〉-〈V_if(t+1)>) If less than 0, avoiding the obstacle anticlockwise;

the method is characterized in that a target is searched in an environment with obstacles, and two situations without obstacle avoidance are provided: the first is that no obstacle is detected in the sensor scanning area of the robot, the second is that there is an obstacle in the sensor scanning area of the robot, but the next desired speed V_ifThere is no obstacle in the (t +1) direction, as shown in fig. 4.

There are two situations in which obstacle avoidance is required: the first is that the robot sensor detects a continuous section of obstacles in the scanning area, including convex obstacles as shown in fig. 5 and non-convex obstacles as shown in fig. 6; the second is that the robot detects a discontinuous obstacle, as shown in fig. 7; after the obstacle is detected, the obstacle is projected to the maximum obstacle perception radius L_BIs on a circular arc of a radius, a simplified diagram of the obstacle is obtained, as shown in fig. 8 and 9.

In the case of obstacle avoidance, the speed of the next step is finally expressed as

V_i(t+1)＝V_max(cos(a),sin(a)) (15)

Suppose P_j(x_j,y_j)、Q₁(x₁,y₁)、Q₂(x₂,y₂) Respectively represent robots R_iDetecting a boundary point, a closest point and a second closest point of the obstacle in the range; k represents the number of boundary points; l_mRepresenting a forced obstacle avoidance distance; sign represents a collision avoidance symbol, Sign is 0,1 and 2, and collision avoidance symbol 0,1 and 2 represent no collision, clockwise collision avoidance and anticlockwise collision avoidance in the previous step respectively;

the angle a is obtained by the following method:

if Q | |₁-X_i(t)||≤l_mAnd then:

b＝F(<V_if(t+1)>-<Q₁-Q₂>) (16)

wherein V_if(t+1)＝(x_i,y_i) (17)

If Q | |₁-X_i(t)||>l_mAnd then:

b₁＝<P_j1-X_i(t)> (18)

wherein P is_j1Represents the boundary point j₁Coordinate of (2), boundary point j₁Satisfies the following conditions:

when sign is 0, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝min|F|；

When sign is 1, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝max(F<0)；

When sign is 2, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝min(F>0)

(1) k 2, i.e. in the first obstacle avoidance situation described above

a＝b₁ (19)

(2) k >2, i.e. in the second obstacle avoidance situation described above

Wherein b, b₁、b₂Are all intermediate variables, boundary points j₁And j₂Adjacent to each other, and no barrier exists between the adjacent two, such as the boundary points of (c), (c) or (c) in fig. 9;

a＝<[||f||₁cos(b₁)+||f₂||cos(b₂),||f₁||sin(b₁)+||f₂||sin(b₂)]> (21)

wherein f is₁，f₂Is a boundary point j₁，j₂To robot R_iSatisfies the formula (5); at each iteration, V is obtained_iAfter (t +1), updating the obstacle avoidance flag sign according to the formula (22):

sign＝1×(F(<V_i(t+1)-V_if(t+1)>)>0)+2×(F(<V_i(t+1)-V_if(t+1)>)<0) (22)

the robot moves position through a plurality of position iterations when a certain robot and a certain target T exist_distance≤L₀When L is₀And when at least one robot of all the targets reaches the condition, the robot is considered to complete the search task.

Demonstration of algorithm convergence

Assuming that the total number of the robots is n and the total number of the targets is k, the swarm robots search the targets by a BSS-TPEPSO method, the set of all the iterative steps is Num, and the length of the search path is recorded as P_athThe relationship between path length and iteration step number can be expressed as:

the following discussion is given as to whether a single robot will converge to the location of the target. For any step t, all roaming steps, the resultant set is Num_r(ii) a The integration of all subtask search steps is Num_g. The single robot path length satisfies:

equation (24) indicates that the path of a single robot can be divided into a roaming search path and a subtask search path, and the roaming part of the single robot is first analyzed. As can be seen from the formula (5) and FIG. 3, if

Robot R_iMust move away from the nearest robot, and is therefore the mostThe terminal robots are uniformly distributed in the search area, and in the roaming process, signals of detected targets start to enter subtask search, so that the roaming process has no influence on whether the robots can search the targets, and only the speed of detected target information is influenced by roaming, so that only the search speed of the targets is influenced by the roaming process.

The convergence of the subtask search, i.e. of TPEPSO, is discussed below. When the robot avoids the obstacle, the speed is constant and is taken as C₄From the formula (11)

For convenient calculation, the parameters are simplified as follows:

substitution of formula (26) into formula (25) and iteration thereof once yields the following formula

Obtaining a difference equation from equations (25) and (27);

X_i(t+2)+(C₁+C₂+C₃-C₀-1)X_i(t+1)+C₀X_i(t)＝C₁pb_i+C₂sgb+C₃T_est+C₄ (28)

the characteristic equation of the second order difference equation is as shown in formula (29)

λ²+(C₁+C₂+C₃-C₀-1)λ+C₀＝0 (29)

It is obvious that formula (29) is a quadratic equation of one unity, Δ ═ C₁+C₂+C₃-C₀-1)²-4C₀Constant A_i(i ═ 0, 1.. 7) is determined by the constant in equation (28).

When is delta<At the time of 0, the number of the first,

lambda has two imaginary solutions, the solution of the difference equation being X_i(t)＝A₀+A₁λ₁ ^t+A₂λ₂ ^t

When Δ is 0, λ is ═ C₁+C₂+C₃-C₀-1)/2, λ has two real multiple roots, and the solution of the difference equation is X_i(t)＝(A₃+A₄t)λ^t

III when delta>At the time of 0, the number of the first,

lambda has two real number solutions, the solution of the difference equation being X_i(t)＝A₅+A₆λ₁ ^t+A₇λ₂ ^t

The convergence of the three cases I, II and III above is discussed, and the limit can be found:

from equation (30), the essential condition for equation convergence is: 0< | | λ | < 1; simultaneous convergence interval

So that appropriate parameters are set to satisfy the convergence condition, the TPEPSO method can converge.

Simulation and verification

In order to analyze the performance of the algorithm, firstly, the obstacle avoidance algorithm and the obstacle avoidance track of the traditional obstacle avoidance algorithm are influenced on the global diffusion rate; then comparing the search algorithm with the traditional algorithm under the condition of no barrier in an ideal environment; and finally, considering the obstacle avoidance method and the search algorithm at the same time, and comparing the target search rate with the traditional algorithm under the condition of the existence of the obstacle. When simulation comparison is performed, parameters of environment, robot, target, search process and the like need to be set, and the parameter setting is as shown in table 2:

TABLE 2 setting of parameters

The invention compares the obstacle avoidance algorithm of boundary scanning, the particle swarm algorithm of position estimation and the comprehensive search method of the two with the traditional method. After initialization, fig. 10, 11, and 12 are shown, respectively. The obstacle is represented as a shape made up of various black lines in an area of 800 × 800 in the upper right part; the target is shown as "major" in the upper right portion 800 × 800 area and "●" in the lower left corner 100 × 100 area is the robot position.

Obstacle avoidance and diffusion rate of boundary point scanning obstacle avoidance strategy

And (4) checking the obstacle avoidance algorithm of the boundary scanning, and comparing the obstacle avoidance algorithm with the traditional obstacle following algorithm. After initialization, the result is shown in fig. 10. Each robot is spread by the virtual force of the adjacent robot. After the start of diffusion, the robot trajectory for obstacle avoidance by the boundary scan strategy is shown in fig. 13, where fig. 13(a) and 13(b) are for obstacle avoidance of a non-convex obstacle, and fig. 13(c) is for obstacle avoidance of a convex obstacle. Fig. 14 shows the obstacle avoidance of the obstacle-following strategy, and it can be seen that the boundary scanning strategy greatly reduces the number of iteration steps consumed in obstacle avoidance and accelerates the obstacle avoidance speed.

In the roaming search, the target signal cannot be detected, and the total area detectable by the swarm robots needs to be as large as possible to acquire the signal of the target, so that the robots need to acquire the detection range as large as possible in as short a time as possible. Here, the ratio of the total area of the two groups of robot detection areas using the boundary scanning strategy and the obstacle-following strategy is compared, the number of robots N is set to 100, and fig. 15 is obtained, so that it is obvious that the boundary scanning strategy can be spread out more quickly, and the detection range is enough to cover the whole search area.

Convergence speed of target position estimation PSO algorithm

After the parameters are set, the robot and the target are initialized to obtain a graph 11, the robot searches the target in the searching area, the position of the target is unknown, and the target is searched only when the robot is close to the target enough, namely the distance T is obtained_distance＜L₀。

The number N of robots is set to 100, fig. 16 is a target search trajectory obtained by a particle swarm algorithm for position estimation, and fig. 17 is a particle swarm algorithm considering kinematics, so that it can be seen that after target position estimation is added, the number of iteration steps from the detection of a target signal to the search of a target by the robot is less, and the required time is shorter. Fig. 18 and 19 show the minimum distance between the cooperative robot corresponding to the target and the target, and when a certain target does not correspond to the robot sub-association, the distance is set to 100, and it can be found by comparison that the speed of the method for joining the target position is faster after the target signal is detected and the cooperative search is started until the target is finally searched. In summary, it can be concluded that the particle swarm algorithm for target location estimation can search for targets faster.

Finally, the target is searched in the environment including the obstacle, and after the parameters are set, the robot, the target position and the obstacle are initialized, so that the graph 12 is obtained. Under the initialization condition, a particle swarm algorithm combination (BSS-TPEPSO) and a cycle fault strategy are obtained by applying a boundary scanning obstacle avoidance strategy and target position estimation, and a particle swarm algorithm combination (SRSMT-SVF) of kinematics is obtained. The operation is independently performed for 100 times according to the sequence of the number of machines N being 20,30,40,50,60,70,80,90 and 100. And comparing the operation results. The comparison result is shown in FIG. 20, which clearly shows that the BSS-TPEPSO method is more excellent.

Analysis of simulation results

From the comparison, it can be concluded that the obstacle avoidance strategy of boundary point scanning and the particle swarm algorithm of position estimation are superior to the conventional obstacle following method and the particle swarm algorithm of kinematics in time consumption.

With the increase of the number scale of the robots, when the BSS-TPEPSO is increased from 20 robots to 30 robots, the efficiency is obviously increased, then the searching efficiency is only slowly increased along with the increase of the robots, and the searching rate of the SRSMT-SVF is gradually increased along with the increase of the number of the robots; but in general, the search efficiency of the former is significantly better than that of the latter.

The method considers the two-dimensional environment containing convex obstacles and non-convex obstacles in the unknown environment, and carries out multi-target search. And optimizing an obstacle avoidance method and a search algorithm, and introducing an obstacle avoidance strategy of boundary scanning and a particle swarm algorithm of position estimation. Finally, the searching speed of the verified algorithm is superior to that of the traditional algorithm.

Claims (8)

1. A swarm robot multi-target search method under unknown environment is characterized by comprising the following steps:

the method comprises the following steps: constructing an unknown environment model and a target response function;

step two: the robot detects a target signal, dynamically divides work based on a target response function, searches a certain target to be regarded as a subtask, and completes the set of the robots of the same subtask to form a sub-alliance;

step three: introducing closed-loop regulation, evaluating resource allocation of each subtask, and forming a new sub-alliance;

step four: the robots which do not form the subunion carry out roaming search; and the robots forming the sub-alliance search the target based on the particle swarm algorithm of position estimation and the obstacle avoidance strategy of boundary scanning.

2. The method for multi-object search of swarm robots in unknown environment according to claim 1, wherein in the first step, the process of constructing the model of unknown environment comprises:

limited two-dimensional space R²In which there is a corpus U ═ { R ═ T ═ B }, where the swarm robots R are present_i|i＝1，2，...，i_max，i_max≥10}，R_iFor the ith robot, i_maxThe total number of the robots; target T ═ { T ═ T_j|j＝1，2，...，j_max，j_max＞1}，T_jIs the jth target, j_maxIs the target total number; obstacle B ═ B_s|s＝1，2，...，s_max，s_max≥1}，B_sIs the s-th obstacle, s_maxIs the total number of obstacles; r_i、T_j、B_sThe positions are respectively represented as X_i(t)、T(x_j，y_j)、B(x_s，y_s) (ii) a In this environment, the swarm robots R are the subject of the search, with the aim of searching all the targets T to be searched, i.e., there is one robot whose distance to a certain target is less than the target arrival threshold, while avoiding collisions with obstacles and other robots.

3. The swarm robot multi-target search method under unknown environment as claimed in claim 2, wherein in the first step, the swarm robot searches for the target by detecting the signal of the target in the environment continuously through its own sensor, and the signal strength is expressed by the target response function as follows:

in the formula I_ijIs a robot R_iDetecting a target T_jThe response function value of (a); m is the attenuation coefficient of the environment, and m is more than 0 and less than 1; q is the constant signal power of the target; η () represents white gaussian noise and satisfies the standard normal distribution; t is_distanceIs the Euclidean distance, L, of the robot from the target_TRepresenting the maximum target perceived radius.

4. The multi-target searching method for swarm robots under unknown environment as claimed in claim 3, wherein in the second step, when the swarm robots search for the target, the total task is to search for all targets, and the search for a target is regarded as a subtask, and the set of robots that complete the same subtask forms a subunion, and the robots form a subunion according to the following way:

the robots which acquire target information are divided into two types, the robot which detects a target signal by a sensor thereof is marked as a type A robot, and the robot is obtained by communication between the robotsThe robot with the known target signal is marked as a B-type robot; setting the target response threshold to I_minIf the A-type robot detects signals of a plurality of targets, firstly selecting I_ij＞I_minThe target is used as an alternative target, and then a current unique subtask searching target is selected from the alternative targets according to a roulette method; roulette is represented by the following two equations:

wherein, t_iIndicating a robot R_iThe number of alternative targets; when P (i, j-1) < Rand < P (i, j), Rand is a random number which satisfies 0-1 uniform distribution, and the robot R_iSelecting a target T_jAs a search target for the subtask; other A-type robots select a target search subtask according to the same principle, and robots with the same subtask form a subunion; and the B-type robot directly joins the sub-alliance where the A-type robot with the minimum distance and already forming the sub-alliance is located in the robot communicating with the B-type robot.

5. The multi-target swarm robot searching method under unknown environment as claimed in claim 4, wherein in the third step, after the sub-task allocation, closed-loop adjustment is introduced, that is, after the preliminary sub-task allocation is finished, the resource allocation of each sub-task is evaluated; judging whether the number G of robots of each sub-alliance is larger than the maximum allowable number G or not_maxWhen G > G_maxSelecting G by the sub-task in the principle of preferred selection_maxThe individual robot is taken as a union individual of the subtasks, and the preferred principle is as follows: the A-class robot is preferentially selected, the one with strong target signal in the A-class robot is preferentially selected, and the one close to the communication robot is selected in the B-class robot.

6. The multi-target swarm robot searching method according to claim 5, wherein in the fourth step, the process of roaming search for robots not forming sub-alliances is as follows:

firstly, a robot motion model is constructed, target search in a two-dimensional environment is considered, and a kinematic equation of the robot is expressed as

In the formula V_i(t) is robot R_iSpeed of movement at time t, | V_i(t) | | is robot R_iMagnitude of speed of movement, theta, at time t_i(t) is robot R_iDirection of speed of movement at time t, x_i(t) and y_i(t) is the horizontal and vertical axis coordinate of velocity, i.e. V_i(t)＝(x_i(t)，y_i(t)), while | | V_i(t)||≤V_max，V_maxIs the maximum value of the speed;

when the swarm robots search for the target, the search objects comprise other robots and obstacles, the search objects have virtual force on the robots and meet the virtual force application function, and the formula (5)

Wherein f is_virtualIs a virtual force, is inversely proportional to the square of the distance, d is the Euclidean distance between the robot and the object, d is the robot time₀＝2L_TWhen the object is an obstacle d₀＝L_B，L_BIs the maximum obstacle sensing radius, c is the virtual forcing function coefficient, l_kTo enhance the obstacle avoidance distance;

repulsion exists between the robot and the barrier, and attraction exists between the barrier boundary and the robot; robot R_iIn roaming state, distance R_iTwo nearest robots to itRepulsion exists, when the obstacle is avoided, the boundary point of the obstacle attracts the robot, a simplified virtual stress model is constructed, and f in the simplified virtual stress model_virtual1And f_virtual2Is off robot R_iTwo nearest robots R₁And R₂To robot R_iVirtual force applied, f_virtual1And f_virtual2Of (a) is | | f_virtual1I and F_virtual2The | | satisfies the formula (5) and forms an angle with the horizontal axis x<f_virtual1>＝θ₁、<f_virtual2〉＝θ₂；f_iIs a robot R_iThe resultant force is the vector sum of two virtual forces, and the magnitude of the resultant force is | | f_i||＝||f_virtual1+f_virtual2Angle | |, angle<f_i>＝<f_virtual1+f_virtual2>I.e. direction of desired speed during roaming, at which the robot moves at maximum speed, V_if(t+1)＝V_max[cos(<f_i>)，sin(<f_i>)]I.e. the magnitude of velocity V_if(t+1)||＝V_maxDirection of rotation<V_if(t+1)>＝<f_i>，V_if(t +1) represents the expected speed of the robot at time t +1 when obstacle avoidance is not considered.

7. The multi-objective swarm robot searching method under unknown environment as claimed in claim 6, wherein in the fourth step, the particle swarm algorithm for position estimation comprises:

the target signal is inversely proportional to the square of the distance as shown in the formula (1), and the robot R_iThe distance to the target obtained by the sensor is d_iWhen three robots detect the target T_iWhen the signals are received, the positions of the three robots are obtained

To the target point T_estDistance of (x, y)

Wherein i₁，i₂，i₃Represents the robot number and takes the value of the interval [10, i_max]An integer of (d); under the condition of no environmental disturbance, the target points are respectively three points (x)_i，y_i)，i＝i₁，i₂，i₃As a center of a circle, d_i，i＝i₁，i₂，i₃Is the intersection of circles of radii;

in the formula (6) (x)_oni，y_oni) Is represented by (x)_i，y_i) As a center of a circle, d_iIs a point on a circle of radius, d_iSatisfies the formula (7), i ═ i₁，i₂，i₃(ii) a The estimated point (x, y) of the target satisfies equation (8):

the common intersection point of the three circles is at most one, so the solution of equation (8) is unique; however, there is white gaussian noise in the environment, so equation (8) has no solution, and the residue norm in equation (9) is obtained by opening root numbers on both sides of equation (8) in consideration of the least square solution:

namely:

when f (x, y) takes a minimum value, i.e. it is leftThe quantity norm obtains the minimum value, and at the moment, (x, y) is the minimum value; solving for the minimum of f (x, y) and its T_estThe value-taking method of (x, y) adopts a particle swarm algorithm based on hybridization, and the step of iteration is 100 by taking 100 as the population size, thus obtaining the approximate solution T_est(x，y)；

Mapping a subtask searching process of the swarm robot with an iterative process of a particle swarm algorithm, wherein the displacement of each step of iteration is represented by the speed in the particle swarm algorithm; when the swarm robots search for the targets, the global information cannot be obtained, only the target information of the subtasks is obtained, so the global optimum is the optimum of a single sub-alliance, namely the semi-global optimum, and the target position estimation obtained by the formula (10) is added to obtain the particle swarm algorithm of the target position estimation, as shown in the formula (11):

wherein V_ipso(t +1) calculating the speed at the time of t +1, namely the speed of the next step, by using a particle swarm algorithm; v_i(t +1) is the next desired velocity after considering the inertia of the robot motion, X_i(t +1) is the next desired position, V_i(t)、X_i(t) the speed and position of the current step; c. C₁、c₂、c₃Respectively a cognitive coefficient, a social coefficient and a target position estimation coefficient; r is₁、r₂、r₃To satisfy random numbers uniformly distributed in the interval (0,1), ω is the inertial weight, α is the inertial link considering kinematics, pb_iFor individual historical optimality, sgb is semi-global optimality, T_estIs the target position estimate obtained by equation (10).

8. The multi-target searching method for swarm robots in unknown environment as claimed in claim 7, wherein in the fourth step, the process of the obstacle avoidance strategy of the boundary scan is as follows:

let the expression of the function F (θ) be as follows:

F(θ)＝θ-2π·sgn(θ)·δ(|θ|-π)-2π＜θ＜2π (12)

wherein:

the obstacle avoidance directions are divided into two types of anticlockwise obstacle avoidance or anticlockwise obstacle avoidance, and the speed V obtained after obstacle avoidance is considered_i(t +1) when F (<V_i(t+1)>-<V_if(t +1) > 0 is clockwise obstacle avoidance, F: (<V_i(t+1)〉-<V_if(t+1)>) If less than 0, avoiding the obstacle anticlockwise;

the method is characterized in that a target is searched in an environment with obstacles, and two situations without obstacle avoidance are provided: the first is that no obstacle is detected in the sensor scanning area of the robot, the second is that there is an obstacle in the sensor scanning area of the robot, but the next desired speed V_ifNo obstacle in the (t +1) direction;

there are two situations in which obstacle avoidance is required: the first is that a continuous section of obstacles is detected in a scanning area of a robot sensor; the second is that the robot detects a discontinuous obstacle; after the obstacle is detected, the obstacle is projected to the maximum obstacle perception radius L_BIs on the arc of radius to obtain the simplified diagram of the barrier; in the case of obstacle avoidance, the speed of the next step is finally expressed as

V_i(t+1)＝V_max(cos(a)，sin(a)) (15)

Suppose P_j(x_j，y_j)、Q₁(x₁，y₁)、Q₂(x₂，y₂) Respectively represent robots R_iDetecting a boundary point, a closest point and a second closest point of the obstacle in the range; k represents the number of boundary points; l_mRepresenting a forced obstacle avoidance distance; sign denotes a collision avoidance symbol, Sign is 0,1,2, and collision avoidance symbols 0,1, and 2 denote no collision, clockwise collision, and inverse collision in the previous step, respectivelyAvoiding collision of the hour hand;

the angle a is obtained by the following method:

if Q | |₁-X_i(t)||≤l_mAnd then:

b＝F(<V_if(t+1)〉-<Q₁-Q₂〉) (16)

if Q | |₁-X_i(t)||＞l_mAnd then:

wherein

Represents the boundary point j₁Coordinate of (2), boundary point j₁Satisfies the following conditions:

when sign is 0, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝min|F|；

When sign is 1, | F | (<P_j1-X_i(t)〉-<V_if(t+1)〉)|＝max(F＜0)；

When sign is 2, | F | (<P_j1-X_i(t)>-<V_if(t+1)>)|＝min(F＞0)

(1) k 2, i.e. in the first obstacle avoidance situation described above

a＝b₁ (19)

(2) k >2, i.e. in the second obstacle avoidance situation described above

Wherein b, b₁、b₂Are all intermediate variables, boundary points j₁And j₂Adjacent without obstacles therebetweenAn agent;

a＝<[||f₁||cos(b₁)+||f₂||cos(b₂)，||f₁||sin(b₁)+||f₂||sin(b₂)]> (21)

wherein f is₁，f₂Is a boundary point j₁，j₂To robot R_iSatisfies the formula (5); at each iteration, V is obtained_iAfter (t +1), updating the obstacle avoidance flag sign according to the formula (22):

sign＝1×(F(<V_i(t+1)-V_if(t+1)>)＞0)+2×(F(<V_i(t+1)-V_if(t+1)〉)＜0) (22)

the robot moves position through a plurality of position iterations when a certain robot and a certain target T exist_distance≤L₀When L is₀And when at least one robot of all the targets reaches the condition, the robot is considered to complete the search task.

Images