An Acoustic Synchronous Localization and Mapping Method Based on Robust Potential Probability Assumption Density

Abstract

This invention discloses an acoustic simultaneous localization and mapping method based on robust potential probability hypothesis density, specifically including: recursively propagating the potential probability hypothesis density of the sound source state and DoA observation data; uniformly distributing the minimum and maximum distances from the sound source to the robot; synthesizing the DoA observation data and missing distance information of the sound source state into the mean of the potential probability hypothesis density of the sound source state; updating the mean and covariance of the potential probability hypothesis density of the sound source state using an unscented Kalman filter; calculating the approximate likelihood probability of the potential probability hypothesis density using a variational Bayesian method; generating robot pose particle weights using a single-feature strategy and resampling the robot pose particle weights; pruning and merging the updated mean and covariance of the potential probability hypothesis density of the sound source state; when the weight of the potential probability hypothesis density is greater than a set threshold, the mean of the corresponding potential probability hypothesis density is the sound source location, and the number of weights of the potential probability hypothesis density that meet the condition is the number of sound sources.

Description

Technical Field

[0001] This invention relates to the field of acoustic simultaneous localization, and more particularly to an acoustic simultaneous localization and mapping method based on robust potential probability assumption density. Background Technology

[0002] Over the past few decades, with technological advancements and increasing demands, Simultaneous Localization and Mapping (SLAM) technology has become a research hotspot and has been widely applied in various fields. These applications include mobile robots, navigation, drones, augmented reality (AR), virtual reality (VR), and autonomous vehicles. The core objective of SLAM is to enable robots or vehicles to self-localize in unknown environments while simultaneously constructing environmental maps using observation data acquired by sensors. SLAM observation data is acquired through onboard sensors for environmental measurement and vehicle localization; therefore, localization performance is closely related to sensor characteristics. Visual sensors often struggle to identify targets in abrupt changes in lighting, darkness, or occlusion conditions, while LiDAR data acquisition is inaccurate under dusty and foggy conditions, leading to a decline in SLAM performance. Notably, traditional SLAM algorithms typically rely on Gaussian noise models, assuming that the noise distribution of the observation data follows a Gaussian distribution. However, in real-world scenarios, observation noise is often influenced by various complex factors, causing measurement noise to deviate from a Gaussian distribution, exhibiting non-Gaussianity or even long-tailed distribution characteristics. Traditional algorithms struggle to handle these data anomalies under non-Gaussian noise, leading to a decrease in the accuracy and robustness of SLAM systems.

[0003] In existing technologies, such as the paper "Acoustic Simultaneous Localization and Mapping (a-SLAM) of a Moving Microphone Array and its Surrounding Speakers," Christine et al. proposed a novel acoustic simultaneous localization and mapping (a-SLAM) method that utilizes a moving microphone array to accurately locate the positions of sound sources and sensors in a real-world environment. The proposed method employs a single-cluster probability hypothesis density (SC-PHD) filter, which exhibits good robustness against challenges such as reverberation and measurement interference. By leveraging the spatial diversity of the moving array, the system can infer the distance between the sound source and the sensor, thereby improving the accuracy of sound source mapping and the localization precision of the microphone array. This innovative approach effectively addresses the difficulties faced by traditional methods in complex environments. Experimental results demonstrate a significant improvement in tracking performance and estimation accuracy, fully showcasing the potential and advantages of acoustic SLAM in practical applications.

[0004] In their paper "A Graph Optimization-Based Acoustic SLAM Edge Computing System Offering Centimeter-Level Mapping Services with Reflector Recognition Capability," Zhou et al. proposed a graph optimization-based acoustic SLAM system. First, a robot moves within an indoor environment, emitting pulsed sound waves, and a four-channel microphone array records the room impulse response (RIR). By extracting the cepstral features of the RIR, the system can identify echoes from different reflective materials, achieving echo labeling. Next, using graph optimization methods, combining acoustic data with inertial measurement unit (IMU) data, a graph model with pose constraints is constructed to eliminate accumulated errors during robot movement. Finally, by optimizing the nodes and edges in the graph, the system can accurately locate the robot's trajectory and generate a detailed indoor map including the positions of walls, doors, and windows. The advantages of this acoustic SLAM system include its ability to operate effectively in low-light environments, avoiding reliance on expensive laser or vision sensors, and reducing the cost of the SLAM system. However, the system also has some drawbacks, such as the fact that it may be limited by the propagation characteristics of sound waves in complex indoor environments, which may lead to a decrease in the accuracy of map recognition and location estimation. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention discloses an acoustic simultaneous localization and mapping method based on robust potential probability assumption density, specifically including the following steps:

[0006] The source state and DoA observation data are modeled as random finite sets, where the noise of the DoA observation data follows a Student's T distribution. The potential probability hypothesis density of the source state and DoA observation data is recursively propagated, where the potential probability hypothesis density includes the potential distribution, weights, mean, and covariance.

[0007] The minimum and maximum distances from the sound source to the robot are evenly distributed, and the randomly generated random numbers from the evenly distributed distribution are used as the missing distance information of the sound source state.

[0008] DoA observation data and source state missing distance information are combined to form the mean of the source state birth potential probability hypothesis density, and the mean of the source state birth potential probability hypothesis density is used as the input information for the potential probability hypothesis density prediction process.

[0009] An unscented Kalman filter based on the Student's T-distribution is used to update the mean and covariance of the probability hypothesis density of the sound source state potential.

[0010] The variational Bayesian method is used to calculate the approximate potential probability hypothesis density likelihood probability, which is then used for the weight update of the potential probability hypothesis density.

[0011] A single-feature strategy is used to generate robot pose particle weights. The robot pose particle weights are then weighted and averaged with Rao-Blackwellized filter particles to obtain robot pose estimation. Finally, the robot pose particle weights are resampled.

[0012] The mean and covariance of the updated potential probability hypothesis density of the sound source state are trimmed and merged. When the weight of the potential probability hypothesis density is greater than the set threshold, the mean of the corresponding potential probability hypothesis density is the sound source location, and the number of weights of the potential probability hypothesis density that meet the conditions is the number of sound sources.

[0013] When modeling the sound source state and DoA observation data as a random finite set: the number of models is N. t The state of the sound source is a random finite set s t,n Indicates the state of each sound source

[0014]

[0015] Where, n t,n It is process noise with covariance Q;

[0016]

[0017] The DoA estimation algorithm is used to infer the instantaneous direction of each sound source relative to the robot at time t. The multi-source measurement process is assumed to consist of the union of measurements generated by the target and clutter caused by reverberation. The measurement data... Including a quantity of M t DoA observation data and robot measurement data Robot measurement data includes robot running speed y t,v and direction, y t,γ The DoA observation data noise follows a Student's T-distribution. The DoA observation data is shown below.

[0018] ω t,m =g(s t,n )+e t,m ,e t,m ~St(0 2×1 ,R t,m ,ν)

[0019] Where St(·) is the student's T-distribution, g(·) is the Cartesian to spherical coordinate transformation, ν is the degree of freedom parameter, and e t,m The covariance is R t,m The error in DoA observation data;

[0020]

[0021] Among them, K t It is a random set of spurious measurements with clutter, and each DoA observation is... Azimuth φ t,m ∈[0,2π) and elevation angle θ t,m ∈[0,π).

[0022] Furthermore, when using uniformly distributed randomly generated numbers as distance information for missing sound source states: estimating the distance of each unmeasured sound source relative to the robot, by using J... b The distance of the missing sound source status Initialize the mean of the assumed density of the birth probability of the i-th particle.

[0023]

[0024] in The minimum distance r from the sound source to the robot. min and maximum value r max The uniform distribution of birth probability, the mean of the assumed density. Constructed into

[0025]

[0026] Among them, g -1(·) represents the transformation from spherical coordinates to Cartesian coordinates.

[0027] Furthermore, in the potential probability assumption density prediction step: Gaussian mixture potential probability assumption density is used to estimate the sound source location. The mean of the Gaussian terms represents the possible location of the sound source, and the predicted potential probability assumption density potential distribution is obtained. It is the sum of the birth and survival target potential distributions, that is, the convolution of the birth and survival target potential distributions. Represented as

[0028]

[0029] in, It is the binomial coefficient, P S It is the probability of the target existing. The birth potential distribution at time step t predicts the strength. It is expressed as follows:

[0030]

[0031] in, Indicates birth intensity, Indicates the intensity of the survival objective.

[0032] Furthermore, when updating the mean and covariance of the source state potential probability hypothesis density: when the microphone array receives a source signal, the probability hypothesis density of each source potential is updated individually, where the updated potential distribution... and update strength It is expressed as follows:

[0033]

[0034] Among them, P D It is the target detection probability. Given n targets and Z, multi-target observations t The likelihood is that the noise in the DoA observation data follows a Student's T-distribution for z∈Ω. t Test Items | Ω t | can be represented as

[0035]

[0036] Where Gamma(·) represents the gamma distribution, α t ~Gamma(ν / 2,ν / 2);

[0037] Known robot pose estimation The mean of the j-th Gaussian component Covariance Using an unscented Kalman filter in the update process of equation (15), the predicted state is obtained from the following equation. covariance matrix and cross-covariance matrix Update

[0038]

[0039]

[0040] The updated probability hypothesis density mean and covariance are obtained using the following formulas;

[0041]

[0042] in, This is the Kalman gain.

[0043] Furthermore, when using the variational Bayesian method to calculate the approximate potential probability hypothesis density likelihood probability: an approximation of the target likelihood is obtained.

[0044]

[0045] The variational Bayesian method obtains the result by minimizing the Kullback-Leibler divergence. and

[0046]

[0047] Variational Bayesian methods approximate nonlinear probabilities through an iterative process. Represents the expected value of the auxiliary random variable. A measure of the difference between observed and predicted values.

[0048] Furthermore, the robot pose particle weights are weighted and averaged with the Rao-Blackwellized filter particles to obtain the robot pose estimate: the Rao-Blackwell particle filter is used to obtain the robot pose estimate. The robot's pose is represented by particles with different weights. in

[0049]

[0050] A single-feature strategy is used to generate the robot pose particle weights. The formula for the single-feature strategy is as follows:

[0051]

[0052] in, and Resampling get

[0053] By employing the above technical solution, this invention provides an acoustic synchronous localization and mapping method based on robust potential probability hypothesis density. This method models the sound source state and orientation observation data as a random finite set. The DoA observation data noise follows a Student's T-distribution, and its probability hypothesis density and potential distribution are recursively propagated. The uniform distribution of the minimum and maximum distances from the sound source to the robot is used as missing distance information to construct the range of random sound source state. Gaussian mixture potential probability hypothesis density (CPHD) is used to estimate the sound source state, where the state update of nonlinear observations is completed by an unscented Kalman filter (UKF). A variational Bayesian method is used to approximate the posterior likelihood of the potential probability hypothesis density. A single-feature strategy is used for the particle weighting process of the Rao-Blackwellized filter. Finally, the number and location of the sound sources and the robot's trajectory are jointly estimated. This method does not require a complex data association process, can adapt to changes in the number of sound sources, and has good localization performance and robustness to non-Gaussian outliers. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of the method of the present invention;

[0056] Figure 2 This is a schematic diagram of acoustic SLAM in this invention;

[0057] Figure 3 This is an example diagram of the acoustic SLAM results of the present invention. Detailed Implementation

[0058] To make the technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention:

[0059] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0060] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0061] A flowchart and schematic diagram of an acoustic simultaneous localization and mapping method based on robust potential probability assumption density are shown below. Figure 1 and Figure 2 As shown, the specific steps include the following:

[0062] S1: Model the sound source state and DoA observation data as a random finite set. The number of sets is N. t The state of the sound source is a random finite set s t,n Indicates the state of each sound source

[0063]

[0064] Where, n t,n It is process noise with covariance Q.

[0065]

[0066] B t Simulate the birth process at time t, P(s) t-1,n () represents the survival objective between t-1 and t;

[0067]

[0068] The DoA estimation algorithm is used to infer the instantaneous direction of each sound source relative to the robot at time t. The multi-source measurement process is assumed to consist of the union of measurements generated by the target and clutter caused by reverberation. The measurement data... Including a quantity of M t DoA observation data and robot measurement data Robot measurement data includes robot running speed y t,v and direction y t,γ The DoA observation data noise follows a Student's T-distribution. The DoA observation data is shown below.

[0069] ω t,m =g(s t,n )+e t,m ,e t,m ~St(0 2×1 ,R t,m ,ν) (4)

[0070] Where St(·) is the student's T-distribution, g(·) is the Cartesian to spherical coordinate transformation, ν is the degree of freedom parameter, and e t,m The covariance is R t,m The DoA observation data error. The student T-distribution is represented as follows:

[0071]

[0072] Where d is the dimension of the DoA observation data, Σ=ω t,m -g(s t,n ).

[0073]

[0074] Among them, K t It is a random set of spurious measurements with clutter, and each DoA observation is... Azimuth φ t,m ∈[0,2π) and elevation angle θ t,m ∈[0,π).

[0075] S2: Distribute the minimum and maximum distances from the sound sources to the robot evenly, and use randomly generated numbers from this uniform distribution as the missing distance information for the sound source states. Estimate the distance of each unmeasured sound source relative to the robot using J... b The distance of the missing sound source status Initialize the mean of the assumed density of the birth probability of the i-th particle.

[0076]

[0077] in, The minimum distance r from the sound source to the robot. min and maximum value r max The uniform distribution of birth probability, the mean of the assumed density. Constructed into

[0078]

[0079] Among them, g -1 (·) represents the transformation from spherical coordinates to Cartesian coordinates.

[0080] S3: Potential Probability Assumption Density Prediction Steps. The sound source location is estimated using a Gaussian mixture potential probability assumption density. The mean of the Gaussian terms represents the possible location of the sound source. The predicted potential probability assumption density potential distribution is shown below. It is the sum of the birth and survival target potential distributions, that is, the convolution of the birth and survival target potential distributions. Represented as

[0081]

[0082] in, It is the binomial coefficient, P S It is the probability of the target existing. This is the birth potential distribution at time step t. The predicted strength... It is expressed as follows:

[0083]

[0084] The strength of a survival objective is expressed as

[0085]

[0086] in, Birth intensity is expressed as

[0087]

[0088] in, The number of Gaussian mixture terms representing birth intensity. and These represent the corresponding prediction weights, mean, and covariance of the Gaussian mixture term.

[0089] S4: The potential probability hypothesis density update step is performed by an unscented Kalman filter. Specifically, when the microphone array receives a sound source signal, the potential probability hypothesis density for each sound source is updated individually. The updated potential distribution... and update strength It is expressed as follows:

[0090]

[0091] Among them, P D It is the target detection probability. For z∈Ω t Test Items | Ω t | can be represented as

[0092]

[0093] DoA observation data noise follows a Student's T-distribution. It can be rewritten as

[0094]

[0095] Where Gamma(·) represents the gamma distribution, α t ~Gamma(ν / 2,ν / 2), Given n targets and Z, multi-target observations t The likelihood, p K,t (·) represents the clutter potential distribution at time step t. It is the permutation coefficient, e j (·) is an elementary symmetric function of order j.

[0096]

[0097] in, Let λ be the probability hypothesis density of a random finite set of clutter at time step t, V be the room volume, and λ be the probability density of the set. c It is the clutter rate, and the clutter follows a uniform distribution. The potential probability assumption is the update intensity of the density. The remaining required parameters are as follows:

[0098]

[0099] The unscented Kalman spectroscopy is applied to the update process of the mean and variance of the probability hypothesis density. The robot's pose is known. The mean of the j-th Gaussian component and variance Weighted sigma points It consists of the following formula

[0100]

[0101] Where ε=α 2 (n+κ)-n, where α=0.01 is the scale parameter, κ=1, β=2. The predicted measurement is generated by the following formula.

[0102] The following formula is used to predict the state. covariance matrix and cross-covariance matrix renew.

[0103]

[0104] The updated probability hypothesis density mean and covariance are obtained using the following formulas;

[0105]

[0106] in, This is the Kalman gain.

[0107] S5: Use the variational Bayesian method to approximate the posterior likelihood of the potential probability hypothesis density. Specifically, the key to the variational Bayesian method is obtaining an approximation of the target likelihood. Due to target likelihood Difficult to calculate, but an approximate expression can be obtained.

[0108]

[0109] The variational Bayesian method calculates the result by minimizing the Kullback-Leibler (KL) divergence. and

[0110]

[0111] Variational Bayesian methods approximate nonlinear probabilities through an iterative process. Represents the expected value of the auxiliary random variable. A measure of the difference between observed and predicted values.

[0112]

[0113] S6: Robot pose estimation is obtained by weighting the robot pose particle weights with the Rao-Blackwellized filter particles. Specifically, to obtain the robot pose estimate, a Rao-Blackwell particle filter is used to estimate... The robot's pose is represented by particles with different weights. in

[0114] A single-feature strategy is used to generate the robot pose particle weights. The formula for the single-feature strategy is as follows:

[0115]

[0116] in, and Resampling get

[0117] S7: The estimation of sound source location involves pruning and merging the updated and optimized potential probability hypothesis density weights, mean, and covariance. The mean of the potential probability hypothesis density corresponding to a weight greater than a set threshold is the sound source location, and the number of weights of the potential probability hypothesis density that meet the conditions is the number of sound sources.

[0118] To verify the effectiveness of this invention, the Imgae model was used to simulate a 6×6×3m... 3 In the room, two static sound sources are located at s1 = (1, 3, 1.7) m and s2 = (5, 4, 1.75) m respectively. The mobile robot is equipped with a microphone array and an IMU, which can obtain DoA observation data and robot speed y. t,v and direction y t,γ Information. To ensure the mobile robot is located within the room, it should be positioned within 1 meter of the room boundary, by allowing... The maximum turning radius was enforced. The sound source was a clean speech signal with a sampling frequency of 16kHz randomly selected from the TIMIT database. DoA was calculated using the classic phase transform generalized cross-correlation algorithm (GCC-PHAT). RMSE (root mean square error), MAE (mean absolute error), and OSPA (ptimal subpattern assignment distance) were used to evaluate the performance of the acoustic simultaneous localization and mapping (SMR) method. The localization results of 50 Monte Carlo experiments under different signal-to-noise ratios (SNR) and reverberation conditions are shown in Tables 1 and 2. The test results show that the impact on localization results is still relatively small under low SNR conditions, meeting the requirements of practical applications. The impact on the method in reverberant environments is also relatively small, meeting the requirements of practical applications. All of these demonstrate the good robustness and good operational stability of this invention. Figure 3 Table 3 shows the localization results of different robot trajectories and sound source locations under the conditions of a reverberation time of 300ms and a signal-to-noise ratio of 25dB.

[0119] Table 1. Localization error of acoustic synchronous localization and mapping methods under different SNRs at a reverberation time of 300ms.

[0120]

[0121] Table 2. Acoustic Synchronous Localization and Map Building Methods with Different Reverberation Times at a Signal-to-Noise Ratio of 25 dB: Localization Error

[0122]

[0123] Table 3 Comparison of different indicators for four groups of trajectories

[0124]

[0125]

[0126] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A robust-based potential probability hypothesis density acoustic simultaneous localization and mapping method, characterized in that The application comprises the following steps: Modeling the sound source state and the DoA observation data in the form of a random finite set, wherein the noise of the DoA observation data obeys a Student's T distribution, recursively propagating the potential probability hypothesis density of the sound source state and the DoA observation data, wherein the potential probability hypothesis density comprises a potential distribution, a weight, a mean value and a covariance; Uniformly distributing the minimum and maximum values of the distance from the sound source to the robot, and generating a random number by randomly generating a uniform distribution as the missing distance information of the sound source state; Synthesizing the DoA observation data and the missing distance information of the sound source state into the mean value of the sound source state birth potential probability hypothesis density, and taking the mean value of the sound source state birth potential probability hypothesis density as the input information of the potential probability hypothesis density prediction process; Updating the mean value and the covariance of the sound source state potential probability hypothesis density by using a Student's T distribution-based unscented Kalman filter; Calculating the approximate potential probability hypothesis density likelihood probability by using a variational Bayesian method, and using the approximate potential probability hypothesis density likelihood probability for weight updating of the potential probability hypothesis density; Generating robot pose particle weights by using a single feature strategy, performing weighted averaging on the robot pose particle weights and Rao-Blackwellized filter particles to obtain robot pose estimation, and performing resampling operation on the robot pose particle weights; Pruning and merging the updated mean value and the covariance of the sound source state potential probability hypothesis density, taking the mean value of the potential probability hypothesis density corresponding to a weight greater than a set threshold as the sound source position, and taking the number of the weights of the potential probability hypothesis densities meeting the condition as the number of the sound sources.

2. The robust-based potential probability hypothesis density acoustic simultaneous localization and mapping method according to claim 1, characterized in that: When modeling the sound source states and DoA observation data as a random finite set form: construct a random finite set of sound source states of number N t s t,n representing each sound source state​ where n t,n is process noise with covariance Q; where B t Simulate the birth process at time s t-1,n P(s) represents the survival target between t-1 and The DoA estimation algorithm is used to infer the instantaneous direction of each sound source relative to the robot at time t. It is assumed that the multi-source measurement process is a union of the measurement values generated by the target and the clutter caused by reverberation. The measurement data includes M t number of DoA observation data and robot measurement data The robot measurement data includes the robot running speed y t,v and direction y t,γ , the noise of the DoA observation data obeys the Student's T distribution, and the DoA observation data is as follows ω t,m = g(s t,n ) + e t,m , e t,m ~ St(0 2×1 , R t,m , v) where St(·) is the Student's T distribution, g(·) is the Cartesian to spherical coordinate transformation, v is the degrees of freedom parameter, e t,m is the DoA observation data error with covariance R t,m . where K t is the set of false measurements due to clutter, each DoA observation is azimuth angle φ t,m ∈ [0, 2π) and elevation angle θ t,m ∈ [0, π).

3. The robust-based potential probability hypothesis density acoustic simultaneous localization and mapping method according to claim 2, characterized in that: When the uniformly distributed randomly generated random number is used as the sound source state missing distance information: estimate the distance of each sound source not measured relative to the robot by using J b sound source state missing distance Initialize the mean of the i-th particle birth probability hypothesis density wherein represents the minimum value r of the distance of the sound source from the robot min and the maximum value r of the distance of the sound source from the robot max the mean of the birth probability assumes the density of the uniform distribution is constructed where g -1 (·) is the spherical to Cartesian coordinate transformation.

4. The robust-based acoustic SLAM method of claim 3, wherein: The prediction process of the potential probability hypothesis density is: using the Gaussian mixture potential probability hypothesis density to estimate the sound source position, the mean of the Gaussian term represents the possible position of the sound source, and the predicted potential probability hypothesis density potential distribution is the sum of the birth and survival target potential distributions, i.e. the convolution of the birth and survival target potential distributions, and is expressed as where, is the binomial coefficient, P S is the target presence probability, is the birth potential distribution at time step t, the predicted intensity is represented as follows: wherein, represents the strength of birth, represents the strength of survival goal.

5. The robust-based, potential probability hypothesis density acoustic simultaneous localization and mapping method of claim 4, wherein: The potential probability hypothesis density update process is as follows: when updating the mean and covariance of the potential probability hypothesis density of the sound source state: when the microphone array receives the sound source signal, each sound source potential probability hypothesis density is updated individually, wherein the updated potential distribution and the update intensity is represented as follows: where P D is the target detection probability, is the likelihood of the given n-target multi-target observation Z t with DoA observation data noise following a Student's T distribution, for z e t The detection term |Ω t | can be expressed as where Gamma(·) denotes the gamma distribution, given the robot's estimated pose mean of the jth Gaussian component and covariance The unscented Kalman filter is used for the update process of equation (15), denotes the weighted sigma points of the unscented Kalman filter, which are updated from the predicted state covariance matrix and cross-covariance matrix ​ The updated probability hypothesis density mean value and the covariance are obtained by using the following formula: wherein, is the Kalman gain.

6. The robust-based, potential probability hypothesis density, acoustic simultaneous localization and mapping method of claim 5, wherein: Using a variational Bayesian approach to compute an approximate potential probability hypothesis density likelihood probability: obtain an approximation to the target likelihood The variational Bayesian method computes by minimizing the Kullback-Leibler divergence and Variational Bayesian methods approximate the non-linear probabilities through an iterative process, denotes the expectation of the auxiliary random variable, denotes a measure of the difference between the observation and the prediction.

7. The robust-based acoustic SLAM method of claim 6, wherein: The robot pose estimate is obtained by weighted average of the robot pose particle weights and the Rao-Blackwellized filter particles The robot pose is represented by particles of different weights wherein Generating robot pose particle weights by using a single feature strategy, and the single feature strategy formula is as follows: wherein, and resampling obtained

Images