Fire-fighting robot enhanced path planning method and system based on SOM network

Abstract

The invention discloses a fire-fighting robot enhanced path planning method and system based on an SOM network. The method comprises the following steps: constructing an initialized SOM network; according to a competition strategy mechanism, selecting a superior neuron for the initialized SOM network; according to a weight adjustment mechanism considering a direction vector, adjusting weights of superior neurons of the initialized SOM network and neurons in the field of the superior neurons; the execution is repeated until the neuron weight does not change any more, and an initial path of each fire-fighting robot is obtained; performing first post-processing operation on the initial path of each fire-fighting robot to obtain a first optimized path; and performing second post-processing operation on the first optimized path to obtain a final path. And the method is of great significance in realizing efficient obstacle avoidance and task collaborative allocation in a complex environment.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular to a firefighting robot enhanced path planning method and system based on a SOM network. Background Art

[0002] With the rapid development of science and technology, firefighting robots have shown great application potential in many fields such as high-rise buildings, petrochemical industrial areas, underground spaces, warehouses, logistics centers, etc. However, in complex tasks, a single robot is often limited by its processing power, sensing range and response speed when performing tasks, and it is difficult to meet the requirements of high real-time and high-precision safety rescue.

[0003] Therefore, multi-machine collaboration technology has attracted widespread attention because it can effectively cope with the challenges of complex tasks. As one of the core technologies of multi-machine collaboration, task allocation and path planning are directly related to the overall performance and task execution efficiency of multiple firefighting robots. Specifically, task allocation determines which robots are responsible for which tasks, while path planning is responsible for planning the optimal or suboptimal path from the starting point to the end point for the robot, ensuring that the robot can complete the task safely and efficiently.

[0004] In view of this, the collaborative optimization problem of task allocation and path planning of fire-fighting robots has become an emerging research focus and is particularly important in complex fire-fighting and rescue mission execution scenarios.

[0005] Currently, mainstream solutions for task allocation and path planning focus on two strategies:

[0006] First, the task allocation result is used as a prerequisite for path planning and solved independently. Although this solution has clear logic, the coupling between algorithms is low and the overall solution architecture tends to be complex.

[0007] The second is to try to couple task allocation with path planning to achieve more efficient collaboration. However, this strategy also faces the challenges of complex collaborative algorithm design and increased difficulty in overall system optimization.

[0008] What is particularly critical is that the effectiveness and adaptability of traditional task allocation algorithms are significantly limited when dealing with small-scale environments with dense obstacles, making it difficult to meet the needs of high-precision and high-flexibility task execution.

[0009] Therefore, exploring a practical algorithm that can efficiently coordinate task allocation and path planning and effectively avoid obstacles in complex indoor fire environments has become one of the technical problems that need to be solved urgently.

[0010] As far as existing research is concerned, the SOM neural network is an unsupervised neural network that can automatically cluster data into a group based on the similarity of the input data. In task allocation, this means that the SOM network can automatically group tasks according to certain similarities (distance, difficulty, resource requirements, etc.), and can automatically generate a path from the starting point to the target point based on the position of neurons at different times, thus completing path planning while allocating tasks.

[0011] Given the inherent greedy selection characteristics of neurons in SOM neural networks, current SOM-based algorithms often have difficulty in ensuring the generation of an optimal path that completely avoids obstacles in path planning. This limitation restricts the application of traditional SOM algorithms in complex and changing environments.

[0012] A common solution to the existing problem is to build a virtual repulsive field around the obstacle to guide the robot to autonomously avoid obstacles during the task and ensure safe operation. However, this method faces significant limitations in complex and changing environments. The core problem is that the number of neurons set by the SOM neural network is fixed and limited, which makes it difficult to fully adapt to the dual requirements of high-precision obstacle avoidance and efficient task allocation in complex environments. In complex scenarios, the limited number of neurons not only limits the flexibility and accuracy of the obstacle avoidance strategy, but also may cause conflicts between obstacle avoidance requirements and task allocation goals, becoming a key factor restricting the overall performance of the system. Summary of the invention

[0013] In order to address the deficiencies of the prior art, the present invention provides an enhanced path planning method and system for a firefighting robot based on a SOM network, and explores an improved SOM algorithm that can dynamically adjust the configuration of neurons and enhance environmental adaptability, which is of great significance for achieving efficient obstacle avoidance and collaborative task allocation in complex environments. When the traditional SOM is applied to task allocation and path planning, the fixed number of neurons is set, which is the main reason for the algorithm to cope with the dual challenges of robot obstacle avoidance and task allocation. Therefore, in order to solve the contradiction between task allocation and obstacle avoidance in the SOM neural network algorithm, the key is to properly select the neuron growth strategy and dynamically adapt to changes in the indoor fire environment.

[0014] On the one hand, an enhanced path planning method for firefighting robots based on SOM network is provided;

[0015] The enhanced path planning method of fire fighting robot based on SOM network includes:

[0016] (1) Constructing an initialized SOM network; selecting the winning neuron from the initialized SOM network according to the competition strategy mechanism;

[0017] (2) According to the weight adjustment mechanism considering the direction vector, the weights of the winning neurons of the initialized SOM network and the neurons in its domain are adjusted;

[0018] (3) Repeat (1) to (2) until the neuron weights no longer change, and the initial path of each firefighting robot is obtained;

[0019] (4) Performing a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path;

[0020] (5) Performing a second post-processing operation on the first optimized path to obtain a final path.

[0021] On the other hand, an enhanced path planning system for firefighting robots based on SOM networks is provided;

[0022] The enhanced path planning system for firefighting robots based on SOM network includes:

[0023] A construction module is configured to: construct an initialized SOM network; select a winning neuron for the initialized SOM network according to a competition strategy mechanism;

[0024] A weight adjustment module, which is configured to: adjust the weights of the winning neurons of the initialized SOM network and the neurons in its domain according to the weight adjustment mechanism considering the direction vector;

[0025] A repetition module is configured to: repeatedly execute the construction module and the weight adjustment module until the neuron weight no longer changes, thereby obtaining an initial path for each firefighting robot;

[0026] A primary processing module is configured to: perform a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path;

[0027] The secondary processing module is configured to perform a second post-processing operation on the first optimized path to obtain a final path.

[0028] On the other hand, there is also provided an electronic device, comprising:

[0029] a memory for non-transitory storage of computer-readable instructions; and

[0030] a processor for executing the computer readable instructions,

[0031] When the computer-readable instructions are executed by the processor, the method described in the first aspect is executed.

[0032] On the other hand, a storage medium is provided, which non-temporarily stores computer-readable instructions, wherein when the non-temporary computer-readable instructions are executed by a computer, the instructions of the method described in the first aspect are executed.

[0033] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program is used to implement the method described in the first aspect when running on one or more processors.

[0034] The above technical solution has the following advantages or beneficial effects:

[0035] This invention innovatively proposes a dual neuron growth strategy to optimize the obstacle avoidance and task allocation of multiple firefighting robots and multiple rescue robots in complex fire scenes, and introduces a novel simplified broken line obstacle avoidance algorithm to ensure that the path is collision-free. Ultimately, the algorithm can achieve efficient task allocation and path planning with a small turning radius for multiple firefighting and rescue robots in complex environments, ensuring that all robots can complete obstacle avoidance with the minimum path, thereby improving the efficiency of firefighting and rescue.

[0036] The present invention generates child neurons by copying parent neurons and inhibiting the activity of parent neurons, effectively solving the contradiction between obstacle avoidance and task allocation when using the original SOM for task allocation and path planning. The overall idea of ​​the solution is clear and the implementation is simple and convenient.

[0037] The present invention implements a strategy of completely avoiding obstacles by adding new neurons between two adjacent neurons, thereby ensuring that the path generated after path planning is completed will not cross any obstacles, thereby ensuring the reliability and safety of the path planning process.

[0038] The present invention proposes a novel post-processing algorithm, which can simplify broken lines and avoid obstacles, reduce path burrs, and reduce curvature. After being applied to SOM task allocation and path planning, a smooth path is generated to ensure that all robots can complete obstacle avoidance with the minimum path, thereby improving the efficiency of fire fighting and rescue. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0040] Figure 1 It is a schematic diagram of the S102 method;

[0041] Figure 2 It is the flowchart of the S102 method;

[0042] Figure 3 It is a schematic diagram of the S104 method;

[0043] Figure 4 It is the flowchart of the S104 method;

[0044] Figure 5 It is a schematic diagram of the method of S105 case 1;

[0045] Figure 6 It is a schematic diagram of the method of S105 case 2;

[0046] Figure 7 This is the final path simulation schematic diagram;

[0047] Among them, 1 is the first position point, 2 is the second position point, and 3 is the third position point. DETAILED DESCRIPTION

[0048] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0049] Embodiment 1

[0050] This embodiment provides an enhanced path planning method for a firefighting robot based on a SOM network;

[0051] The enhanced path planning method of fire fighting robot based on SOM network includes:

[0052] S101: construct an initialized SOM network; select the winning neuron for the initialized SOM network according to the competition strategy mechanism;

[0053] S102: According to the weight adjustment mechanism considering the direction vector, the weights of the winning neuron of the initialized SOM network and the neurons in its domain are adjusted;

[0054] S103: Repeat S101 to S102 until the neuron weights no longer change, and obtain the initial path of each firefighting robot;

[0055] S104: performing a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path;

[0056] S105: performing a second post-processing operation on the first optimized path to obtain a final path.

[0057] Furthermore, the S101: constructing an initialization SOM network includes:

[0058] According to the number k of firefighting robots and the number m of target points, an initialized SOM network is constructed, and the number of neurons is set to k groups, each group of neurons includes m neurons; the weight of each group of neurons is set to the initial position information of the corresponding robot.

[0059] Among them, the neuron weight is expressed as R ij represents, i=1,...,k, j=1,...,m, R ij Represents the weight of the jth neuron in the i-th group of neurons. ij It also represents the coordinate position of the jth neuron in the i-th group of neurons.

[0060] It should be understood that the SOM network is divided into an input layer and an output layer. The number of neurons in the input layer is determined by the dimension of the input vector, and one neuron corresponds to one feature; a node in the output layer represents a clustered class.

[0061] Furthermore, the S101: according to the competition strategy mechanism, the winning neuron is selected for the initialized SOM network, and the specific competition strategy adopts a greedy strategy.

[0062] The greedy strategy is to take the best or optimal choice under the current state in each step of selection, hoping that the result is the best or optimal algorithm strategy in the world.

[0063] The greedy strategy selects the closest neuron as the winning neuron based on the distance between the neuron and the target point, thereby achieving efficient self-organization of the network.

[0064] Furthermore, if Figure 1 and Figure 2 As shown, the step S102: adjusting the weights of the winning neurons of the initialized SOM network and the neurons in their domain according to the weight adjustment mechanism considering the direction vector, includes:

[0065] Step (1-1): Assume that the neuron R 11 It has become the winning neuron through the greedy strategy;

[0066] Step (1-2): Through the traditional neuron weight adjustment strategy, neuron R 11 The weight is updated to the first position point 1, and the updated neuron is checked to see if it is within the obstacle;

[0067] Step (1-3): When neuron R is detected 11When located inside an obstacle, the firefighting robot is taken as the origin, and the line connecting the firefighting robot and the first position point 1 is rotated clockwise and counterclockwise respectively. The line is rotated twice in the clockwise direction to obtain the second position point 2, and the line is rotated twice in the counterclockwise direction to obtain the third position point 3; the line is used as the direction vector;

[0068] The rotation angle formula is as follows:

[0069] θ=tβ

[0070] Among them, θ represents the angle that the vector needs to rotate, t represents the number of times the vector needs to rotate, and t=t is taken as clockwise. 1 (t 1 <0), counterclockwise is t=t 2 (t 2 >0), β represents the angle of each rotation, which can be 15°, 30° and 45°. In the present invention, β=15°;

[0071] Furthermore, when the second position point 2 is obtained, t 1 = -2, when the third position point 3 is obtained, t 2 =2;

[0072] Step (1-4): Assume that the second position point 2 is outside the obstacle, and the third position point 3 is still inside the obstacle. Therefore, in order to make the third position point 3 outside the obstacle, it is necessary to make t 2 ≥3, at this time |t 1 |<|t 2 |, therefore, the second position point 2 obtained by rotating clockwise is selected as the final weight point. At this time, neuron R 11 Update to the second position point 2;

[0073] Step (1-5): The second position point 2 is obtained by rotating the direction vector. Therefore, the parent neuron R is copied at the second position point 2. 11 Get daughter neurons And set the parent neuron R 11 no longer participate in the competitive process;

[0074] Steps (1-6): Daughter Neurons Participate in the competition and repeat steps (1-1), (1-2), (1-3), (1-4) and (1-5) until the weights of the group of neurons no longer change, and an initial path from the starting point to the end point is obtained.

[0075] Furthermore, the traditional neuron weight adjustment strategy is formulated as follows:

[0076]

[0077] Among them, β is the learning rate, η is a small positive constant, and D min is the minimum distance between any two neurons, R km (t+1) is the updated weight of the neuron, R km (t) is the weight of the neuron before updating, T i (t) is the i-th target point, D ikm is the distance between the i-th target point and R km (t) weighted distance, domain function f(d j ,G) is defined as follows:

[0078]

[0079] Among them, d j is the distance between the jth neuron and the winning neuron, j = 1, 2, ..., m, G is the neighbor function gain, and r is the neighbor range radius.

[0080] Furthermore, if Figure 3 and Figure 4 As shown, the S104: performing a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path includes:

[0081] According to the arrangement order of neurons of each robot's initial path, starting from the starting neuron, check one by one to the last neuron, and check whether there are obstacles between each pair of adjacent neurons in turn;

[0082] If there is an obstacle between the paths formed by adjacent neurons A and B, a new neuron R is added to the corresponding neuron group. add , which is located between neurons A and B, and the new neuron R add The weight of is initialized to the arithmetic mean of the weights of neurons A and B;

[0083] Check the new neurons R at this time add Is it in an obstacle? If not, keep the neuron R add , continue to check from the starting neuron to the last neuron one by one;

[0084] If the new neuron R add If the path is in an obstacle, then the perpendicular bisector of path AB is drawn. At this time, the newly added neuron R add Located on the vertical foot, with AB as the dividing line and neuron R addThe initial position is the starting point, and the movement is carried out along the first direction and the second direction away from the foot of the perpendicular on the vertical line. The distance of each movement is dist. After each movement of dist, it is determined whether the position after the movement is within the obstacle. If the position after the movement is not within the obstacle, it means that the position after the movement is neuron R. add The candidate position points are obtained from the first direction, where the candidate position point C is obtained from the first direction, and the candidate position point D is obtained from the second direction. add The number of moves required from the initial position to the two candidate positions C and D respectively, select the candidate position C with the fewest moves as the final position of the newly added neuron, and take the path ACB passing through points A, C and B in sequence as the first optimization path.

[0085] It should be understood that this adjustment process is intended to guide new neurons to "go around" obstacles, thereby ensuring the continuity and unobstructedness of the path, and specifically includes the following steps:

[0086] Step (2-1): Check whether a certain path AB passes through an obstacle;

[0087] Step (2-2): After determining that path AB passes through an obstacle, add a neuron R at the midpoint of the path. add , and check R add Is it inside an obstacle? Confirm R add If it falls into the obstacle, the subsequent operation will be performed;

[0088] Step (2-3): Move it to both sides along the direction perpendicular to the path AB by a fixed distance dist, and record the number of times it moves away from the obstacle s 1 and 2 , until reaching points C and D;

[0089] Step (2-4): Compare the number of moves to leave the obstacle s 1 and 2 , select the C point with fewer times as the newly added neuron after updating;

[0090] Step (2-5): Connect ACBs to form a new path.

[0091] Furthermore, if Figure 5 As shown, the S105: performing a second post-processing operation on the first optimized path to obtain a final path, case 1 includes:

[0092] Step (3-1): First, mark the starting point of the path and several set position points. Assume that the path passes through points C, D, E, and F in sequence. Assume that points C and D are marked points, and points E and F are unmarked points. Marked points are manually set path points that must be passed, and unmarked points are candidate path points in the path.

[0093] Step (3-2): Starting from point C, passing through the middle path point D, connecting CE, and determining whether path point D is a marked point. If path point D is a marked point, it means that the path must pass through point D, and point D cannot be removed. Therefore, CE is not connected, and CD and DE are retained as paths.

[0094] Step (3-3): Starting from point D, passing through the middle path point E, connecting DF. Since the middle path point E is not marked, it is determined whether the line segment DF passes through an obstacle. If the line segment DF passes through an obstacle, it means that DF cannot be used as a simplified path and point E cannot be eliminated.

[0095] Step (3-4): Get the final path CDEF.

[0096] Furthermore, if the line DF connecting the previous point D and the next point F of an unmarked point E in the path passes through an obstacle, then the unmarked point E cannot be eliminated; if the line DF connecting the previous point D and the next point F of an unmarked point E in the path does not pass through an obstacle, then the unmarked point E can be eliminated.

[0097] Furthermore, if Figure 6 As shown, the S105: performing a second post-processing operation on the first optimized path to obtain a final path, the second situation includes:

[0098] Step (4-1): First, mark the starting point of the path and the points that must be passed through to the target, indicating that these points are indispensable in the path. Point C is marked, and points D, E, and F are unmarked points;

[0099] Step (4-2): Starting from point C, passing through the middle path point D, connecting CE, determining whether CE passes through an obstacle and determining whether the middle path point D is a marked point. In this case, both are negative, indicating that the path does not need to pass through point D. Point D is removed, and CE is connected; path CDE is deleted, and path CE is retained;

[0100] Step (4-3): Still starting from point C, crossing the middle path point E, connecting CF. Since CF passes through an obstacle, point E cannot be eliminated, so CF is not connected, and CE and EF are retained as paths;

[0101] Step (4-4): Get the final path CEF.

[0102] It should be understood that the second post-processing operation adopts a discounted post-processing algorithm, which is intended to simplify the broken line path and avoid obstacles, reduce burrs, and reduce the curvature after kinematic planning, thereby generating a smooth path. The final path is as follows Figure 7 As shown in the figure, R represents the robot and T represents the target point.

[0103] By introducing a dual neuron growth strategy, it is integrated into the competitive learning mechanism and path post-processing stage of SOM. This strategy effectively solves the conflict between task allocation and obstacle avoidance in complex indoor fire environments, and significantly optimizes the task allocation and path planning capabilities of multiple firefighting robots during firefighting and rescue. In addition, the present invention innovatively introduces a simplified broken line obstacle avoidance algorithm, which intelligently selects two strategies based on different scenarios, which not only greatly reduces the redundant turns in the path, reduces the overall bending degree of the path, but also ensures that the path is collision-free. Ultimately, the algorithm can achieve efficient task allocation and low-bending path planning for multiple robots in complex environments, optimize the path length while ensuring that all robots can safely avoid obstacles, improve firefighting and rescue efficiency, and successfully complete firefighting and rescue tasks.

[0104] Embodiment 2

[0105] This embodiment provides an enhanced path planning system for fire fighting robots based on a SOM network;

[0106] The enhanced path planning system for firefighting robots based on SOM network includes:

[0107] A construction module is configured to: construct an initialized SOM network; select a winning neuron for the initialized SOM network according to a competition strategy mechanism;

[0108] A weight adjustment module, which is configured to: adjust the weights of the winning neurons of the initialized SOM network and the neurons in its domain according to the weight adjustment mechanism considering the direction vector;

[0109] A repetition module is configured to: repeatedly execute the construction module and the weight adjustment module until the neuron weight no longer changes, thereby obtaining an initial path for each firefighting robot;

[0110] A primary processing module is configured to: perform a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path;

[0111] The secondary processing module is configured to perform a second post-processing operation on the first optimized path to obtain a final path.

[0112] It should be noted that the above-mentioned construction module, weight adjustment module, repetition module, primary processing module and secondary processing module correspond to steps S101 to S105 in Embodiment 1, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Embodiment 1. It should be noted that the above-mentioned modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0113] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0114] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0115] Embodiment 3

[0116] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory so that the electronic device executes the method described in the above embodiment one.

[0117] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0118] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0119] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software.

[0120] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0121] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0122] Embodiment 4

[0123] This embodiment further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method described in the first embodiment is completed.

[0124] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The enhanced path planning method of fire fighting robot based on SOM network is characterized by: include: (1) Construct and initialize the SOM network; According to the competition strategy mechanism, select the winning neuron for the initialized SOM network; (2) According to the weight adjustment mechanism considering the direction vector, the weights of the winning neurons of the initialized SOM network and the neurons in its domain are adjusted; (3) Repeat (1) to (2) until the neuron weights no longer change, and the initial path of each firefighting robot is obtained; (4) Performing a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path; (5) Performing a second post-processing operation on the first optimized path to obtain a final path.

2. The enhanced path planning method for fire fighting robots based on SOM network as claimed in claim 1 is characterized in that: Construct and initialize the SOM network, including: According to the number k of firefighting robots and the number m of target points, an initialized SOM network is constructed, and the number of neurons is set to j groups, each group of neurons includes m neurons; the weight of each group of neurons is set to the initial position information of the corresponding robot.

3. The enhanced path planning method for fire fighting robots based on SOM network as claimed in claim 1 is characterized in that: According to the competition strategy mechanism, the winning neuron is selected for the initialized SOM network. The specific competition strategy adopts the greedy strategy, which selects the neuron with the closest distance as the winning neuron based on the distance between the neuron and the target point.

4. The enhanced path planning method for fire fighting robots based on SOM network as claimed in claim 1 is characterized in that: According to the weight adjustment mechanism considering the direction vector, the weights of the winning neurons of the initialized SOM network and the neurons in its field are adjusted, including: Step (1-1): Assume that the neuron R 11 It has become the winning neuron through the greedy strategy; Step (1-2): Through the traditional neuron weight adjustment strategy, neuron R 11 The weights are updated to the position of the first position point, and the updated neuron is checked to see if it is within the obstacle; Step (1-3): When neuron R is detected 11 When located in an obstacle, the fire-fighting robot is taken as the origin, and the line connecting the fire-fighting robot and the first position point is rotated clockwise and counterclockwise respectively. The line is rotated twice in the clockwise direction to obtain the second position point, and the line is rotated twice in the counterclockwise direction to obtain the third position point; the line is used as the direction vector; Step (1-4): Assume that the second position point is outside the obstacle, while the third position point is still inside the obstacle. Therefore, in order to make the third position point outside the obstacle, t2≥3 is set. At this time, |t1|<|t2|. Therefore, the second position point obtained by rotating clockwise is selected as the final weight point. At this time, neuron R 11 Update to a second position point; Step (1-5): The second position point is obtained by rotating the direction vector, so the parent neuron R is copied at the second position point. 11 Get daughter neurons And set the parent neuron R 11 no longer participate in the competitive process; Steps (1-6): Daughter Neurons Participate in the competition and repeat steps (1-1), (1-2), (1-3), (1-4) and (1-5) until the weights of the group of neurons no longer change, and an initial path from the starting point to the end point is obtained.

5. The enhanced path planning method for fire fighting robots based on SOM network as claimed in claim 1 is characterized in that: For the initial path of each firefighting robot, the first post-processing operation is performed to obtain the first optimized path, including: According to the arrangement order of neurons of each robot's initial path, starting from the starting neuron, check one by one to the last neuron, and check whether there are obstacles between each pair of adjacent neurons in turn; If there is an obstacle between the paths formed by adjacent neurons A and B, a new neuron R is added to the corresponding neuron group. add , which is located between neurons A and B, and the new neuron R add The weight of is initialized to the arithmetic mean of the weights of neurons A and B; Check the new neurons R at this time add Is it in an obstacle? If not, keep the neuron R add , continue to check from the starting neuron to the last neuron one by one; If the new neuron R add If the path is in an obstacle, then the perpendicular bisector of path AB is drawn. At this time, the newly added neuron R add Located on the vertical foot, with AB as the dividing line and neuron R add The initial position is the starting point, and the movement is carried out along the first direction and the second direction away from the foot of the perpendicular on the vertical line. The distance of each movement is dist. After each movement of dist, it is determined whether the position after the movement is within the obstacle. If the position after the movement is not within the obstacle, it means that the position after the movement is neuron R. add The candidate position points are obtained from the first direction, where the candidate position point C is obtained from the first direction, and the candidate position point D is obtained from the second direction. add The number of moves required from the initial position to the two candidate positions C and D respectively, select the candidate position C with the fewest moves as the final position of the newly added neuron, and take the path ACB passing through points A, C and B in sequence as the first optimization path.

6. The enhanced path planning method for fire fighting robots based on SOM network as claimed in claim 1, characterized in that: A second post-processing operation is performed on the first optimized path to obtain a final path. Case 1 includes: Step (3-1): First, mark the starting point of the path and several set position points. Assume that the path passes through points C, D, E, and F in sequence. Assume that points C and D are marked points, and points E and F are unmarked points. Marked points are manually set path points that must be passed, and unmarked points are candidate path points in the path. Step (3-2): Starting from point C, passing through the middle path point D, connecting CE, and determining whether path point D is a marked point. If path point D is a marked point, it means that the path must pass through point D, and point D cannot be removed. Therefore, CE is not connected, and CD and DE are retained as paths. Step (3-3): Starting from point D, passing through the middle path point E, connecting DF. Since the middle path point E is not marked, it is determined whether the line segment DF passes through an obstacle. If the line segment DF passes through an obstacle, it means that DF cannot be used as a simplified path and point E cannot be eliminated. Step (3-4): Get the final path CDEF.

7. The enhanced path planning method for fire fighting robots based on SOM network as claimed in claim 1 is characterized in that: The second post-processing operation is performed on the first optimized path to obtain the final path. Case 2 includes: Step (4-1): First, mark the starting point of the path and the points that must be passed through to the target, indicating that these points are indispensable in the path. Point C is marked, and points D, E, and F are unmarked points; Step (4-2): Starting from point C, passing through the middle path point D, connecting CE, determining whether CE passes through an obstacle and determining whether the middle path point D is a marked point. In this case, both are negative, indicating that the path does not need to pass through point D. Point D is removed, and CE is connected; path CDE is deleted, and path CE is retained; Step (4-3): Still starting from point C, crossing the middle path point E, connecting CF. Since CF passes through an obstacle, point E cannot be eliminated, so CF is not connected, and CE and EF are retained as paths; Step (4-4): Get the final path CEF.

8. The enhanced path planning system of fire fighting robot based on SOM network is characterized by: include: A construction module is configured to: construct an initialized SOM network; select a winning neuron for the initialized SOM network according to a competition strategy mechanism; A weight adjustment module, which is configured to: adjust the weights of the winning neurons of the initialized SOM network and the neurons in its domain according to the weight adjustment mechanism considering the direction vector; A repetition module is configured to: repeatedly execute the construction module and the weight adjustment module until the neuron weight no longer changes, thereby obtaining an initial path for each firefighting robot; A primary processing module is configured to: perform a first post-processing operation on the initial path of each firefighting robot to obtain a first optimized path; The secondary processing module is configured to perform a second post-processing operation on the first optimized path to obtain a final path.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; as well as a processor for executing the computer readable instructions, When the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitory stored, wherein when the non-transitory computer-readable instructions are executed by a computer, the instructions of the method according to any one of claims 1 to 7 are executed.

Images