Underwater robot sensor information fusion method and system based on evidence theory

Abstract

The invention provides an underwater robot sensor information fusion method and system based on an evidence theory, and the method comprises the steps: collecting and preprocessing the measurement data of each sensor of an underwater robot under a single control algorithm, and obtaining data without obvious conflicts; based on the simplified JDL data fusion model and a Pearson coefficient, evaluating the data without obvious conflicts, and based on an evaluation result, performing initial confidence distribution on the data without obvious conflicts; checking the conflict degree between the initial confidence distribution based on the conflict coefficient, and combining and outputting the confidence of which the conflict degree meets a preset threshold based on a D-S evidence theory combination rule; and performing approximate normalization processing on the combined output based on the Bayesian theory to obtain the probability of accuracy of measurement data of each sensor, and completing underwater robot sensor information fusion based on the evidence theory. According to the method, the simplified JDL model and the Pearson coefficient are introduced to perfect the fusion mechanism, and the accuracy and rationality of fused data are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of underwater robot information processing, and in particular relates to an underwater robot sensor information fusion method and system based on evidence theory. Background Art

[0002] With breakthroughs in various technological fields, the field of underwater robots has also ushered in a technological blowout stage. Relying on the birth of various high-precision sensors, the intelligence of underwater robots has become within reach. Under today's social development prospects and trends, the field of underwater robots will surely receive widespread attention from all walks of life. Achievements in the field of underwater exploration will surely be one of the important references for future social evaluations. Therefore, the development of underwater robot-related technologies is particularly important.

[0003] Nowadays, many excellent control algorithms have been developed, but no matter which advanced control algorithm or fault-tolerant algorithm, accurate sensor data is required, and single sensor data has no anti-interference ability and cannot obtain effective and accurate data. Therefore, multi-sensor information fusion is a very important link in the field of underwater robots. With the mutual reference of information from multiple sensors, the interference information can be reasonably and effectively processed, thereby realizing safe and reliable navigation of underwater robots.

[0004] The key to multi-sensor information fusion is to establish the trust distribution of sensors. It is not scientific to give confidence to sensors through expert systems. It is just a rule of thumb. DS evidence theory provides a natural and powerful method for the expression and synthesis of uncertain information. This algorithm can effectively combine the homogeneous or heterogeneous information of other sensors to make all-round, low-error and high-stability realistic estimates and predictions on the basis of fully retaining the uncertainty of data. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an underwater robot sensor information fusion method and system based on evidence theory, so as to effectively solve the problem of poor control accuracy of the underwater robot.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] An underwater robot sensor information fusion method based on evidence theory includes the following steps:

[0008] S1: Collect and pre-process the measurement data of each sensor of the underwater robot under a single control algorithm to obtain data without obvious conflicts;

[0009] S2: Based on the simplified JDL data fusion model and the Pearson coefficient, the data without obvious conflict is evaluated, and initial confidence is assigned to the data without obvious conflict based on the evaluation result;

[0010] S3: checking the conflict degree between the initial confidence allocations based on the conflict coefficient, and combining and outputting the confidences whose conflict degrees meet a preset threshold based on the DS evidence theory combination rule;

[0011] S4: Based on the Bayesian theory, the combined output is approximately normalized to obtain the final probability of the accuracy of the measurement data of each sensor, and the measurement data is weighted summed based on the final probability to obtain the optimal data, thereby completing the underwater robot sensor information fusion based on evidence theory.

[0012] Preferably, in step S1,

[0013] The measurement data of each sensor is preprocessed based on Kalman filtering to obtain Kalman posterior estimation value;

[0014] Performing a residual test on the Kalman posterior estimate and the measurement data of each sensor to obtain a standardized residual;

[0015] Determine whether the standardized residual is greater than a preset residual threshold, and if so, determine that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and set the sensor confidence at the corresponding moment;

[0016] If not, the data without obvious conflict is obtained.

[0017] Preferably, in step S2,

[0018] Taking the data without obvious conflict as evidence bodies, calculating the Pearson coefficients between two evidence bodies in the simplified JDL data fusion model, and obtaining a multi-evidence body Pearson coefficient matrix;

[0019] Based on the Pearson coefficient matrix of the multiple evidence bodies, determine whether the correlation between the data without obvious conflict is lower than a preset correlation threshold; if higher, determine that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and set the sensor confidence at the corresponding moment;

[0020] If it is lower than , the initial confidence level is assigned to the measurement data of the corresponding sensor based on the Pearson coefficient matrix to form a body of evidence.

[0021] Preferably, in step S3, the calculation formula for combining and outputting the evidence body data with a conflict degree lower than a preset threshold based on the DS evidence theory combination rule is as follows:

[0022]

[0023] In the formula, K is a normalization constant, A represents the proposition group or the event group, n is the number of assumptions in the proposition, i represents the i-th body of evidence, and P represents the assumption A in the proposition group. j The final probability value of m represents the body of evidence, B and C are the hypothetical codes for the proposition in the body of evidence, and m i (A j ) represents the jth data of the i-th evidence body in the A proposition group, Represents DS combination rule operation.

[0024] Preferably, in step S4, the formula for approximate normalization processing of the combined output by Bayesian theory is as follows:

[0025]

[0026] In the formula, m i '(A j ) represents the Bayesian normalized value of the jth data of the i-th evidence body in proposition group A, and the numerator represents the evidence body m of proposition group A. i All containing A j The set of hypothetical propositions, the denominator represents the body of evidence m for the proposition group A i Each A j The product of the assumed proposition and its number p in the proposition group;

[0027] Then recalculate the normalization constant K', the formula is as follows:

[0028]

[0029] The final probability of the accuracy of the measurement data of each sensor normalized by Bayesian approximation is obtained, and the measurement data is weighted and output based on the final probability. The calculation formula is as follows:

[0030]

[0031] In the formula, E represents the optimal data output by the multi-sensor, data(A j ) represents A j Assume the measurement data of the jth sensor.

[0032] The present invention also provides an underwater robot sensor information fusion system based on evidence theory, which is used to implement the method, comprising:

[0033] The data acquisition module is used to collect and pre-process the measurement data of each sensor of the underwater robot under a single control algorithm to obtain data without obvious conflicts;

[0034] An initial confidence allocation module, used for evaluating the data without obvious conflict based on a simplified JDL data fusion model and a Pearson coefficient, and allocating initial confidence to the data without obvious conflict based on the evaluation result;

[0035] A DS combination module, used to check the conflict degree between the initial confidence degree allocations based on the conflict coefficient, and combine and output the confidence degrees whose conflict degrees meet a preset threshold based on the DS evidence theory combination rule;

[0036] The normalization module is used to perform approximate normalization processing on the combined output based on Bayesian theory to obtain the final probability of the accuracy of the measurement data of each sensor, and to perform weighted summation on the measurement data based on the final probability to obtain the optimal data, thereby completing the underwater robot sensor information fusion based on evidence theory.

[0037] Preferably, the data acquisition module includes:

[0038] A Kalman processing unit, used for preprocessing the measurement data of each sensor based on Kalman filtering to obtain a Kalman posterior estimation value;

[0039] A residual test unit, used for performing a residual test on the Kalman posterior estimate and the measurement data of each sensor to obtain a standardized residual;

[0040] The first judgment unit is used to judge whether the standardized residual is greater than a preset residual threshold. If so, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if not, the data without obvious conflict is obtained.

[0041] Preferably, the initial confidence allocation module includes:

[0042] A Pearson coefficient calculation unit, used to use the data without obvious conflict as evidence, calculate the Pearson coefficients between two evidence bodies in the simplified JDL data fusion model, and obtain a multi-evidence body Pearson coefficient matrix;

[0043] The second judgment unit is used to judge whether the correlation between the data without obvious conflict meets the preset correlation threshold based on the Pearson coefficient matrix of the multiple evidence bodies. If not, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if so, the measurement data of the corresponding sensor is initially assigned confidence based on the Pearson coefficient.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows: the underwater robot sensor information fusion method based on evidence theory adopted by the present invention can obtain a comprehensive, low-error, and highly stable controller input by combining information from multiple sensors;

[0045] The underwater robot sensor information fusion method based on evidence theory adopted in the present invention can filter out the influence of faulty sensors, reduce the amount of calculation to obtain accurate expected values, improve the rapidity of the system, and provide more accurate expected values ​​for other components that rely on sensor information. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0047] Figure 1 It is a flow chart of the underwater robot sensor information fusion method based on evidence theory according to an embodiment of the present invention;

[0048] Figure 2 Schematic diagram of the JDL model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] Embodiment 1

[0052] like Figure 1 As shown, a sensor information fusion method for underwater robots based on evidence theory includes the following steps:

[0053] S1: Collect and pre-process the measurement data of each sensor of the underwater robot under a single control algorithm to obtain data without obvious conflicts; when there is only one measurement data obtained, calculate the sensor average value and output it directly, otherwise, execute step S1.

[0054] A further implementation method is that in step S1,

[0055] The measurement data of each sensor is preprocessed based on Kalman filtering to obtain Kalman posterior estimation value;

[0056] Perform residual test on Kalman posterior estimation value and measurement data of each sensor to obtain standardized residual;

[0057] In this embodiment, there are many types of data collected by the underwater robot, and data of the same type are processed here. Kalman filtering is performed on the original data.

[0058] x(t)=G*x(t-1)+f*W(t-1)

[0059] Z(t)=H*data(t)+V(t)

[0060] X_hat_pre=G*X_hat(t-1)+W(t)

[0061] Pk_pre=G*Pk(t-1)*G'+Q

[0062] Kk=Pk_pre*H' / (H*Pk_pre*H'+R)

[0063] X_hat(t)=X_hat_pre+Kk*(Z(t)-H*X_hat_pre)

[0064] Pk(t)=(1-Kk*H)*Pk_pre

[0065] Where R and Q are the variances of the measurement error and process error, respectively. G, f, and H represent the state transfer matrix, noise matrix, and observation matrix, respectively. x(t) and Z(t) are the Kalman prediction value and observation value, respectively. W and V are the variance of process noise and observation noise, respectively, and both satisfy Gaussian distribution. X_hat_pre, Pk_pre, Kk, and X_hat represent the prior estimate, prior estimate covariance, Kalman gain, and posterior estimate, respectively.

[0066] It is determined whether the standardized residual is greater than a preset residual threshold. If so, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if not, data without obvious conflict is obtained.

[0067] In this embodiment, the specific test formula is as follows:

[0068] r t =z t -H t *X_hat_pre

[0069]

[0070] where r t is the residual at time t, S t is the residual covariance matrix at time t, v t is the standardized residual, and T represents the transpose of the matrix.

[0071] The standardized residual is compared with the set threshold. If the threshold is exceeded, it is determined that the data of the sensor is inaccurate, and a very small confidence level is given to the sensor. In this embodiment, the sensor confidence level at this moment is set to 0.01.

[0072] S2: Based on the simplified JDL data fusion model and the Pearson coefficient, the data without obvious conflict is evaluated, and the initial confidence level is assigned to the data without obvious conflict based on the evaluation results;

[0073] Among them, Figure 2 As shown, the JDL model includes data processing, target assessment, situation assessment, threat assessment and process refinement.

[0074] Information and processing are the first steps of the entire model. Data preprocessing can screen out suitable data in advance, ensure data accuracy, and eliminate data with large errors.

[0075] The first level of processing is target evaluation, also known as object refinement. In this stage, the input information is transformed into a consistent data structure by using the data processed in the previous layer to perform spatiotemporal alignment, association, correlation, clustering or grouping techniques.

[0076] The second level of processing is situation assessment, which is to determine possible situations based on observed events and acquired data. This part is different from the focus of the first level of processing, which is more focused on individual targets, such as specific objects, people or events, and aims to analyze and evaluate these individual targets, including their attributes, characteristics, threats, etc. Situation assessment focuses on a broader scenario or situation, involving a comprehensive analysis of multiple targets, events and factors, and aims to obtain various relationship information in the entire scenario, including the relationship between targets and the development trend of events, etc.

[0077] The third level of processing is threat assessment, which evaluates the impact of the scenarios detected in the second level and the target activities to obtain appropriate views and make future predictions on the current views to identify and analyze possible risks, vulnerabilities and possible countermeasures in the future. In general, this level of processing includes (1) risk or threat assessment of the identified scenarios and (2) future prediction and analysis of countermeasures based on the assessment results.

[0078] The fourth level is process refinement or process detailing. The purpose of this level of processing is to monitor and evaluate the entire process, provide resource and sensor management for the entire process, and consider task priority, scheduling, and control of available resources, so as to improve the real-time performance of the system.

[0079] The present invention is only applied to the information preprocessing, the second level process and the third level process of the model.

[0080] Rational analysis and optimization of pre-processed data can completely screen out obviously conflicting data, reduce the complexity of subsequent programs, and thus achieve the goal of improving the real-time performance of the system. In addition, since data with higher conflicts is eliminated in advance, the subsequent conflict detection process will become faster and more efficient.

[0081] A further implementation method is that in step S2,

[0082] Taking the data without obvious conflict as evidence, in the simplified JDL data fusion model, the Pearson coefficient between two evidence bodies is calculated to obtain the Pearson coefficient matrix of multiple evidence bodies.

[0083] In this embodiment, the Pearson coefficient is calculated

[0084]

[0085] Where R 12 represents the Pearson coefficient between two bodies of evidence, m1 and m2 are two bodies of evidence, cov represents covariance, and E represents mathematical expectation. and The calculation formulas are:

[0086]

[0087] If there are multiple evidence bodies, calculate the Pearson coefficient matrix R between each pair of evidence bodies. ij :

[0088]

[0089] Based on the Pearson coefficient matrix of multiple evidence bodies, determine whether the correlation between the data without obvious conflict is lower than the preset correlation threshold. If not, determine that the measurement data of the corresponding sensor at the corresponding time is inaccurate, and set the sensor confidence at the corresponding time;

[0090] Specifically, the above matrix can clearly understand the correlation between the two evidence bodies at the current moment. Combined with the relevant Pearson coefficient, a comprehensive judgment is made on whether the sensor is faulty. The specific method is that if the correlation between the values ​​of a certain sensor and other sensors is very poor, it means that the data of the sensor at this moment is not accurate enough. The specific sensor is determined and a confidence level of 0.01 is also given.

[0091] If so, an initial confidence assignment is performed on the measurement data of the corresponding sensor based on the Pearson coefficient.

[0092] Specifically, the other data are then assigned initial confidence and combined into a body of evidence. The formula is as follows:

[0093] m i =1 / n

[0094]

[0095] Where true(m i ) is the evidence body m i The confidence ratio of true(m i )∈[0,1] and

[0096] So the initial confidence distribution is:

[0097]

[0098] S3: Check the degree of conflict between the initial confidence allocations based on the conflict coefficient, and combine and output the confidences whose conflict degrees meet the preset threshold based on the DS evidence theory combination rule; in this embodiment, step S3 also includes checking the degree of conflict between the initial confidence allocations and checking the sensor confidence that has been set to 0.01 in steps S1 and S2.

[0099] A further implementation method is that in step S3, based on the DS evidence theory combination rule, the calculation formula for combining and outputting the confidence level of the conflict degree that meets the preset threshold is as follows:

[0100] The output of DS evidence theory is obtained by performing orthogonal operations on multiple evidence bodies. Generally, ⊕ is used to represent the combination operation, that is,

[0101] The DS combination rule for two evidence bodies m1 and m2 is defined as:

[0102]

[0103] In the formula, K is the normalization constant, A represents the proposition group or the event group, n is the number of assumptions in the proposition, i represents the i-th body of evidence, and P represents the assumption A in the proposition group.j The final probability value of m represents the body of evidence, B and C are the hypothetical codes for the proposition in the body of evidence, and m i (A j ) represents the jth data of the i-th evidence body in the A proposition group, Represents DS combination rule operation.

[0104] In addition, the DS combination rule satisfies the associative law and the commutative law in mathematics, that is, the result of the multi-evidence combination operation is independent of the calculation order.

[0105] S4: Based on Bayesian theory, the combined output is approximately normalized to obtain the probability of the accuracy of each sensor's measurement data, and the underwater robot sensor information fusion based on evidence theory is completed.

[0106] A further implementation method is that, in step S4, the formula for approximate normalization of the combined output based on Bayesian theory is as follows:

[0107]

[0108] In the formula, m i '(A j ) represents the Bayesian normalized value of the jth data of the i-th evidence body in proposition group A, and the numerator represents the evidence body m of proposition group A. i All containing A j The set of hypothetical propositions, the denominator represents the body of evidence m for the proposition group A i Each A j The product of the assumed proposition and its number p in the proposition group.

[0109] Then recalculate the normalization constant K', the formula is as follows:

[0110]

[0111] The final probability of the accuracy of the measurement data of each sensor normalized by Bayesian approximation is obtained, and the measurement data is weighted and output based on the final probability. The calculation formula is as follows:

[0112]

[0113] In the formula, E represents the optimal data output by the multi-sensor, data(A j ) represents A j Assume the measurement data of the jth sensor.

[0114] Embodiment 2

[0115] The present invention also provides an underwater robot sensor information fusion system based on evidence theory, which is used to implement the method, including:

[0116] The data acquisition module is used to collect and pre-process the measurement data of each sensor of the underwater robot under a single control algorithm to obtain data without obvious conflicts;

[0117] An initial confidence allocation module is used to evaluate the data without obvious conflict based on the simplified JDL data fusion model and the Pearson coefficient, and to allocate initial confidence to the data without obvious conflict based on the evaluation results;

[0118] The DS combination module is used to check the conflict degree between the initial confidence allocations based on the conflict coefficient, and combine and output the confidences whose conflict degree meets the preset threshold based on the DS evidence theory combination rule;

[0119] The normalization module is used to perform approximate normalization processing on the combined output based on Bayesian theory to obtain the final probability of the accuracy of the measurement data of each sensor, and to perform weighted summation on the measurement data based on the final probability to obtain the optimal data, thereby completing the underwater robot sensor information fusion based on evidence theory.

[0120] A further implementation is that the data acquisition module comprises:

[0121] A Kalman processing unit, used for preprocessing the measurement data of each sensor based on Kalman filtering to obtain a Kalman posterior estimation value;

[0122] A residual test unit is used to perform a residual test on the Kalman posterior estimate and the measurement data of each sensor to obtain a standardized residual;

[0123] The first judgment unit is used to judge whether the standardized residual is greater than a preset residual threshold. If so, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if not, data without obvious conflict is obtained.

[0124] A further implementation is that the initial confidence allocation module includes:

[0125] The Pearson coefficient calculation unit is used to use the data without obvious conflict as the evidence body, calculate the Pearson coefficient between two evidence bodies in the simplified JDL data fusion model, and obtain the Pearson coefficient matrix of multiple evidence bodies;

[0126] The second judgment unit is used to judge whether the correlation between the data without obvious conflict meets the preset correlation threshold based on the Pearson coefficient matrix of multiple evidence bodies. If not, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if so, the initial confidence is assigned to the measurement data of the corresponding sensor based on the Pearson coefficient.

[0127] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A sensor information fusion method for underwater robots based on evidence theory, characterized in that: The following steps are involved: S1: Collect and pre-process the measurement data of each sensor of the underwater robot under a single control algorithm to obtain data without obvious conflicts; S2: Based on the simplified JDL data fusion model and the Pearson coefficient, the data without obvious conflict is evaluated, and initial confidence is assigned to the data without obvious conflict based on the evaluation result; S3: checking the conflict degree between the initial confidence allocations based on the conflict coefficient, and combining and outputting the confidences whose conflict degrees meet a preset threshold based on the DS evidence theory combination rule; S4: Based on the Bayesian theory, the combined output is approximately normalized to obtain the final probability of the accuracy of the measurement data of each sensor, and the measurement data is weighted summed based on the final probability to obtain the optimal data, thereby completing the underwater robot sensor information fusion based on evidence theory.

2. The underwater robot sensor information fusion method based on evidence theory according to claim 1 is characterized in that: In step S1, The measurement data of each sensor is preprocessed based on Kalman filtering to obtain Kalman posterior estimation value; Performing a residual test on the Kalman posterior estimate and the measurement data of each sensor to obtain a standardized residual; Determine whether the standardized residual is greater than a preset residual threshold, and if so, determine that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and set the sensor confidence at the corresponding moment; If not, the data without obvious conflict is obtained.

3. The underwater robot sensor information fusion method based on evidence theory according to claim 1 is characterized in that: In step S2, Taking the data without obvious conflict as evidence, in the simplified JDL data fusion model, calculating the Pearson coefficients between the source data and the Kalman filter, and obtaining the Pearson coefficient matrix of multiple evidence bodies; Based on the Pearson coefficient matrix of the multiple evidence bodies, determine whether the correlation between the data without obvious conflict is lower than a preset correlation threshold; if higher, determine that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and set the sensor confidence at the corresponding moment; If it is lower than , the initial confidence level is assigned to the measurement data of the corresponding sensor based on the Pearson coefficient matrix to form a body of evidence.

4. The underwater robot sensor information fusion method based on evidence theory according to claim 1 is characterized in that: In step S3, the calculation formula for combining and outputting the evidence body data with a conflict degree lower than a preset threshold based on the DS evidence theory combination rule is as follows: In the formula, K is a normalization constant, A represents the proposition group or the event group, n is the number of assumptions in the proposition, i represents the i-th body of evidence, and P represents the assumption A in the proposition group. j The final probability value of m represents the body of evidence, B and C are the hypothetical codes for the proposition in the body of evidence, and m i (A j ) represents the jth data of the i-th evidence body in the A proposition group, Represents DS combination rule operation.

5. The underwater robot sensor information fusion method based on evidence theory according to claim 1 is characterized in that: In step S4, the formula for approximate normalization of the combined output using Bayesian theory is as follows: In the formula, m′ i (A j ) represents the Bayesian normalized value of the jth data of the i-th evidence body in proposition group A, and the numerator represents the evidence body m of proposition group A. i All containing A j The set of hypothetical propositions, the denominator represents the body of evidence m for the proposition group A i Each A j The product of the assumed proposition and its number p in the proposition group; Recalculate the normalization constant K', the formula is as follows: The final probability of the accuracy of the measurement data of each sensor normalized by Bayesian approximation is obtained, and the measurement data is weighted and output based on the final probability. The calculation formula is as follows: In the formula, E represents the optimal data output by the multi-sensor, data(A j ) represents A j Assume the measurement data of the jth sensor.

6. An underwater robot sensor information fusion system based on evidence theory, used to implement the method described in any one of claims 1 to 5, characterized in that: include: The data acquisition module is used to collect and pre-process the measurement data of each sensor of the underwater robot under a single control algorithm to obtain data without obvious conflicts; An initial confidence allocation module, used for evaluating the data without obvious conflict based on a simplified JDL data fusion model and a Pearson coefficient, and allocating initial confidence to the data without obvious conflict based on the evaluation result; A DS combination module, used to check the conflict degree between the initial confidence degree allocations based on the conflict coefficient, and combine and output the confidence degrees whose conflict degrees meet a preset threshold based on the DS evidence theory combination rule; The normalization module is used to perform approximate normalization processing on the combined output based on Bayesian theory to obtain the final probability of the accuracy of the measurement data of each sensor, and to perform weighted summation on the measurement data based on the final probability to obtain the optimal data, thereby completing the underwater robot sensor information fusion based on evidence theory.

7. The underwater robot sensor information fusion system based on evidence theory according to claim 6 is characterized in that: The data acquisition module comprises: A Kalman processing unit, used for preprocessing the measurement data of each sensor based on Kalman filtering to obtain a Kalman posterior estimation value; A residual test unit, used for performing a residual test on the Kalman posterior estimate and the measurement data of each sensor to obtain a standardized residual; The first judgment unit is used to judge whether the standardized residual is greater than a preset residual threshold. If so, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if not, the data without obvious conflict is obtained.

8. The underwater robot sensor information fusion system based on evidence theory according to claim 6 is characterized in that: The initial confidence allocation module comprises: A Pearson coefficient calculation unit, used to use the data without obvious conflict as evidence, calculate the Pearson coefficients between two evidence bodies in the simplified JDL data fusion model, and obtain a multi-evidence body Pearson coefficient matrix; The second judgment unit is used to judge whether the correlation between the data without obvious conflict is lower than a preset correlation threshold based on the Pearson coefficient matrix of the multiple evidence bodies. If not, it is determined that the measurement data of the corresponding sensor at the corresponding moment is inaccurate, and the sensor confidence at the corresponding moment is set; if so, the measurement data of the corresponding sensor is initially assigned confidence based on the Pearson coefficient.

Images