Autonomous task allocation system for multiple heterogeneous robots for underground space exploration

Abstract

The multi-heterogeneous robot autonomous task allocation system for underground space detection disclosed by the present invention belongs to the field of robot technology, and includes a segmentation area and heterogeneous robot cluster relationship establishment unit, a problem modeling unit, a fault online monitoring unit, an interference online observation and prediction unit, a transfer learning unit, and a task and heterogeneous robot mapping unit. By combining the segmentation area and heterogeneous robot cluster relationship establishment unit, the problem modeling unit, the fault online monitoring unit, the interference online observation and prediction unit, the transfer learning unit, and the task and heterogeneous robot mapping unit, the underground space detection task is completed, and the heterogeneous robots are endowed with strong autonomous, strong adaptability, and strong survival intelligent behavior capabilities; thereby effectively improving the efficiency and safety of urban underground pipe gallery detection task allocation, and having automation, digitization, and intelligent characteristics.

Description

Technical Field

[0001] The present invention belongs to the field of robot technology, and in particular relates to a multi-heterogeneous robot autonomous task allocation system for underground space detection. Background Art

[0002] In recent years, with the increasing demand for underground unstructured irregular space detection, the difficulty of detection has become increasingly high. How to achieve global task allocation for detection robots has become the key to underground space detection. Underground unstructured irregular space detection tasks, such as urban underground pipe gallery detection tasks, have the characteristics of time and space constraints, multi-source heterogeneous interference, and high dynamic multi-target, which leads to problems such as limited efficiency of task allocation of traditional detection robots. Faced with such a complex scenario, the single intelligent system or homogeneous robot in the existing technology is difficult to complete the detection task of underground space because it cannot fully utilize the information of the body, environment, geography, etc., while multi-heterogeneous robots can overcome the above problems and complete the detection task.

[0003] It can be seen that the global task allocation of multiple heterogeneous robots has become the basis for successful detection; however, in the face of multi-target task allocation in satellite denial and unstructured underground space environments, previous task allocation methods, such as linear programming-related methods, heuristic methods, market mechanism-based methods, and learning-based methods, are difficult to apply. Therefore, it is particularly important to develop an autonomous task allocation method for multiple heterogeneous robots for underground space detection. Summary of the invention

[0004] The present invention aims at the above-mentioned problems, makes up for the deficiencies of the prior art, and provides a multi-heterogeneous robot autonomous task allocation system for underground space exploration; the present invention effectively improves the autonomy, safety and reliability of underground space exploration.

[0005] To achieve the above purpose, the present invention adopts the following technical solution.

[0006] The multi-heterogeneous robot autonomous task allocation system for underground space exploration provided by the present invention includes a segmentation area and heterogeneous robot cluster relationship establishment unit, a problem modeling unit, a fault online monitoring unit, an interference online observation and prediction unit, a transfer learning unit, and a task and heterogeneous robot mapping unit;

[0007] The segmented area and heterogeneous robot cluster relationship establishment unit establishes the segmented area and heterogeneous robot cluster relationship through a finite-time Bernstein network, divides the exploration area of ​​the underground space into different sub-areas according to the workload, and divides the heterogeneous robots into multiple heterogeneous robot colonies according to the execution capabilities of the heterogeneous robots themselves, so as to form a functionally complementary heterogeneous robot cluster;

[0008] On the basis of the relationship between the segmented area and the heterogeneous robot cluster, the target optimization function and constraint conditions are characterized by the problem modeling unit; the operating status of each heterogeneous robot is monitored by the fault online monitoring unit. Once a heterogeneous robot fails, the heterogeneous robot exits the detection task, which serves as a constraint condition of the problem modeling unit; the interference factors in the task allocation are characterized and predicted by the interference online observation and prediction unit. Once external interference is observed, it is also used as a constraint condition of the problem modeling unit;

[0009] Based on the problem modeling unit, the fault online monitoring unit, and the interference online observation and prediction unit, the transfer learning unit performs reasoning on past similar cases and rule reasoning, and uses past experience to speed up task allocation;

[0010] On the basis of the transfer learning unit, the task and heterogeneous robot mapping unit constructs the relationship between each task in the segmented area and the heterogeneous robots in each group through a market auction mechanism to complete task allocation.

[0011] As a preferred solution of the present invention, the finite-time Bernstein network is composed of an input layer, an intermediate layer and an output layer. The expression of the finite-time Bernstein network is as follows:

[0012] ;

[0013] in, To expand to The Bernstein network; is the number of expansions; is a finite time term; ; is a symbolic function; is the weight coefficient; ; is the number of middle-layer nodes; is a real-valued function defined on [0,1]; Input for the network; is the number of input items.

[0014] As another preferred embodiment of the present invention, the previous similar case reasoning includes calculating the similarity between the source case and the target case, and retrieving the historical case with the highest similarity to the target case from the case library. The calculation formula is as follows:

[0015] ;

[0016] in, , represent the source case and the target case respectively; is the normalized Euclidean distance between the source case and the target case; k To retrieve the attribute number; n To retrieve the total number of attributes; Number k Attribute weights; Number the source case and the target case k The distance after normalization on the attribute dimension.

[0017] As another preferred embodiment of the present invention, the rule reasoning includes establishing a rule base and a decision reasoning mechanism; the rule base is derived from the allocation rules formed by previous task allocation cases, and the decision reasoning mechanism completes the mechanism construction through reasoning technology, expert system, Agent technology, knowledge acquisition and knowledge representation.

[0018] As another preferred embodiment of the present invention, the task and heterogeneous robot mapping unit further includes the following situations according to the relationship between the number of heterogeneous robots in the group and the number of tasks in the group: (1) when the number of tasks in the group is equal to the number of heterogeneous robots, a market auction mechanism is used to allocate tasks; (2) when the number of tasks in the group is greater than the number of heterogeneous robots, the efficiency of task allocation needs to consider the constraints of timing factors, and the tasks are released according to the timing, and then a market auction mechanism is used to allocate tasks.

[0019] As another preferred embodiment of the present invention, the market auction mechanism includes allocating tasks to heterogeneous robots through bidding. In the auction algorithm, each heterogeneous robot submits a bid, indicating the price that this heterogeneous robot is willing to complete the task; finally, the task is allocated to the heterogeneous robot with the highest bid.

[0020] As another preferred solution of the present invention, the finite-time Bernstein network optimizes its parameters using a learning algorithm, and the learning algorithm adopts any one of the Levenberg-Marquardt algorithm, the recursive least squares algorithm, the back propagation algorithm, and the extended Kalman filter algorithm.

[0021] As another preferred solution of the present invention, the original signal is analyzed by the signal processing technology to remove the noise signal; the signal processing technology adopts Fourier transform or wavelet analysis.

[0022] Beneficial effects of the present invention:

[0023] The present invention provides a method for allocating autonomous tasks of multiple heterogeneous robots for underground space detection. By combining the segmented area and heterogeneous robot cluster relationship establishment unit, the problem modeling unit, the fault online monitoring unit, the interference online observation and prediction unit, the transfer learning unit, and the task and heterogeneous robot mapping unit, the underground space detection task is completed, and the heterogeneous robots are endowed with intelligent behavioral capabilities with strong autonomy, strong adaptability, and strong survival. This can effectively improve the efficiency and safety of urban underground pipe gallery detection task allocation, and has automation, digitization, and intelligent characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a schematic block diagram of the structure of the multi-heterogeneous robot autonomous task allocation system for underground space exploration according to the present invention.

[0025] Markings in the figure: 1 is the segmentation area and heterogeneous robot cluster relationship establishment unit, 2 is the problem modeling unit, 3 is the fault online monitoring unit, 4 is the interference online observation and prediction unit, 5 is the transfer learning unit, and 6 is the task and heterogeneous robot mapping unit. DETAILED DESCRIPTION

[0026] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be understood that the specific implementation methods described herein are only used to explain the present invention and are not used to limit the present invention.

[0027] Combination Figure 1As shown, the multi-heterogeneous robot autonomous task allocation system for underground space exploration provided by an embodiment of the present invention includes a segmentation area and heterogeneous robot cluster relationship establishment unit 1, a problem modeling unit 2, a fault online monitoring unit 3, an interference online observation and prediction unit 4, a transfer learning unit 5, and a task and heterogeneous robot mapping unit 6; the segmentation area and heterogeneous robot cluster relationship establishment unit 1, the problem modeling unit 2, the fault online monitoring unit 3, the interference online observation and prediction unit 4, the transfer learning unit 5, and the task and heterogeneous robot mapping unit 6 are coordinated and connected; the segmentation area and heterogeneous robot cluster relationship establishment unit 1 establishes the segmentation area and heterogeneous robot cluster relationship through a finite-time Bernstein network, divides the exploration area of ​​the underground space into different sub-areas according to the workload, and divides the heterogeneous robots into multiple heterogeneous robot communities according to the execution capabilities of the heterogeneous robots themselves to form a functionally complementary heterogeneous robot cluster; in the segmentation area On the basis of the relationship between the region and the heterogeneous robot cluster, the target optimization function and constraint conditions are characterized by the problem modeling unit 2; the operating status of each heterogeneous robot is monitored by the fault online monitoring unit 3. Once a heterogeneous robot fails, the heterogeneous robot exits the detection task, which serves as a constraint condition of the problem modeling unit 2; the interference factors in task allocation are characterized and predicted by the interference online observation and prediction unit 4. Once external interference is observed, it is also used as a constraint condition of the problem modeling unit 2; on the basis of the problem modeling unit 2, the fault online monitoring unit 3, and the interference online observation and prediction unit 4, the transfer learning unit 5 performs previous similar case reasoning and rule reasoning, and uses past experience to speed up task allocation; on the basis of the transfer learning unit 5, the task and heterogeneous robot mapping unit 6 constructs the relationship between each task in the segmented region and the heterogeneous robots in each group through a market auction mechanism to complete task allocation.

[0028] The segmented area and heterogeneous robot cluster relationship establishment unit 1: First, according to the workload of the exploration area of ​​the underground space, it is divided into different sub-areas, and the obtained segmented area serial numbers are recorded as ; Then, according to the execution capabilities of the heterogeneous robots themselves, the heterogeneous robots are divided into multiple heterogeneous robot colonies to form functionally complementary heterogeneous robot clusters. The number of heterogeneous robots in each cluster is recorded as ; Finally, the relationship between the segmented area and the heterogeneous robot cluster is established through the finite time Bernstein network, and the corresponding relationship between the segmented area and the heterogeneous robot cluster is determined according to the weight value obtained by the finite time Bernstein network; the established functional relationship is: , , ,in, represents a heterogeneous robot swarm, represents the number of heterogeneous robot groups, Indicates the number of segmented regions. is the connection weight, Indicates the segmentation area; It can be described by a finite-time Bernstein network.

[0029] The problem modeling unit 2 specifically includes: utilizing the functional complementarity of heterogeneous robots and considering the space-time constraints to achieve task collaboration and information sharing; considering 12 heterogeneous robots of 5 categories, including 2 multi-rotor drones, denoted as A1 and A2; 3 tilt-rotor drones, denoted as A3, A4, and A5; 3 tracked robots, denoted as A6, A7, and A8; 2 legged robots, denoted as A9, A10, A111, A12, A13, A14, A15, A16, A17, A18, A19, A20, A21, A22, A23, A24, A25, A26, A27, A28, A29, A30, A31, A32, A33, A34, A35, A36, A37, A38, A39, A40, A41, A42, A43, A44, A45, A46, A47, A48, A49, A49, A50, A51, A52, A53, A54, A55, A56, A57, A58, A59, A60, A61, A62, A63, A74, A75, A76, A77, A78, A79, A80, A81, A82, A83, A84, A85, A86, A87, A90, A91, A92, A93, A94, A95, A96, A97, A98, A99, A100, A111, A112 10 ; 2 air-ground dual-purpose robots, denoted as A 11 , A 12 . Each of the 12 heterogeneous robots of 5 categories is equipped with a high-definition camera, an infrared imager, a temperature sensor, and a harmful gas sensor to achieve monitoring and early warning of underground space detection tasks. In addition, the number of tasks to be performed is set to M In addition, set the number of tasks to be executed to M, denoted as T1, T2, ..., T M ; Issue start commands and mission information through the command and control terminal to complete the task allocation of underground space detection.

[0030] Based on the relationship between the segmented areas and the heterogeneous robot clusters, the task completion and global task reward of each cluster are maximized, and the following multi-objective optimization function is established: ; Its constraints are: ; In the formula, the task allocation matrix is for: ;in, , ; s is the sequence number of the segmented area; Heterogeneous robots i Ability to perform tasks; Heterogeneous robots i The task sequence package stores heterogeneous robots i The tasks to be performed and their sequencing; To perform tasks j The reward value is and A non-negative function of the correlation; is the number of iterations for each group to converge; The number of heterogeneous robots assigned to each segmented area; Heterogeneous robots i The upper limit of the number of tasks to be executed.

[0031] Among them, the fault online monitoring unit 3 is used to establish a fault monitoring model through a finite time Bernstein network, combine signal processing technology and historical data, establish monitoring standards, and convert it into a binary classification problem, 0 represents no fault, 1 represents a fault, and monitor the abnormal state of each heterogeneous robot; wherein, the fault monitoring model is established as follows: using a finite time Bernstein network as a fault classifier, the input layer uses the robot operation monitoring data after signal processing, and the output layer is respectively fault-free data and faulty data, marking the fault-free data as "0" and the faulty data as "1", and training the network. When a heterogeneous robot is monitored and diagnosed to have a fault, the heterogeneous robot exits the detection task, and other fault-free heterogeneous robots take over.

[0032] Among them, the interference online observation and prediction unit 4 is used to establish an interference observer through a finite-time Bernstein network to quantify and characterize the interference impact, wherein the interference observer is established as follows: using a finite-time Bernstein network, the input layer uses observation data after signal processing, and the output layer is interference data to train the network; and is used to use signal processing technology and historical data, and at the same time combine the use of a finite-time Bernstein network to establish an interference prediction model to predict interference; complete early intervention in interference and reduce the impact of interference, wherein the interference prediction model is established as follows: using a finite-time Bernstein network, the input layer uses historical interference data after signal processing, and the output layer is interference prediction data to train the network.

[0033] The transfer learning unit 5 is used to speed up task allocation based on reasoning based on previous similar cases and rule reasoning, and to use past experience;

[0034] The task and heterogeneous robot mapping unit 6 is used to construct the relationship between each task in the segmented area and the heterogeneous robots in each group through a market auction mechanism to complete task allocation, taking into account interference and failure factors.

[0035] Specifically, the finite-time Bernstein network is composed of an input layer, an intermediate layer and an output layer. It has strong nonlinear approximation, fast convergence, strong generalization ability, self-learning ability, robustness and anti-interference performance. It can be well used as an approximation model and a classification model. The expression of the finite-time Bernstein network is as follows:

[0036] ;

[0037] in, To expand to The Bernstein network; is the number of expansions; is a finite time term; ; is a symbolic function; is the weight coefficient; ; is the number of middle-layer nodes; is a real-valued function defined on [0,1]; Input for the network; is the number of input items.

[0038] Specifically, the previous similar case reasoning includes calculating the similarity between the source case and the target case, and retrieving the historical case with the highest similarity to the target case from the case library. The calculation formula is as follows:

[0039] ;

[0040] in, , represent the source case and the target case respectively; is the normalized Euclidean distance between the source case and the target case; k To retrieve the attribute number; n To retrieve the total number of attributes; Number k Attribute weights; Number the source case and the target case k The distance after normalization on the attribute dimension. The rule reasoning mentioned above includes the establishment of a rule base and a decision reasoning mechanism; the rule base is derived from the allocation rules formed by previous task allocation cases, and the decision reasoning mechanism is completed through reasoning technology, expert system, Agent technology, knowledge acquisition and knowledge representation.

[0041] Specifically, the task and heterogeneous robot mapping unit 6 also includes the following situations according to the relationship between the number of heterogeneous robots in the group and the number of tasks in the group: (1) when the number of tasks in the group is equal to the number of heterogeneous robots, the market auction mechanism is used to allocate tasks; (2) when the number of tasks in the group is greater than the number of heterogeneous robots, the efficiency of task allocation needs to consider the constraints of timing factors, and the tasks are released according to the timing, and then the market auction mechanism is used to allocate tasks.

[0042] Specifically, the market auction mechanism includes allocating tasks to heterogeneous robots through bidding. In the auction algorithm, each heterogeneous robot submits a bid, which indicates the price at which the heterogeneous robot is willing to complete the task. Finally, the task is allocated to the heterogeneous robot with the highest bid. In this way, the auction algorithm can reduce computational complexity and communication overhead while ensuring the effect of task allocation.

[0043] Specifically, the finite-time Bernstein network optimizes its parameters using a learning algorithm, and the learning algorithm uses any one of the Levenberg-Marquardt algorithm, the recursive least squares algorithm, the back propagation algorithm, and the extended Kalman filter algorithm; it should be noted that the learning algorithm can also use other algorithms that can meet the optimization function, not limited to the above learning algorithms.

[0044] Specifically, the original signal is analyzed by the signal processing technology, the noise signal is removed, and the accuracy of fault monitoring is improved; the signal processing technology in the fault online monitoring unit 3 and the interference online observation and prediction unit 4 adopts Fourier transform or wavelet analysis. It should be noted that the signal processing technology can also adopt processing technology other than Fourier transform or wavelet analysis, as long as it can achieve the analysis and denoising of the original signal.

[0045] In summary, through the embodiments provided by the present invention, the following outstanding advantages of the present invention can be summarized:

[0046] (1) The present invention integrates task allocation technology, prediction technology, anti-interference technology, fault monitoring and diagnosis technology, intelligent optimization technology and artificial intelligence technology to design a multi-heterogeneous robot autonomous task allocation system for underground space detection. By combining the segmentation area and heterogeneous robot cluster relationship establishment unit 1, the problem modeling unit 2, the fault online monitoring unit 3, the interference online observation and prediction unit 4, the transfer learning unit 5, and the task and heterogeneous robot mapping unit 6, the underground space detection task is completed, and the heterogeneous robots are endowed with strong autonomous, strong adaptability, and strong survival intelligent behavior capabilities, thereby improving the efficiency and safety of urban underground pipe gallery detection task allocation, and having the beneficial effects of automation, digitization, and intelligence.

[0047] (2) The present invention integrates task allocation technology and artificial intelligence technology to determine the corresponding relationship between the segmented area and the heterogeneous robot cluster according to the size of the weight.

[0048] (3) The present invention integrates fault monitoring and diagnosis technology, signal processing technology and artificial intelligence technology, uses a finite-time Bernstein network to establish a fault monitoring model, uses historical data to establish monitoring standards, converts them into a binary classification problem, and monitors the abnormal state of each heterogeneous robot.

[0049] (4) The present invention integrates artificial intelligence technology, based on previous similar case reasoning and rule reasoning, and uses past experience to speed up task allocation.

[0050] (5) The present invention integrates anti-interference technology, prediction technology, signal processing technology and artificial intelligence technology, and uses a finite-time Bernstein network to establish an interference observer to quantify and characterize the impact of interference. With the help of historical data, a prediction model is established using a finite-time Bernstein network to predict interference, thereby completing early intervention in interference and reducing the impact of interference.

[0051] (6) The present invention integrates task allocation technology and artificial intelligence technology, adopts a market auction mechanism, and uses the market auction mechanism to build the relationship between each task and each heterogeneous robot in each group, taking into account interference and failure factors, to complete task allocation, effectively improve task allocation efficiency, and reduce computational complexity and communication overhead.

[0052] (7) The finite-time Bernstein network adopted by the present invention, which integrates artificial intelligence technology, has strong nonlinear approximation, fast convergence, strong generalization ability, self-learning ability, robustness and anti-interference performance, and can be well applied as an approximation model and classification model.

[0053] It can be understood that the above specific description of the present invention is only used to illustrate the present invention and is not limited to the technical solutions described in the embodiments of the present invention. Those skilled in the art should understand that the present invention can still be modified or replaced by equivalents to achieve the same technical effects; as long as the use requirements are met, they are within the protection scope of the present invention.

Claims

1. An autonomous task allocation system for multiple heterogeneous robots for underground space exploration, characterized by: It includes a segmentation area and heterogeneous robot cluster relationship establishment unit, a problem modeling unit, a fault online monitoring unit, an interference online observation and prediction unit, a transfer learning unit, and a task and heterogeneous robot mapping unit; The segmented area and heterogeneous robot cluster relationship establishment unit establishes the segmented area and heterogeneous robot cluster relationship through a finite-time Bernstein network, divides the exploration area of ​​the underground space into different sub-areas according to the workload, and divides the heterogeneous robots into multiple heterogeneous robot colonies according to the execution capabilities of the heterogeneous robots themselves, so as to form a functionally complementary heterogeneous robot cluster; Based on the relationship between the segmented areas and the heterogeneous robot cluster, the target optimization function and constraint conditions are described through the problem modeling unit; The operating status of each heterogeneous robot is monitored by the fault online monitoring unit. Once a heterogeneous robot fails, the heterogeneous robot exits the detection task, which serves as a constraint condition of the problem modeling unit; the interference factors in the task allocation are characterized and predicted by the interference online observation and prediction unit. Once external interference is observed, it also serves as a constraint condition of the problem modeling unit; Based on the problem modeling unit, the fault online monitoring unit, and the interference online observation and prediction unit, the transfer learning unit performs reasoning on past similar cases and rule reasoning, and uses past experience to speed up task allocation; On the basis of the transfer learning unit, the task and heterogeneous robot mapping unit constructs the relationship between each task in the segmented area and the heterogeneous robots in each group through a market auction mechanism to complete task allocation; The problem modeling unit utilizes the functional complementarity of heterogeneous robots and considers spatiotemporal constraints to achieve task collaboration and information sharing; based on the relationship between the segmented area and the heterogeneous robot cluster, the task completion degree and global task reward of each cluster are maximized, and the following multi-objective optimization function is established: ; The constraints are: ; In the formula, the task allocation matrix is for: ;in, , ; s is the sequence number of the segmented area; Heterogeneous robots i Ability to perform tasks; Heterogeneous robots i The task sequence package stores heterogeneous robots i The tasks to be performed and their sequencing; To perform tasks j The reward value is and A non-negative function of the correlation; is the number of iterations for each group to converge; The number of heterogeneous robots assigned to each segmented area; Heterogeneous robots i The upper limit of the number of tasks to be executed.

2. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 1 is characterized by: The finite-time Bernstein network is composed of an input layer, an intermediate layer, and an output layer. The expression of the finite-time Bernstein network is as follows: ； in, To expand to The Bernstein network; is the number of expansions; is a finite time term; ; is a symbolic function; is the weight coefficient; ; is the number of middle-layer nodes; is a real-valued function defined on [0,1]; Input for the network; is the number of input items.

3. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 2 is characterized by: The segmented region and heterogeneous robot cluster relationship establishing unit determines the corresponding relationship between the segmented region and the heterogeneous robot cluster according to the weight value obtained by the finite time Bernstein network; the established functional relationship is: , , ,in, represents a heterogeneous robot swarm, represents the number of heterogeneous robot groups, Indicates the number of segmented regions. is the connection weight, Indicates the segmentation area; Described by a finite-time Bernstein network.

4. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 2 is characterized by: The fault online monitoring unit is used to establish a fault monitoring model through a finite-time Bernstein network, combine signal processing technology and historical data, establish a monitoring standard, and convert it into a binary classification problem, 0 represents no fault, 1 represents a fault, and perform abnormal state monitoring on each heterogeneous robot; wherein, the fault monitoring model is established as follows: a finite-time Bernstein network is used as a fault classifier, the input layer uses the robot operation monitoring data after signal processing, and the output layer is respectively fault-free data and faulty data, the fault-free data is marked as "0", and the faulty data is marked as "1", and the network is trained; when a heterogeneous robot is monitored and diagnosed to have a fault, the heterogeneous robot exits the detection task, and other fault-free heterogeneous robots take over.

5. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 2 is characterized by: The interference online observation and prediction unit is used to establish an interference observer through a finite-time Bernstein network to quantify and characterize the interference impact; wherein the interference observer is established as follows: using a finite-time Bernstein network, the input layer uses observation data after signal processing, and the output layer is interference data, and the network is trained; and it is used to establish an interference prediction model through signal processing technology and historical data, and at the same time combine the use of a finite-time Bernstein network to predict interference and complete early intervention in interference; wherein the interference prediction model is established as follows: using a finite-time Bernstein network, the input layer uses historical interference data after signal processing, and the output layer is interference prediction data, and the network is trained.

6. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 1 is characterized by: The previous similar case reasoning includes calculating the similarity between the source case and the target case, and retrieving the historical case with the highest similarity to the target case from the case library. The calculation formula is as follows: ； in, , represent the source case and the target case respectively; is the normalized Euclidean distance between the source case and the target case; k To retrieve the attribute number; n To retrieve the total number of attributes; Number k Attribute weights; Number the source case and the target case k The distance after normalization on the attribute dimension.

7. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 1 is characterized by: The rule reasoning includes establishing a rule base and a decision reasoning mechanism; the rule base is derived from the allocation rules formed by previous task allocation cases, and the decision reasoning mechanism completes the mechanism construction through reasoning technology, expert system, Agent technology, knowledge acquisition and knowledge representation.

8. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 1 is characterized by: The task and heterogeneous robot mapping unit also includes the following situations according to the relationship between the number of heterogeneous robots in the group and the number of tasks in the group: (1) when the number of tasks in the group is equal to the number of heterogeneous robots, a market auction mechanism is used to allocate tasks; (2) when the number of tasks in the group is greater than the number of heterogeneous robots, the efficiency of task allocation needs to consider the constraints of timing factors, and the tasks are released according to the timing, and then a market auction mechanism is used to allocate tasks.

9. The multi-heterogeneous robot autonomous task allocation system for underground space exploration according to claim 8, characterized in that: The market auction mechanism includes allocating tasks to heterogeneous robots through bidding. In the auction algorithm, each heterogeneous robot submits a bid, which represents the price that the heterogeneous robot is willing to complete the task; finally, the task is allocated to the heterogeneous robot with the highest bid.

Images