Multi-robot system target searching method based on block chain

Abstract

The invention relates to the technical field of target search, in particular to a block chain-based multi-robot system target search method, which comprises the following steps of: establishing a multi-robot shared network in a block chain, and forming a multi-robot system; establishing a grid map; when the position of the target point is known, the optimal path of the robot is obtained through an improved A * target algorithm; when the position of the target point is unknown, drawing a target probability map; in the target probability map, judging whether the moving direction falls into a local optimum problem or not; and if the robot is trapped in the local optimum problem, the robot search grid area range is expanded until the robot is not trapped in the local optimum problem, and the robot leaves the obstacle through a D * Lite target search path algorithm so as to obtain the optimal path of the robot. According to the method, the requirement of the robot on the computing power can be reduced, the target searching efficiency is improved, and the target point searching purposiveness of a multi-robot system is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of target search technology, and in particular to a multi-robot system target search method based on blockchain. Background Art

[0002] At present, multi-robot systems are used in many fields, such as search, rescue, and surveillance. A multi-robot system can be composed of two or more single robots, mainly to complete tasks in the same workspace. Multi-robot systems can be divided into many types, such as homogeneous multi-robot systems and heterogeneous multi-robot systems, centralized control multi-robot systems and distributed control multi-robot systems, etc. Most of them are classified according to the functions and control methods of multi-robot systems. From the perspective of its virtual appearance, blockchain technology is composed of "blocks" + "chains". "Blocks" are transactions between nodes. Data information is packaged into blocks in the form of Merkle trees. "Chains" are connected together in some form according to the order of generation. Therefore, the structure of the multi-robot system has a certain degree of fit with the blockchain. It can solve the problem of false information released by malicious nodes in the multi-robot system, so that the system can have the ability to distinguish, and can resist data attacks on the multi-robot system by hackers behind the scenes that may lead to catastrophic results, so that the multi-robot system can perform tasks safely and reliably.

[0003] The existing multi-robot target search algorithms are mainly expanded from single-robot target search algorithms. In the process of robot target search paths, the most ideal state is that the paths of multiple robots are distributed as much as possible in the area to be searched, and the total energy consumed by the multi-robot system path can be kept to a minimum. According to the characteristics of multi-robot target search, target search algorithms can be divided into two categories: classic target search algorithms and bionic target search algorithms. Classic algorithms include grid-based target search algorithms and sampling-based target search algorithms. Bionic target search algorithms are mainly inspired by biological populations in nature, including particle swarm algorithms, genetic algorithms, and ant colony algorithms. Most other bionic algorithms are based on improvements or derivatives of these algorithms.

[0004] In the target search algorithm of multi-robot system, although the principle of bionic target search algorithm is simple and the adjustment parameters are few, it requires the robot to have certain computing power. For multi-robot system, using bionic target search algorithm requires the multi-robot system to have a strong sense of integrity. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a multi-robot system target search method based on blockchain, which can reduce the demand for robot computing power, improve the efficiency of target search, and enhance the purposefulness of the multi-robot system's search for target points.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A multi-robot system target search method based on blockchain, comprising the following steps:

[0008] S1. Build a shared network of multiple robots on the blockchain so that robots can obtain data through the blockchain and form a multi-robot system;

[0009] S2. One of the robots establishes a grid map through SLAM-Gmapping to represent the location of the real scene and obstacles through the grid map, and the robot binarizes the grid map and uploads it to the blockchain;

[0010] S3. When the position of the target point is known, the cost of the robot from the initial point to the target point is estimated by the improved A* target algorithm to obtain the optimal path of the robot;

[0011] S4. When the position of the target point is unknown, the robot predicts the target point through the target probability function of the grid map, so that the multi-robot system can draw a target probability map on the grid map through the different perception strengths of the target signal by the robot;

[0012] S5. In the target probability map, the robot calculates the moving direction according to the search benefit function and determines whether the moving direction falls into a local optimal problem;

[0013] S6. If the robot falls into the local optimal problem, the search grid area is expanded and the search benefit function is calculated again to carry out the next step of target search path planning. After the robot does not fall into the local optimal problem, the robot leaves the obstacle through the D*Lite target search path algorithm, and after leaving the obstacle, the robot calculates the moving direction according to the search benefit function until it searches near the target point to obtain the optimal path of the robot;

[0014] S7. In the target probability map, if the robot cannot determine the location of the target point, it returns to step S5 to retry. When the number of retries is greater than a threshold, the robot search task is forcibly terminated.

[0015] Furthermore, in step S1, each node of the robot generates a pair of public-private key pairs in the private blockchain, the public key is uploaded to the blockchain and broadcast to other blockchain nodes through the blockchain, and the robot stores the private key in its own safe location; the robot hashes the plaintext data through a hash algorithm to obtain a hash value, and the robot encrypts the hash value through its own private key and sends it to the blockchain.

[0016] Furthermore, in step S2, the robot performs hash calculation on the binarized grid map to obtain a hash value, and the robot signs the binarized hash value of the grid with its own private key and uploads it to the blockchain.

[0017] Furthermore, in step S3, the improved A* target algorithm estimates the cost from the target point to the end point through an evaluation function, and the evaluation function is:

[0018] f(n)=g(n)+(1-lgp)h(n)+(1+lgp)T(n) Formula (1)

[0019] Among them, g(n) is the evaluation function; g(n) is the actual movement cost consumed by the robot to move from the starting point to the current position node n; P is the obstacle rate of the grid map; h(n) is the heuristic function; T(n) is the number of bends in the path from the starting point to the current node n.

[0020] Furthermore, the obstacle rate calculation method of the grid map is:

[0021]

[0022] Where S is the number of obstacle grids in the grid map; (x b ,y b ) is the starting point of the robot; (x e ,y e ) is the end point of the robot.

[0023] Further, in step S4, when the position of the target point is unknown, the position of the target point is relative to any grid in the environment grid map where the robot may appear, and the target probability function is:

[0024]

[0025] Among them, f(x, y) is the probability that the target point is located at the grid (x, y); φ(x, y) is the target point position estimated by multiple robots; Ω is the entire task environment.

[0026] Furthermore, the method for estimating the target point position by multiple robots is:

[0027]

[0028] Among them, (x i ,y i ) is the position of the multi-robot in the grid map; (x, y) is the estimated point of the target point position in the grid map; μ is a constant; k is the number of multi-robots.

[0029] Further, in step S5, at a certain time t, the robot is located at position N in the grid map. After time t+1, the robot moves to position N+1 at this time. The search benefit function is expressed as φ(N+1,t+1). The search benefit function calculation method is:

[0030] φ(N+1,t+1)=0.6I A (N+1,t+1)-0.1I B (N+1,t+1)+0.2I C (N+1,t+1) Formula (5)

[0031] Among them, I A (N+1, t+1) is the target gain obtained by the robot when it goes to grid N+1 at time t+1; I B (N+1,t+1) is the aggregation cost of the robot; I C (N+1, t+1) is the exploration benefit of the robot; when an obstacle appears between the target point and the robot, and the target point is close to the robot, the robot falls into a local optimal problem.

[0032] Further, in step S6, a function Rhs is introduced into the D*Lite target search path algorithm to represent the estimated value from the target point to the point where the robot falls into the local optimal problem:

[0033]

[0034] Among them, k(s) is the basis for the order of node detection, indicating the priority amount that the robot processes first, and k(s) includes two values ​​[k1(s); k2(s)]; g(s) is the actual value that the robot needs to consume from the local optimal problem point to the target point; h(s,s start ) is the estimated value from the point where the robot falls into the local optimal problem to the starting point; k m Used to rationalize node sorting calculations; when k1(s) = k2(s), the robot target search path planning is completed.

[0035] Furthermore, the estimated value from the target point to the point where the robot falls into the local optimal problem is expressed as:

[0036]

[0037] Where S is the point where the robot falls into the local optimal problem; S goal is the target point; c(s, s′) is the value consumed by the robot moving from the local optimal problem point to the adjacent node;

[0038] According to formula (7), the estimated value from the point where the robot falls into the local optimal problem to the starting point is calculated as:

[0039]

[0040] Among them, h(s′,s goal ) is the value consumed in moving from the adjacent node of the local optimal problem point where the robot is trapped to the target point.

[0041] The beneficial effects of the present invention are:-

[0042] The traditional A* target search algorithm can find the optimal solution of the path in a limited space, but the time it takes to calculate the path becomes longer as the space increases. The present invention improves the planning efficiency of the path by changing the heuristic function. The improved algorithm can adapt to different obstacle environments, so that the target search efficiency of the robot can be improved, and the number of path bends can be reduced to a certain extent, thereby improving the speed of the multi-robot system's global search for targets. The evaluation function of the present invention mainly depends on the computing power of the individual robot, and it is necessary to perform multi-robot target search under the premise of known target points. Therefore, the present invention solves the problem of the robot falling into the local optimal situation by constructing a target probability map and introducing the D*Lite algorithm on the basis of the target probability map, which not only improves the search efficiency of the multi-robot system, but also enhances the multi-robot system's purposefulness in searching for target points. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a preferred implementation of the present invention.

[0044] Figure 2 It is a flow chart of data information sharing verification of a multi-robot system according to a preferred embodiment of the present invention.

[0045] Figure 3 It is a flow chart of a preferred implementation of the present invention for solving the problem of a robot falling into a local optimum.

[0046] Figure 4 It is a path trajectory comparison diagram based on a multi-robot system target search algorithm according to a preferred embodiment of the present invention.

[0047] Figure 5 This is an initial position diagram of multiple robots and target points according to a preferred embodiment of the present invention.

[0048] Figure 6 It is a target probability map of a preferred embodiment of the present invention.

[0049] Figure 7 This is a target probability map when the number of iterations is 50 according to a preferred embodiment of the present invention.

[0050] Figure 8 The present invention is a multi-robot target probability search trajectory diagram according to a preferred embodiment of the present invention.

[0051] Fig. 9 It is a schematic diagram of map construction of a real environment according to a preferred embodiment of the present invention.

[0052] Fig.10 It is a diagram of the initial positions and target point positions of multiple robots in a preferred embodiment of the present invention.

[0053] Fig.11 It is a trajectory diagram of a multi-robot system based on an improved A* target search algorithm in a preferred embodiment of the present invention.

[0054] Fig.12 It is a target probability map of multiple robots at the initial moment of a preferred implementation of the present invention.

[0055] Fig.13 It is a search trajectory diagram of a multi-robot system based on a target probability map algorithm according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0058] Please also see Figure 1 As shown in the figure, a multi-robot system target search method based on blockchain according to a preferred embodiment of the present invention comprises the following steps:

[0059] S1. Build a shared network of multiple robots on the blockchain so that robots can obtain data through the blockchain and form a multi-robot system.

[0060] In step S1, each robot's node generates a public-private key pair in the private blockchain. The public key is uploaded to the blockchain and broadcast to other blockchain nodes through the blockchain. The robot stores the private key in its own safe location. The robot uses a hash algorithm to hash the plaintext data to obtain a hash value, and the robot encrypts the hash value with its own private key and sends it to the blockchain.

[0061] In this embodiment, the multi-robot system data information sharing verification is as follows Figure 2 As shown in the figure, robot a node generates a pair of public and private key pairs in the private blockchain, where the public key is uploaded to the blockchain and broadcast to other blockchain nodes through the blockchain, and robot a stores the private key in its own safe location; in order to reduce the excessive block storage space occupied by robot data information, robot a uses a hash algorithm to hash the plaintext data to obtain a hash value, and then encrypts the hash value with its own private key and sends it to the blockchain. Only robot c is shown in the figure because robot c and robot b perform the same operation. When robots b and c in the system obtain the encrypted data information through the blockchain, robots b and c obtain the public key of robot a node on the blockchain, decrypt the data information and obtain the hash value; robots b and c obtain the plaintext data of robot a through traditional data transmission and calculate the hash value, and determine whether the data has been tampered with during the transmission process by judging whether the hash values ​​are equal. If the two hash values ​​are different, it can be said that the plaintext data sent by robot a has been tampered with, and robots b and c that receive the plaintext data will automatically remove the data.

[0062] The multi-robot system uses the public and private key pairs of blockchain technology to prevent malicious robot nodes from joining. The robot's sensor data and other information are signed with its private key and stored on the blockchain, and the robot obtains data through the blockchain. Uploading various information of the robot to the blockchain can ensure the security of the data information of the multi-robot system, and at the same time ensure the reliability of the data information source, providing secure data information interaction for the target search of the multi-robot system.

[0063] S2. One of the robots builds a grid map through SLAM-Gmapping to represent the real scene and the location of obstacles through the grid map, and the robot binarizes the grid map and uploads it to the blockchain.

[0064] In step S2, the robot performs hash calculation on the binarized grid map to obtain a hash value, and the robot signs the hash value of the binarized grid with its own private key and uploads it to the blockchain. The binarized grid map is hashed to obtain a hash value, which can reduce the storage space occupied by the image data information.

[0065] Robot a will pass the binary grid map to robot b and robot c through the ROS communication module. After receiving the binary image, robot b and robot c will perform hash calculation to obtain the hash value of the binary image. At the same time, the public key of robot a will be used to obtain and decrypt the hash value data encrypted by robot a from the blockchain. After obtaining the two hash values, robot b and robot c will compare whether they are equal. If they are equal, robot b and robot c will accept the binary image. If not, the binary image will be discarded. Robots a, b, and c will share sensor data information in the same way.

[0066] S3. When the position of the target point is known, the cost of the robot from the initial point to the target point is estimated by the improved A* target algorithm to obtain the optimal path of the robot.

[0067] In step S3, the improved A* target algorithm estimates the cost from the target point to the end point through an evaluation function, and the evaluation function is:

[0068] f(n)=g(n)+(1-lgp)h(n)+(1+lgp)T(n) Formula (1)

[0069] Among them, g(n) is the evaluation function; g(n) is the actual movement cost consumed by the robot to move from the starting point to the current position node n; P is the obstacle rate of the grid map; h(n) is the heuristic function; T(n) is the number of bends in the path from the starting point to the current node n.

[0070] The obstacle rate calculation method of the raster map is:

[0071]

[0072] Where S is the number of obstacle grids in the grid map; (x b ,y b ) is the starting point of the robot; (x e ,y e ) is the robot’s end point.

[0073] By changing the weight of the heuristic function, the improved A* target search algorithm can adjust the weights of the heuristic function and the curvature function at any time in the multi-robot system according to the different environmental obstacles. This can not only improve the efficiency of the target search path, but also reduce unnecessary path curvatures, thereby improving the speed of the global target search of the multi-robot system.

[0074] S4. When the position of the target point is unknown, the robot predicts the target point through the target probability function on the grid map, so that the multi-robot system can draw a target probability map on the grid map through the different perception strengths of the robots to the target signal.

[0075] In step S4, when the position of the target point is unknown, the position of the target point is relative to any grid in the environment grid map where the robot may appear, and the target probability function is:

[0076]

[0077] Among them, f(x, y) is the probability that the target point is located at the grid (x, y); φ(x, y) is the target point position estimated by multiple robots; Ω is the entire task environment.

[0078] The method for multi-robots to estimate the target point position is:

[0079]

[0080] Among them, (x i ,y i ) is the position of the multi-robot in the grid map; (x, y) is the estimated point of the target point in the grid map; μ is a constant; k is the number of multi-robots.

[0081] Based on the different perceptions of target signal strength, the robot predicts the target points in the grid map through the target probability function, so that the multi-robot system can draw a target probability map based on the grid map through the different perception strengths of the robots to the target signal.

[0082] In this embodiment, the relationship between the target position and the target release signal strength is:

[0083]

[0084] Among them, I d is the strength of the target release signal that can be sensed by the robot; Q is the strength of the target release signal, which is usually replaced by a constant; T d is the maximum distance the robot can estimate to the target.

[0085] When the robot is searching for a target, the relationship between the target signal strength and the distance to the target position perceived by the robot is usually unknown. Formula (9) is an unsolvable formula. If the robot can infer the distance between the target and the robot by the signal strength, it needs to know the relationship between the two. Therefore, by keeping a certain distance between the robot and the target, the robot can just sense the weak signal released by the target. The distance at this time is recorded as d0. The straight line d0 is evenly divided into τ segments. For the accuracy of the experimental data, the average value between each segment τ and the signal strength is calculated through multiple measurements. Finally, the line segment d is obtained by interpolation calculation method. ξ The average target signal strength I ξ The relationship between the robot's target distance T d With the target signal strength Iij The approximate functional relationship between them is shown in formula (10):

[0086]

[0087] Among them, I d Release signals for targets that the robot can perceive, I ξ with I ξ+1 are two adjacent interpolation nodes.

[0088] The straight-line distance between the robot and the target point can be roughly estimated by formula (11). Assume that there are three robots in the grid map, and their three positions are (x1, y1), (x2, y2), and (x3, y3). The target point position at this time is T j (x j ,y j ), then the distances from the three robots to the target point are d1, d2, and d3. When there is no external interference in the environment, the positions of the three robots are taken as the centers of the circles, and the distances from the three robots to the target point are taken as the radii. The intersection of the three circles is the approximate location of the target point.

[0089] Target point estimated position T est (x,y) satisfies formula (11):

[0090]

[0091] When there is no solution to formula (11), the robot's search environment is disturbed by external factors. In order to obtain an approximate solution to the equation, the least squares solution of the equation can be calculated, as shown in formula (11):

[0092]

[0093] Right now:

[0094]

[0095] Where T(x,y)≈T est (x, y) = arg min (φ (x, y)) is the approximate estimated position of the target point. Formula (12) is a specific method for estimating the approximate position of the target point by the three robots. The method for obtaining the target point position estimated by the multiple robots in this embodiment according to formula (12) is shown in formula (4).

[0096] S5. In the target probability map, the robot calculates the moving direction according to the search benefit function and determines whether the moving direction falls into the local optimal problem.

[0097] In step S5, at a certain time t, the robot is located at position N in the grid map. After time t+1, the robot moves to position N+1. The search benefit function is expressed as φ(N+1, t+1). The calculation method of the search benefit function is:

[0098] φ(N+1,t+1)=0.6I A (N+1,t+1)-0.1I B (N+1,t+1)+0.2I C (N+1,t+1) Formula (5)

[0099] Among them, I A (N+1, t+1) is the target gain obtained by the robot when it goes to grid N+1 at time t+1; I B (N+1,t+1) is the aggregation cost of the robot; I C (N+1, t+1) is the exploration benefit of the robot; when an obstacle appears between the target point and the robot, and the target point is close to the robot, the robot falls into a local optimal problem.

[0100] S6. If the robot falls into the local optimal problem, the search grid area will be expanded and the search benefit function will be calculated again to carry out the next step of target search path planning. After the robot does not fall into the local optimal problem, it will leave the obstacle through the D*Lite target search path algorithm. After leaving the obstacle, the robot will calculate the moving direction according to the search benefit function until it searches near the target point to obtain the optimal path of the robot.

[0101] In step S6, the function Rhs is introduced into the D*Lite target search path algorithm to represent the estimated value from the target point to the point where the robot falls into the local optimal problem:

[0102]

[0103] Among them, k(s) is the basis for the order of node detection, indicating the priority of the robot to process first, and k(s) includes two values ​​[k1(s); k2(s)]; g(s) is the actual value consumed by the robot from the local optimal problem point to the target point; h(s,s start ) is the estimated value from the point where the robot falls into the local optimal problem to the starting point; k m Used to rationalize node sorting calculations; when k1(s) = k2(s), the robot target search path planning is completed.

[0104] The estimated value from the target point to the point where the robot falls into the local optimal problem is expressed as:

[0105]

[0106] Among them, S is the point where the robot falls into the local optimal problem; S goal is the target point; c(s,s′) is the value consumed by the robot moving from the local optimal problem point to the adjacent node;

[0107] According to formula (7), the estimated value from the point where the robot falls into the local optimal problem to the starting point is calculated as:

[0108]

[0109] Among them, h(s′,s goal ) is the value consumed by the robot moving from the adjacent node to the target point when it falls into the local optimal problem point.

[0110] The D*Lite algorithm uses a reverse search method, so it calculates the estimated value from the target point to the end point based on the Rhs (right-hand side) function, and uses the Rhs function to calculate the estimated value of the 8-domain grid adjacent to the target point, and then determines whether k1 and k2 are equal. If not, it uses the estimated minimum value in multiple nodes as the center, and so on until k1 and k2 are equal; if they are equal, the algorithm ends, and the robot's target search path ends. The D*Lite algorithm is well suited for target search algorithms in unknown environments. Due to its reverse search path method, the planning efficiency of the target path is improved.

[0111] In this embodiment, the problem of the robot falling into the local optimum is solved by expanding the search grid range and introducing the D*Lite algorithm based on the traditional search benefit function. The flow chart is as follows: Figure 3 shown.

[0112] S7. In the target probability map, if the robot cannot determine the location of the target point, it returns to step S5 to retry. When the number of retries is greater than a threshold, the robot search task is forcibly terminated.

[0113] In order to prove the effectiveness of the improved A* target search algorithm in this embodiment, the following experiment is conducted:

[0114] The improved A* target search algorithm is compared with the traditional A* target search algorithm, reference 1 (Xu Wei, Tang Junwei, Zhang Chi. Improved A* and dynamic window algorithm for mobile robot path planning [J]. Computer Simulation, 2023, 40(03): 447-452.), reference 2 (

[63]

[0115] Li B, Dong C, Chen Q, et al. Path planning of mobile robots based on animproved A* algorithm[C]. Proceedings of the 20204th High Performance Computing and Cluster Technologies Conference&20203rd International Conference on BigData and Artificial Intelligence. 2020: 49-53.) conducted comparative experimental simulations to verify the effectiveness of the improved A* target search algorithm. The experimental environment is to perform target search simulation in MATLAB2016a, such as Figure 4 As shown in the figure, the three blue circles represent the initial positions of the robots, the red five-pointed star represents the target search position of the multi-robot, and the black square represents the position of the obstacle.

[0116] As shown in Table 1, the experimental data comparison results of the average path planning time, average number of turns and average path length of the four target search algorithms in multi-robots under grid maps are shown.

[0117] Table 1 Experimental results of 4 target search algorithms

[0118]

[0119] As can be seen from Table 1, the path planning time based on the improved A* target search algorithm is shortened by an average of 42.94% compared with the traditional A* target search algorithm, and is shortened by 29.18% compared with Reference 1. Although it is slightly higher than Reference 2, its path planning length is reduced by 1.95% compared with Reference 2. The average number of bends in the improved algorithm is reduced by 50.00% compared with Reference 1, and increased by 24.98% compared with Reference 2. The experimental results show that the improved method proposed in this embodiment may not have an advantage over other target search algorithms in terms of time consumption, but from the perspective of the length of path planning and the number of bends, the improved A* algorithm achieves path shortening and reduces the number of path bends, meeting the expectations of improving the traditional A* target search algorithm.

[0120] This embodiment is improved on the basis of the traditional A* target search algorithm to solve the problems that the target search path is prone to bifurcation and falling into local optimality due to the increase of obstacles. The key to the improvement is to consider the influence of the proportion of obstacles in the grid map on the weights of the heuristic function and the curvature function. The weight will change with the different proportions of the environment occupied by obstacles between the starting point and the end point, which not only improves the efficiency of the target path search, but also optimizes the number of bends in the search target path.

[0121] In order to verify the effectiveness of the improved multi-robot target probability map search algorithm, the following experiments were conducted:

[0122] The simulation experiment was conducted using MATLAB 2016a. The environment space is a 30*30 grid. Unlike the experimental environment in Chapter 3, the target location is unknown when the robot is searching for a target. However, in the map based on target probability, the robot can sense the approximate distance between itself and the target location through sensors. Multiple robots can determine the approximate direction of the target location through information interaction.

[0123] In the simulation experiment, Figure 5 As shown, multiple robots are represented by three blue dots, and the location of the target point is represented by a red five-pointed star.

[0124] At the initial moment, the multi-robots draw the target probability map based on the perception of the signal strength released by the target. Figure 6 As shown, Figure 6 (a) is the target probability map stereogram, Figure 6 (b) is the target probability map plane. The red dots represent the initial positions of the robots, while the brighter areas indicate that the multi-robots perform the target probability map algorithm and infer the approximate location of the target area.

[0125] When the number of iterations of the improved target probability map search algorithm is 50, the target probability map is as follows Figure 7 As shown, through Figure 6 (b) Compared with the target probability map, the brightness of the target area has decreased, indicating that the multi-robots gradually determine the location of the target area.

[0126] According to the continuous update and iteration of the multi-robot target probability map, until the multi-robot finds the target point. In order to verify the effectiveness of the improved target probability map search algorithm, Figure 8 Shown are the multi-robot trajectory diagrams of the traditional target probability map search algorithm and the improved target probability search algorithm.

[0127] from Figure 8It can be seen that the multi-robot target search algorithm based on the improved target probability map successfully found the target, and the robot path trajectory avoided falling into the local optimal problem. The traditional algorithm that has not been modified failed the robot's target search task due to the limitations of the search benefit function.

[0128] In order to test the performance of the improved algorithm of this embodiment, the algorithm is compared with the unimproved traditional algorithm, document 3 (Ruan Xiaogang, Zhou Jing, Zhang Jingjing, et al. Robot goal-oriented RRT path planning algorithm based on sub-goal search [J]. Control and Decision, 2020, 35(10): 2543-2548.2019.0043.), document 4 (Huang Y, Li M, Zhao T. A Multi-robot Coverage Path Planning Algorithm Based on Improved DARP Algorithm [J]. arXiv preprint arXiv: 2304.09741, 2023.) and the improved A* algorithm of this embodiment by simulation experimental data. The experimental results are shown in Table 2.

[0129] Table 2 Experimental comparison of target search algorithms

[0130]

[0131] Under the same experimental environment, the improved target probability map search algorithm used in this embodiment increases the average target search path planning length by 2.98% compared to the improved A* algorithm of this embodiment, but shortens the average target search path planning length by 28.23% compared to document 3, and shortens the average target search path planning length by 11.48% compared to document 4. Although the target search algorithm of this embodiment does not perform as well as other algorithms in terms of target search time, the improved target probability map search algorithm still has great advantages from other data. It can be seen from its target search path trajectory that it has a strong purpose in searching for targets.

[0132] This embodiment is experimented in a real scene. In this embodiment, a ROS robot intelligent car is used, which is equipped with an IMU inertial test unit, a high-performance radar and a chassis controller. The main controller of the ROS car is Jetson Nano, and the motion controller uses STM32F103 as a microcontroller to control the movement of the car.

[0133] In this embodiment, the Fisco Bcos alliance chain is used as the blockchain platform, and the Jetson Nano car and computer are used as blockchain nodes to form a blockchain network to achieve the security of robot data information interaction and realize information data sharing between robots through blockchain technology.

[0134] The experimental site of this embodiment is selected as a relatively empty site indoors, such as Fig. 9 (a) shows the SLAM-Gmapping composition result. Fig. 9 (b) shows; the raster map after binarization is as follows Fig. 9 (c) as shown.

[0135] The initial positions of the multiple robots and the target point are as follows: Fig.10 As shown, according to the improved A* target search algorithm in step S2, the multi-robot target search trajectory diagram is obtained as shown in Fig.11 As shown. Fig.11 The robots in the multi-robot system can accurately find the target by using the improved A* target search algorithm. The path trajectory diagrams of the three robots show that the planned paths of the multiple robots do not intersect and the three robots do not collide with obstacles.

[0136] The target probability map obtained in step S4 is as follows: Fig.12 As shown, according to the algorithm of steps S5 and S6, the search trajectory diagram is obtained, such as Fig.13 shown.

[0137] from Fig.11 , Fig.13 It can be found that the robot's actual target search path is different from the experimental simulation path trajectory. The main reason for this is that there is an error in the robot's position odometer IMU.

[0138] In order to verify whether the blockchain technology has an impact on the path planning time, path length, and path traversal times of the multi-robot target search algorithm with the improved target probability map, Table 3 shows a comparison of the target search data of multiple robots with and without blockchain technology.

[0139] Table 3 Comparison of multi-robot target search algorithms under blockchain technology

[0140]

[0141] The application of blockchain technology for target search in a multi-robot system increases the path planning time. This is because it takes a certain amount of time for the robot to obtain and verify data messages in the blockchain, which prolongs the path planning time. However, blockchain technology can ensure the reliability of the data source and the immutability of the data, which is crucial for the security and trust building of the multi-robot system. Although the target search path planning time will be extended, the path length and traversal times under blockchain technology have little effect on the multi-robot system.

Claims

1. A multi-robot system target search method based on blockchain, characterized in that: The steps include: S1. Build a shared network of multiple robots on the blockchain so that robots can obtain data through the blockchain and form a multi-robot system; S2. One of the robots establishes a grid map through SLAM-Gmapping to represent the location of the real scene and obstacles through the grid map, and the robot binarizes the grid map and uploads it to the blockchain; S3. When the position of the target point is known, the cost of the robot from the initial point to the target point is estimated by the improved A* target algorithm to obtain the optimal path of the robot; S4. When the position of the target point is unknown, the robot predicts the target point through the target probability function of the grid map, so that the multi-robot system can draw a target probability map on the grid map through the different perception strengths of the target signal by the robot; S5. In the target probability map, the robot calculates the moving direction according to the search benefit function and determines whether the moving direction falls into a local optimal problem; S6. If the robot falls into the local optimal problem, the search grid area is expanded and the search benefit function is calculated again to carry out the next step of target search path planning. After the robot does not fall into the local optimal problem, the robot leaves the obstacle through the D*Lite target search path algorithm, and after leaving the obstacle, the robot calculates the moving direction according to the search benefit function until it searches near the target point to obtain the optimal path of the robot; S7. In the target probability map, if the robot cannot determine the location of the target point, it returns to step S5 to retry. When the number of retries is greater than a threshold, the robot search task is forcibly terminated.

2. According to claim 1, a multi-robot system target search method based on blockchain is characterized in that: In step S1, each node of the robot generates a pair of public and private key pairs in the private blockchain, the public key is uploaded to the blockchain and broadcast to other blockchain nodes through the blockchain, and the robot stores the private key in its own safe location; the robot hashes the plaintext data through the hash algorithm to obtain a hash value, and the robot encrypts the hash value through its own private key and sends it to the blockchain.

3. According to claim 2, a multi-robot system target search method based on blockchain is characterized in that: In step S2, the robot performs hash calculation on the binarized grid map to obtain a hash value, and the robot signs the binarized hash value of the grid with its own private key and uploads it to the blockchain.

4. According to a blockchain-based multi-robot system target search method according to claim 1, it is characterized by: In step S3, the improved A* target algorithm estimates the cost from the target point to the end point through an evaluation function, and the evaluation function is: f(n)=g(n)+(1-lgp)h(n)+(1+lgp)T(n) Formula (1) Among them, g(n) is the evaluation function; g(n) is the actual movement cost consumed by the robot to move from the starting point to the current position node n; P is the obstacle rate of the grid map; h(n) is the heuristic function; T(n) is the number of bends in the path from the starting point to the current node n.

5. According to claim 4, a multi-robot system target search method based on blockchain is characterized in that: The obstacle rate calculation method of the grid map is: Wherein, S is the number of obstacle grids in the grid map; (x b ,y b ) is the starting point of the robot; (x e ,y e ) is the end point of the robot.

6. The multi-robot system target search method based on blockchain according to claim 1, characterized in that: In step S4, when the position of the target point is unknown, the position of the target point is relative to any grid in the environment grid map where the robot may appear, and the target probability function is: Among them, f(x, y) is the probability that the target point is located at the grid (x, y); φ(x, y) is the target point position estimated by multiple robots; Ω is the entire task environment.

7. A multi-robot system target search method based on blockchain according to claim 6, characterized in that: The method for multi-robots to estimate the target point position is: Among them, (x i ,y i ) is the position of the multi-robot in the grid map; (x, y) is the estimated point of the target point position in the grid map; μ is a constant; k is the number of multi-robots.

8. The multi-robot system target search method based on blockchain according to claim 1, characterized in that: In step S5, at a certain time t, the robot is located at position N in the grid map. After time t+1, the robot moves to position N+1. The search benefit function is expressed as φ(N+1, t+1). The calculation method of the search benefit function is: φ(N+1,t+1)=0.6I A (N+1,t+1)-0.1I B (N+1,t+1)+0.2I C (N+1,t+1) Formula (5) Among them, I A (N+1, t+1) is the target gain obtained by the robot when it goes to grid N+1 at time t+1; I B (N+1,t+1) is the aggregation cost of the robot; I C (N+1, t+1) is the exploration benefit of the robot; when an obstacle appears between the target point and the robot, and the target point is close to the robot, the robot falls into a local optimal problem.

9. The multi-robot system target search method based on blockchain according to claim 8, characterized in that: In step S6, a function Rhs is introduced into the D*Lite target search path algorithm to represent the estimated value from the target point to the point where the robot falls into the local optimal problem: Among them, k(s) is the basis for the order of node detection, indicating the priority amount that the robot processes first, and k(s) includes two values ​​[k1(s); k2(s)]; g(s) is the actual value that the robot needs to consume from the local optimal problem point to the target point; h(s,s start ) is the estimated value from the point where the robot falls into the local optimal problem to the starting point; k m Used to rationalize node sorting calculations; when k1(s) = k2(s), the robot target search path planning is completed.

10. A multi-robot system target search method based on blockchain according to claim 9, characterized in that: The estimated value from the target point to the point where the robot falls into the local optimal problem is expressed as: Where S is the point where the robot falls into the local optimal problem; S goal is the target point; c(s, s′) is the value consumed by the robot moving from the local optimal problem point to the adjacent node; According to formula (7), the estimated value from the point where the robot falls into the local optimal problem to the starting point is calculated as: Among them, h(s′,s goal ) is the value consumed in moving from the adjacent node of the local optimal problem point where the robot is trapped to the target point.

Images