A decoupling adaptation method for a special robot system suitable for all-terrain operations

Abstract

The present invention discloses a decoupling and adaptation method for a special robot system that is adaptable to all-terrain operations. The method takes into account three-dimensional decoupling and adaptation criteria, and adopts a multi-view consensus clustering method that integrates spectral clustering and non-negative matrix decomposition to achieve the solution of the optimal decoupling and adaptation scheme for the design of different subsystem functional units such as suspension, shock absorption, steering, sensor planning, and motion control of the special robot for all-terrain operations. The method actively decouples and adapts the special robot system that is adaptable to all-terrain operations from the top-level design level, which can help effectively improve the upper limit of its achievable comprehensive performance level from a global perspective.

Description

Technical Field

[0001] The present invention relates to a decoupling and adaptation method for a robot system, and more particularly to a decoupling and adaptation method for a special robot system suitable for all-terrain operations. Background Art

[0002] Special robots adapted to all-terrain operations have important research and application value in fields such as logistics handling, safety search and rescue, field operations, deep space exploration, and military operations. Ordinary robots generally consider structured operating environments and flat terrain. It is assumed that the impact of terrain factors on the robot's operating performance is relatively small or can be regarded as a passive environmental disturbance. Special robots adapted to all-terrain operations are closely related to terrain attributes. The environment in which the robot is located may be an unknown or incompletely known dangerous environment, with rocks, potholes, soft terrain, rugged terrain, and complex geological conditions. When moving in such an unstructured environment, the complexity of the environmental terrain has a great impact on the robot's operating performance and terrain passability. Therefore, for special robot systems adapted to all-terrain operations, the robot and the terrain environment are inseparable as a whole. When designing the system, it is necessary to fully consider the possible impact of terrain factors on the robot's operating performance.

[0003] In the current related research and applications, the focus is generally on specific application scenarios and specific terrain environments, and then on the corresponding aspects such as mechanism design, autonomous motion, multi-sensor fusion, environmental modeling, visual navigation, terrain classification, path planning, configuration planning, wireless communication and intelligent control that meet the operational needs. It is more inclined to bottom-up performance optimization design, while neglecting top-down comprehensive performance optimization. It lacks a general method to focus on improving the comprehensive performance level of special robot systems for sustainable operations in different complex terrain environments from the perspective of the top-level design of the special robot system coupling, while taking into account the cognitive ambiguity in the top-level design of the system. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a decoupling and adaptation method for a special robot system that is suitable for all-terrain operations. The proposed method takes into account the decoupling and adaptation criteria of three dimensions, and adopts a multi-view consensus clustering method that integrates spectral clustering and non-negative matrix decomposition to achieve the optimal decoupling and adaptation solution for the design of different subsystem functional units such as suspension, shock absorption, steering, sensor planning, and motion control of the special robot for all-terrain operations. Active decoupling and adaptation of the special robot system that is suitable for all-terrain operations is performed from the top-level design level, which can help effectively improve the upper limit of its comprehensive performance level that can be achieved from a global perspective.

[0005] To achieve the above object, the technical solution provided by the present invention comprises the following steps:

[0006] Step 1: Calculate and obtain the similarity matrix {G (1) , G (2) , G (3)}, set the decoupling cluster number range [c min ,c max ];

[0007] Step 2: Take the number of decoupled clusters c = c min , run the multi-view consensus decoupling adaptation algorithm, initialize {F (1) , F (2) , F (3)} and H, let F (j) G (j) The eigenvectors of the first c largest eigenvalues ​​of are composed of, and, depending on whether the condition is met Determine whether to continue iteratively updating H and F (j) , Q (j) If the value of H and F is satisfied, iterative update (j) , Q (j) , the optimal decoupling adaptation solution under the decoupling cluster number c is output, and then the optimal decoupling adaptation solution under the condition that c≤c max In this case, the multi-view consensus decoupling adaptation algorithm is used to calculate the clustering solutions under different decoupling clustering numbers c in turn;

[0008] Step three, after obtaining the clustering schemes under different decoupling cluster numbers, taking into account the three-dimensional decoupling adaptation criteria, calculate the quantitative evaluation value corresponding to each clustering scheme, and then, based on the needs, select the optimal clustering scheme as the final decoupling adaptation scheme of the system based on fuzzy comprehensive evaluation or weighted method.

[0009] As a further improvement of the present invention, the specific process of implementing the multi-view consensus decoupling adaptation algorithm in step 2 is as follows:

[0010] The objective of multi-view consensus clustering is defined as:

[0011]

[0012] Among them, ||.|| 2，1 is the L21 norm; G (v) is the normalized similarity matrix corresponding to the v-th view, w ij Indicates the similarity between two sample points; F (v) is the embedding matrix corresponding to the vth view; H is the clustering label matrix shared by multiple views.

[0013] According to the derivation, the above target declaration can be equivalently transformed into the following formula:

[0014]

[0015] in, is the residual; Q is a diagonal matrix, Its value update depends on the size of the residual in the previous iteration, which can be regarded as the weight of the v-th view, showing adaptability; and Refers to F (v) and h i,: The current solution of .

[0016] In order to find the optimal solution to the above equation, it is necessary to iteratively update Q (v) 、F (v) and H.

[0017] When fixed H updates F (v) hour:

[0018] F (v) =UV T

[0019] Among them, U and V are matrices S (v) The singular value decomposition S (v) =G (v) Q (v) H=UΣV T U and V in, U is given by (S (v) ) T S (v) The eigenvectors of V are composed of S (v) (S (v) ) T The feature vector of .

[0020] When fixing F to update H:

[0021]

[0022] After normalization, {G (v) ∈R n×n} and the number of clusters c as input, the main steps of the multi-view consensus clustering algorithm include: initialization {F (v)} and H. Use G (v) Initialize F with the eigenvectors corresponding to the first c eigenvalues ​​in descending order (v) ; For each view v, update Q (v) 、F (v) , H; judge whether it satisfies (J t is the optimization objective function value at the tth iteration), if satisfied, the iteration is stopped and the cluster label matrix H is output, otherwise it returns to continue iterating and updating Q(v) 、F (v) , H.

[0023] As a further improvement of the present invention, the three-dimensional decoupling adaptation criterion in step 3 includes the following criteria:

[0024] Functional interaction dependency clustering criterion: to increase the cohesion of functional interaction dependency within the clustering granularity, while reducing the coupling of functional interaction dependency between clustering granularities;

[0025] Performance characteristic association clustering criterion: to increase the similarity of performance characteristic requirements within the clustering granularity, while reducing the coupling of performance characteristic requirements between clustering granularities;

[0026] Clustering principle for comprehensive adaptability of terrain environment: to increase the adaptability of the corresponding subsystem set to different complex terrain environments after increasing the granularity of clustering, that is, to maximize the terrain environment adaptability evaluation value of each clustering granularity on the basis of priority.

[0027] As a further improvement of the present invention, the quantitative evaluation index of the functional interaction dependency clustering criterion is specifically expressed by the following formula:

[0028]

[0029] Among them, VI is the variance of the cohesion degree of each clustering granularity, ICP i The clustering result PA i The degree of dependency coupling, CA i T i The clustering coefficient, ICA g is the particle size Tp i degree of cohesion.

[0030] As a further improvement of the present invention, the quantitative evaluation index of the functional interaction dependency clustering criterion is obtained by the following steps:

[0031] Assume that the matrix IC n×n is the functional interaction relationship matrix between functional units, when ICD ij +ICD ij >0, Ic ij =Ic ji =1, otherwise 0; Matrix RA n×z is the n functional units and the clustering results PA i The corresponding relationship matrix of the z partition granularities, Ra ij The value is 1 or 0, indicating the functional unit T i Whether it belongs to the particle size Tp i ;ICP z×z is the functional interaction relationship matrix between granularities. If Icp ij=Icp ji =1, otherwise it is 0, and the particle size Tp i The degree of association within a set of functional units is defined as:

[0032]

[0033] Then refer to the definition of clustering coefficient in network analysis and calculate T i The aggregation coefficient CA i and its particle size Tp i The average clustering coefficient of the internal functional unit set is CAM g for:

[0034]

[0035] So we can get the particle size Tp i ICA g for:

[0036] ICA g =CAM g ×RCA g

[0037] For the correlation between granularities, the functional correlation degree across granularities and the clustering coefficient between granularities are considered. i RCT of the functional correlation between mesoscale particles i It can be calculated as:

[0038]

[0039] Particle size Tp i The clustering coefficient CP g And the clustering results PA i The average clustering coefficient CPM of the particle size set within i It can be calculated as:

[0040]

[0041] Thus, the clustering result PA is obtained i Dependency coupling degree ICP i for:

[0042] ICP i =CPM i ×RCT i

[0043] Finally, the quantitative evaluation index of the functional interaction dependency clustering criterion is obtained as follows:

[0044]

[0045] Where VI is the variance of the cohesion degree of each particle size.

[0046] As a further improvement of the present invention, the quantitative evaluation index of the performance characteristic association clustering criterion is as follows:

[0047]

[0048] Among them, IQ j Q is the performance characteristic j Importance of QCM i Reflects the average level of convergence and independence of performance targets at each granularity; IQP i and VQP i The performance characteristics Q j The mean and variance of the impact on each partition granularity.

[0049] As a further improvement of the present invention, the quantitative evaluation index of the performance characteristic association clustering criterion is obtained by the following steps:

[0050] Assume QT k×n is the correlation matrix between k performance characteristics and n functional units, Qt ij Indicates the degree of influence of the i-th performance characteristic on the j-th functional unit, BCT k×n For QT k×n The Boolean value of the conversion matrix, when Qt ij >0, Bct ij =1, otherwise 0, comprehensive matrix BCT k×n and the matrix RA n×z , we can get the Boolean relationship matrix QP between k performance characteristics and z partition granularities k×z , performance characteristics Q i The average IQP of the impact on each partition granularity i and VQP i It can be calculated as:

[0051]

[0052] For the clustering results PA i , the quantitative evaluation index of the performance characteristic association clustering criterion is defined as:

[0053]

[0054] Among them, IQ j Q is the performance characteristic j Importance of QCM i It reflects the average level of convergence and independence of performance characteristics of each particle size.

[0055] As a further improvement of the present invention, the quantitative evaluation index of the terrain comprehensive adaptability clustering principle is as follows:

[0056] For the clustering result PA containing z partition granularities i , the quantitative evaluation index of the terrain comprehensive matching clustering criterion is defined as:

[0057]

[0058] Among them, wtp i is the particle size Tp i The weight of Y can be obtained by comparing and integrating the weights of each functional unit contained in the granularity; j is the number of functional units contained in the granularity; wa j is the weight of the jth functional unit within the granularity; RC i Tp i VR is the variance of the terrain adaptability evaluation values ​​of all z granularities. As a further improvement of the present invention, the clustering scheme in step eight includes a standard spectral clustering algorithm, a non-negative matrix factorization algorithm, a multi-view RMS clustering algorithm combining spectral clustering with NMF, and a design intent clustering method based on multi-view consensus.

[0059] The beneficial effects of the present invention are as follows: the existing methods lack a focus on bottom-up performance optimization design, while relatively ignoring top-down comprehensive performance optimization, and lack a focus on the top-level design level of special robot system coupling. While taking into account the cognitive ambiguity existing in the top-level design of the system, the present invention focuses on the problem of a general method for improving the comprehensive performance level of special robot systems that can adapt to sustainable operations in different complex terrain environments. The present invention actively decouples and adapts special robot systems that can adapt to all-terrain operations from the top-level design level, which can help effectively improve the upper limit of the comprehensive performance level that can be achieved from a global perspective, and at the same time has the advantages of fewer hyperparameter settings and high computational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic flow chart of a decoupling adaptation method for a special robot system adapted to all-terrain operations according to the present invention;

[0061] Figure 2 Schematic diagram of the multidimensional decoupling adaptation criterion designed for special robotic systems. DETAILED DESCRIPTION

[0062] The present invention will be further described below in detail with reference to the embodiments shown in the accompanying drawings.

[0063] Reference Figure 1 As shown, a decoupling adaptation method of a special robot system adapted to all-terrain operations in this embodiment includes the following steps:

[0064] Step 1: Calculate and obtain the similarity matrix {G (1) , G (2) , G (3)}, set the cluster number range [c min ,c max ];

[0065] Step 2: Take the number of clusters c = c min ;

[0066] Step 3: Initialize {F (1) , F (2) , F (3)} and H, let F (j) G (j) The eigenvectors of the first c largest eigenvalues ​​of , where j = 1, 2, 3, let j = 1, t = 0, J 0 =0;

[0067] Step 4: When 1≤j≤3, set j=j, otherwise, set j=1; then update H and F in turn by the formula (j) and G (j) The value of

[0068] Step 5: Let t = t + 1 to calculate J t ;

[0069] Step 6: Determine J t Is it satisfied? If not, set j=j+1 and return to step 4. If satisfied, proceed to the next step.

[0070] Step 7: Obtain the optimal clustering solution under the cluster number c based on H, and determine whether c>c max If not, set c=c+1 and return to step 3. If satisfied, proceed to the next step.

[0071] Step eight, based on the decoupled clustering criteria in three dimensions, namely, functional interaction correlation, performance characteristic coupling correlation, and terrain environment comprehensive adaptability, calculate the quantitative evaluation value corresponding to each clustering scheme, and then select the optimal scheme based on fuzzy comprehensive evaluation or weighted method according to the needs.

[0072] The three-dimensional clustering method in step 8 is shown in Figure 2 As shown in the figure, its connotation and quantitative evaluation implementation are described in detail as follows:

[0073] (1) Functional interaction dependency clustering criterion: to increase the cohesion of functional interaction dependencies within the clustering granularity and reduce the coupling of functional interaction dependencies between clustering granularities.

[0074] Assume that the matrix IC n×n is the functional interaction relationship matrix between functional units, when ICD ij +ICD ij >0, Ic ij =Ic ji =1, otherwise 0; Matrix RA n×z is the n functional units and the clustering results PA i The corresponding relationship matrix of the z partition granularities, Ra ij The value is 1 or 0, indicating the functional unit T i Whether it belongs to the particle size Tp i ;ICP z×z is the functional interaction relationship matrix between granularities. If Icp ij =Icp ji =1, otherwise it is 0, and the particle size Tp i The degree of association within a set of functional units is defined as:

[0075]

[0076] Then refer to the definition of clustering coefficient in network analysis and calculate T i The aggregation coefficient CA i and its particle size Tp i The average clustering coefficient of the internal functional unit set is CAM g for:

[0077]

[0078] So we can get the particle size Tp i ICA g for:

[0079] ICA g =CAM g ×RCA g

[0080] For the correlation between granularities, the functional correlation degree across granularities and the clustering coefficient between granularities are considered. i RCT of the functional correlation between mesoscale particles i It can be calculated as:

[0081]

[0082] Particle size Tp i The clustering coefficient CP g And the clustering results PA i The average clustering coefficient CPM of the particle size set within iIt can be calculated as:

[0083]

[0084] Thus, the clustering result PA is obtained i Dependency coupling degree ICP i for:

[0085] ICP i =CPM i ×RCT i

[0086] Finally, the quantitative evaluation index of the functional interaction dependency clustering criterion is obtained as follows:

[0087]

[0088]

[0089] Where VI is the variance of the cohesion degree of each particle size.

[0090] (2) Performance characteristic association clustering criterion: to increase the similarity of performance characteristic requirements within the clustering granularity, while reducing the coupling of performance characteristic requirements between clustering granularities.

[0091] Assume QT k×n is the correlation matrix between k performance characteristics and n functional units, Qt ij Indicates the degree of influence of the i-th performance characteristic on the j-th functional unit, BCT k×n For QT k×n The Boolean value of the conversion matrix, when Qt ij >0, Bct ij =1, otherwise 0, comprehensive matrix BCT k×n and the matrix RA n×z , we can get the Boolean relationship matrix QP between k performance characteristics and z partition granularities k×z , performance characteristics Q i The average IQP of the impact on each partition granularity i and VQP i It can be calculated as:

[0092]

[0093] For the clustering results PA i , the quantitative evaluation index of the performance characteristic association clustering criterion is defined as:

[0094]

[0095] Among them, IQ j Q is the performance characteristic jImportance of QCM i It reflects the average level of convergence and independence of performance characteristics of each particle size.

[0096] (3) Clustering principle of comprehensive adaptability of terrain environment: Increase the adaptability of the corresponding subsystem set to different complex terrain environments after increasing the granularity of clustering division, that is, maximize the terrain environment adaptability evaluation value of each clustering granularity on the basis of considering the priority.

[0097] For the clustering result PA containing z partition granularities i , the quantitative evaluation index of the terrain comprehensive matching clustering criterion is defined as:

[0098]

[0099] Among them, wtp i is the particle size Tp i The weight of Y can be obtained by comparing and integrating the weights of each functional unit contained in the granularity; j is the number of functional units contained in the granularity; wa j is the weight of the jth functional unit within the granularity; RC i Tp i The terrain adaptability evaluation value of ; VR is the variance of the terrain adaptability evaluation values ​​of all z granularities.

[0100] The method corresponding to steps 2 to 7 in this embodiment is a multi-view consensus clustering method. The calculation process of this method is described in detail below in conjunction with the clustering problem of clustering a sample X containing V view information into c cluster groups:

[0101] The objective of multi-view consensus clustering is defined as:

[0102]

[0103] Among them, ||.|| 2，1 is the L21 norm; G (v) is the normalized similarity matrix corresponding to the v-th view, w ij Indicates the similarity between two sample points; F (v) is the embedding matrix corresponding to the vth view; H is the clustering label matrix shared by multiple views.

[0104] According to the derivation, the above target declaration can be equivalently transformed into the following formula:

[0105]

[0106] in, is the residual; Q is a diagonal matrix, Its value update depends on the size of the residual in the previous iteration, which can be regarded as the weight of the v-th view, showing adaptability; and Refers to F (v) and h i,: The current solution of .

[0107] In order to find the optimal solution to the above equation, it is necessary to iteratively update Q (v) 、F (v) and H.

[0108] When fixed H updates F (v) hour:

[0109] F (v) =UV T

[0110] Among them, U and V are matrices S (v) The singular value decomposition S (v) =G (v) Q (v) H=UΣV T U and V in, U is given by (S (v) ) T S (v) The eigenvectors of V are composed of S (v) (S (v) ) T The feature vector of .

[0111] When fixing F to update H:

[0112]

[0113] After normalization, {G (v) ∈R n×n} and the number of clusters c as input, the main steps of the multi-view consensus clustering algorithm include: initialization {F (v)} and H. Use G (v) Initialize F with the eigenvectors corresponding to the first c eigenvalues ​​in descending order (v) ; For each view v, update Q (v) 、F (v) , H; judge whether it satisfies (J t is the optimization objective function value at the tth iteration). If it is satisfied, the iteration is stopped and the cluster label matrix H is output. Otherwise, it returns to step 2 to continue the iteration.

[0114] To sum up, the decoupling and adaptation method of a special robot system suitable for all-terrain operations in this embodiment takes into account three-dimensional decoupling and adaptation criteria, and adopts a multi-view consensus clustering method that integrates spectral clustering and non-negative matrix decomposition to achieve the optimal decoupling and adaptation solution for the design of different subsystem functional units such as suspension, shock absorption, steering, sensor planning, and motion control of the special robot for all-terrain operations. The special robot system suitable for all-terrain operations is actively decoupled and adapted from the top-level design level to help improve the upper limit of its achievable comprehensive performance level from a global perspective.

[0115] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A decoupling adaptation method for a special robot system adapted to all-terrain operations, characterized in that: The steps include: Step 1: Calculate and obtain the similarity matrix {G (1) , G (2) , G (3) }, set the decoupling cluster number range [c min ,c max ]; Step 2: Take the number of decoupled clusters c = c min , run the multi-view consensus decoupling adaptation algorithm, initialize {F (1) , F (2) , F (3) } and H, let F (j) G (j) The eigenvectors of the first c largest eigenvalues ​​of are composed of, and, depending on whether the condition is met Determine whether to continue iteratively updating H and F (j) , Q (j) If the value of H and F is satisfied, iterative update (j) , Q (j) , the optimal decoupling adaptation solution under the decoupling cluster number c is output, and then the optimal decoupling adaptation solution under the condition that c≤c max In this case, the multi-view consensus decoupling adaptation algorithm is used to calculate the clustering solutions under different decoupling clustering numbers c in turn; Step 3: After obtaining the clustering schemes under different decoupling clustering numbers, the quantitative evaluation values ​​corresponding to each clustering scheme are calculated taking into account the three-dimensional decoupling adaptation criteria. Then, based on the needs, the optimal clustering scheme is selected as the final decoupling adaptation scheme of the system based on fuzzy comprehensive evaluation or weighted method. The specific process of implementing the multi-view consensus decoupling adaptation algorithm in step 2 is as follows: The objective of multi-view consensus clustering is defined as: Among them, ||.|| 2，1 is the L21 norm; G (v) is the normalized similarity matrix corresponding to the v-th view, w ij Indicates the similarity between two sample points; F (v) is the embedding matrix corresponding to the vth view; H is the clustering label matrix shared by multiple views; According to the derivation, the above target formula is equivalent to the following formula: in, is the residual; Q is a diagonal matrix, Its value update depends on the size of the residual in the previous iteration, which can be regarded as the weight of the v-th view, showing adaptability; and Refers to F (v) and h i,: The current solution of In order to find the optimal solution to the above equation, it is necessary to iteratively update Q (v) 、F (v) and H. When fixed H updates F (v) hour: F (v) =UV T Among them, U and V are matrices S (v) The singular value decomposition S (v) =G (v) Q (v) H=UΣV T U and V in, U is given by (S (v) ) T S (v) The eigenvectors of V are composed of S (v) (S (v) ) T The feature vector of When fixing F to update H: After normalization, {G (v) ∈R n×n } and the number of clusters c as input, the multi-view consensus clustering algorithm steps include: initialization {F (v) } and H, use G (v) Initialize F with the eigenvectors corresponding to the first c eigenvalues ​​in descending order (v) ; For each view v, update Q (v) 、F (v) , H; judge whether it satisfies J t is the optimization objective function value at the tth iteration. If it is satisfied, the iteration stops and the cluster label matrix H is output. Otherwise, it returns to continue iterating and updating Q. (v) 、F (v) , H.

2. The decoupling adaptation method of a special robot system adapted to all-terrain operations according to claim 1 is characterized in that: The three-dimensional decoupling adaptation criteria in step 3 include the following criteria: Functional interaction dependency clustering criterion: to increase the cohesion of functional interaction dependency within the clustering granularity, while reducing the coupling of functional interaction dependency between clustering granularities; Performance characteristic association clustering criterion: to increase the similarity of performance characteristic requirements within the clustering granularity, while reducing the coupling of performance characteristic requirements between clustering granularities; Clustering criterion for comprehensive adaptability of terrain environment: the adaptability of the corresponding subsystem set to different complex terrain environments after increasing the granularity of clustering division, that is, the terrain environment adaptability evaluation value of each clustering granularity is increased as much as possible on the basis of considering the priority.

3. The decoupling adaptation method of a special robot system adapted to all-terrain operations according to claim 2 is characterized in that: The quantitative evaluation index of the functional interaction dependency clustering criterion is specifically expressed by the following formula: Among them, VI is the variance of the cohesion degree of each clustering granularity, ICP i The clustering result PA i The degree of dependency coupling, CA i T i The clustering coefficient, ICA g is the particle size Tp i degree of cohesion.

4. The decoupling adaptation method of a special robot system adapted to all-terrain operations according to claim 3 is characterized in that: The quantitative evaluation index of the functional interaction dependency clustering criterion is obtained by the following steps: Assume matrix IC n×n is the functional interaction relationship matrix between functional units, when ICD ij +ICD ij >0, Ic ij =Ic ji =1, otherwise 0; Matrix RA n×z is the n functional units and the clustering results PA i The corresponding relationship matrix of the z partition granularities, Ra ij The value is 1 or 0, indicating the functional unit T i Whether it belongs to the particle size Tp i ;ICP z×z is the functional interaction relationship matrix between granularities. If Icp ij =Icp ji =1, otherwise it is 0, and the particle size Tp i The degree of association within a set of functional units is defined as: Then refer to the definition of clustering coefficient in network analysis and calculate T i The aggregation coefficient CA i and its particle size Tp i The average clustering coefficient of the internal functional unit set is CAM g for: So we can get the particle size Tp i ICA g for: ICA g =CAM g ×RCA g For the correlation between granularities, the functional correlation degree across granularities and the clustering coefficient between granularities are considered; the clustering result PA i RCT of the functional correlation between mesoscale particles i Calculated as: Particle size Tp i The clustering coefficient CP g And the clustering results PA i The average clustering coefficient CPM of the particle size set within i Calculated as: Thus, the clustering result PA is obtained i Dependency coupling degree ICP i for: ICP i =CPM i ×RCT i Finally, the quantitative evaluation index of the functional interaction dependency clustering criterion is obtained as follows: Where VI is the variance of the cohesion degree of each particle size.

5. The decoupling adaptation method of a special robot system adapted to all-terrain operations according to claim 2 is characterized in that: The quantitative evaluation index of the performance characteristic association clustering criterion is as follows: Among them, IQ j Q is the performance characteristic j Importance of QCM i Reflects the average level of convergence and independence of performance targets at each granularity; IQP i and VQP i The performance characteristics Q j The mean and variance of the impact on each partition granularity.

6. The decoupling adaptation method of a special robot system adapted to all-terrain operations according to claim 5 is characterized in that: The quantitative evaluation index of the performance characteristic association clustering criterion is obtained by the following steps: Assume QT k×n is the correlation matrix between k performance characteristics and n functional units, Qt ij Indicates the degree of influence of the i-th performance characteristic on the j-th functional unit, BCT k×n For QT k×n The Boolean value of the conversion matrix, when Qt ij >0, Bct ij =1, otherwise 0, comprehensive matrix BCT k×n and the matrix RA n×z , we can get the Boolean relationship matrix QP between k performance characteristics and z partition granularities k×z , performance characteristics Q i The average IQP of the impact on each partition granularity i and VQP i Calculated as: For the clustering results PA i , the quantitative evaluation index of the performance characteristic association clustering criterion is defined as: Among them, IQ j Q is the performance characteristic j Importance of QCM i It reflects the average level of convergence and independence of performance characteristics of each particle size.

7. The decoupling adaptation method of a special robot system adapted to all-terrain operations according to claim 2 is characterized in that: The quantitative evaluation index of the clustering criterion of the comprehensive adaptability of the terrain environment is as follows: For the clustering result PA containing z partition granularities i , the quantitative evaluation index of the terrain comprehensive matching clustering criterion is defined as: Among them, wtp i is the particle size Tp i The weight of Y can be obtained by comparing and integrating the weights of each functional unit contained in the granularity; j is the number of functional units contained in the granularity; wa j is the weight of the jth functional unit within the granularity; RC i Tp i The terrain adaptability evaluation value of ; VR is the variance of the terrain adaptability evaluation values ​​of all z granularities.

Images