A SLAM method based on rodent models and WIFI fingerprints

Abstract

The present invention discloses a SLAM method based on a rodent model and WIFI fingerprint, which relates to the fields of bionics and wireless signal networks. Using a fingerprint recognition method based on WIFI signal strength, it replaces the local scene cell network in the original rodent model with WIFI fingerprint information. An on-position fingerprint database is established in the offline stage, and the Bayesian algorithm is used to complete the matching of WIFI signal strength fingerprint information in the online positioning stage, correct the activity of the pose cell network, and finally obtain a more accurate experience map. By using the SLAM method based on a rodent model and WIFI fingerprint of the present invention, not only is the positioning accuracy of the mobile robot higher, but also the stability of the system is improved, and it has good positioning performance.

Description

Technical Field

[0001] This invention relates to the fields of bionics and wireless signal networks, specifically to a SLAM method based on rodent models and WIFI fingerprints. Background Technology

[0002] Simultaneous localization and mapping (SLAM) are major challenges currently facing mobile robots. This is because mobile robots are essentially mobile sensor platforms, and while sensors vary in type and capability, they widely suffer from problems such as odometry drift and varying degrees of noise. Through continuous research, biomimetic robots have gradually shown promising application prospects, exhibiting perfect biological rationality and a high degree of adaptability to the natural environment. Among these, rodent models have received the most research.

[0003] This model integrates path integrals and visual scene information into a pose-aware cell model, enabling the mobile robot to have a certain ability to update and predict. Simultaneously, it establishes an experience mapping algorithm based on temporal, spatial location, and behavioral information, which is now widely used in robot localization and navigation, solving many problems that SLAM struggles with. However, the visual scene information and odometry information acquired by the rodent model have certain errors. To address the odometry error, FAB-MAP (fast appearance based mapping), a closed-loop detection algorithm based on a historical model, is introduced. Through real-time keyframe matching, it can improve system stability, but the localization accuracy is unstable and its robustness is weak. Therefore, the rodent model alone needs further improvement in terms of localization accuracy and robustness. Summary of the Invention

[0004] The purpose of this invention is to provide a SLAM method based on rodent models and WIFI fingerprints to overcome the aforementioned deficiencies in the prior art.

[0005] A SLAM method based on a rodent model and WIFI fingerprinting, the method comprising the following steps:

[0006] Step 1) Using a competitive attractor network, head orientation cells and position cells are modeled as pose sensing cells. After processing by the pose cells, the robot's self-centered and non-centered information stimuli affect the activity of the pose cells.

[0007] Step 2) Use the wireless signal network WIFI as a sensor in the rodent model. In the offline stage, store the sensing snapshot of the WIFI signal strength in the environment to form a WIFI strength template.

[0008] Step 3) In the online phase, the newly input WIFI strength is matched with the WIFI strength in the offline phase. The active factors of the pose cell network are activated and corrected. The combination of the two can largely prevent the occurrence of mismatch and generate a more accurate experience map.

[0009] Preferably, in step 1), the dynamics of the attractor competition network always manipulate the activities inside the pose sensing cell network. Its internal dynamic process can be divided into excitation update, global inhibition of pose sensing cells, and normalization of pose sensing cell activity.

[0010] Preferably, in step 2), the WIFI fingerprint acquires relevant environmental information and interacts with the pose sensing cells. By establishing a connection between the WIFI strength fingerprint and the pose sensing cells, the energy in the activated WIFI fingerprint is injected into the pose sensing cells. The WIFI fingerprint is then connected to the available robot position, and the updated connection strength... It can then be expressed as

[0011]

[0012] In the formula: R i Q represents the signal strength at a specific location in a Wi-Fi fingerprint. xyθ To sense the activity level of cells based on their position.

[0013] Preferably, in step 2), a location area map is drawn, a series of test reference points are planned, and the RSSI values ​​of each reference point from different APs are measured sequentially by a WIFI strength receiving device. That is, the average value of multiple measurements is taken as the WIFI signal strength of the AP at that reference point, and recorded in a certain format in the location fingerprint database, which is also called the location fingerprint map.

[0014] Preferably, in step 3), each experience has an activity level, determined by the proximity of the energy peak in the pose-aware cell and the WIFI fingerprint to each experience-related unit. Each experience has a corresponding active region in the pose-aware cell and the WIFI fingerprint. When the energy peak is within these active regions, the stimulus is immediately activated. These regions are continuous within the pose-aware cell, but discontinuous in the corresponding regions of the WIFI fingerprint. Each experience e i From the experience of activity level E i WIFI signal strength R i The decision,

[0015] e i ={E i ,R i}

[0016] An experienced energy level Exyθ and the total energy level E of the i-th experience i The following formula can be used to calculate:

[0017]

[0018]

[0019] In the formula: x pc y pc and θ pc x represents the coordinates of the cell with the highest activity posture; i y i θ i The coordinates of the pose-aware cells associated with this experience; r a θ is the regional constant of the (x,y) plane; a R is the region constant in the θ dimension. curr R represents the current Wi-Fi signal strength. i The Wi-Fi signal strength associated with experience i.

[0020] Preferably, in step 3), the online phase uses a Bayesian algorithm to match WIFI fingerprint information, which corrects the activation level of pose cells and generates an experience map. The Bayesian algorithm online positioning is divided into two phases: the selection of access point (AP) adopts the mutual information minimization strategy and the Bayesian algorithm is used for location estimation.

[0021] The advantages of this invention are as follows: It employs a SLAM method based on a rodent model and WIFI fingerprints, utilizing a fingerprint recognition method based on WIFI signal strength. This method replaces the local scene cell network in the original rodent model with WIFI fingerprint information. An offline location fingerprint database is established, and in the online localization stage, a Bayesian algorithm is used to match the WIFI signal strength fingerprint information, correcting the activity of the pose cell network, and ultimately obtaining a more accurate experience map. This not only improves the localization accuracy of the mobile robot but also enhances the system's stability and provides excellent localization performance.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0023] Figure 1 This is a structural diagram of the rodent model system in the SLAM method based on rodent models and WIFI fingerprints of the present invention.

[0024] Figure 2 This is a schematic diagram of the WIFI localization principle in a SLAM method based on a rodent model and WIFI fingerprints according to the present invention.

[0025] Figure 3 This is a structural diagram of a WIFI-based rodent model, which is part of the SLAM method based on a rodent model and WIFI fingerprints according to the present invention. Detailed Implementation

[0026] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0027] like Figure 1 As shown, in the rodent model, visual scene cells acquire local scenes through visual perception. Furthermore, the model merges head orientation cells and position cells to form a new cell type called pose cells. The robot's pose information is encoded in the individual competitive attractor network (CAN) modules of the pose-aware cells. After processing by each pose-aware cell, the robot's egocentric and non-centric information stimuli influence the activity of the pose cells. The pose cells and visual cells generate a pose-visual scene, which in turn produces an experience map.

[0028] like Figure 2 As shown: The WIFI positioning method is a location fingerprinting method. Similar to fingerprint recognition in the conventional sense, location fingerprinting mainly relies on a database representing target features for identification. RSSI-based location fingerprinting consists of two stages: an offline acquisition stage and an online positioning stage.

[0029] The main task of the offline acquisition phase is to collect feature information of each reference point in the positioning area and establish a location fingerprint database. First, a map of the positioning area is drawn, a series of test reference points are planned, and the RSSI values ​​of each reference point from different APs are measured sequentially (multiple measurements are taken and the average value is taken) as the signal feature number of the AP at that reference point. The data is then recorded in the location fingerprint database in a certain format. This database is also called the location fingerprint map.

[0030] The method for establishing a WIFI fingerprint map in an indoor environment is as follows: Reference points are selected in the positioning environment according to certain rules, and the signal strength of the access point (AP) is continuously sampled at each reference point for a period of time to obtain the average value for each AP. Stored in the database, forming a location fingerprint (IM):

[0031]

[0032] in, L i =(x i ,y i () represents the position of the reference point, and k is the number of reference points. A represents the set of positions of all reference points; A = {AP1, AP2, ..., AP...} R} represents the set of all observed APs on the map:

[0033]

[0034] To locate the set of all means in a fingerprint, where Let be the mean of the j-th AP at the reference point Li, and MAC be the mean of the j-th AP. i This represents the MAC address value of the i-th reference point.

[0035] The online location phase consists of the following two steps:

[0036] Step 1: AP Selection Strategy

[0037] (1) Assuming there are T available reference points (APs) in the indoor positioning environment, selecting an optimized subset of S APs can reduce the signal space dimension from T to S, thus reducing the computational load. For each of the selected S APs, pairwise combinations are performed, and the mutual information of each combination is calculated using the following formula. The combination with the minimum mutual information is then identified, along with the corresponding APs. m AP n AP serves as the two initial reference points;

[0038] MI(AP m AP n )=H(AP m )+H(AP n )-H(AP m AP n )

[0039] In the formula: MI(AP) m AP n H(AP) represents the mutual information between two different APs; m AP n ) represents the combined information entropy of two APs.

[0040] (2) Calculate the mutual information of the remaining S-2 APs and the two initial APs respectively according to the following formula.

[0041] MI(AP m AP n AP i ) = H(AP m AP n )+H(AP i )-H(AP m AP n AP i )

[0042] Find the AP that minimizes MI and use it as the 3rd AP in the optimal AP subset.

[0043] (3) Following the same procedure as in step (2), select the next optimal AP, iterating until S optimal APs are selected. The formula for selecting the Rth optimal AP is:

[0044] MI(AP1,AP2,…,AP R )=H(AP1,AP2,…,AP R-1 )+

[0045] H(AP R )-H(AP m AP n AP R )

[0046] Step 2: Bayesian Location Estimation Strategy

[0047] To further optimize the AP selection strategy that minimizes mutual information, Bayesian posterior estimation is used, which greatly improves the location estimation accuracy and reliability of the WIFI fingerprint positioning algorithm.

[0048] The basic principle of Bayesian posterior estimation is

[0049]

[0050] In the formula: RSSI represents the RSSI observations of multiple APs at the location estimation point; p(L i |RSSI) indicates position L i The conditional probability of the location point appearing in L given the RSSI, i.e., given the observed RSSI vector. i The probability of p(RSSI|L) i ) indicates position L i The probability of p(L) i ) indicates position L i The probability of RSSI is usually assumed to be equal, without considering the differences between fingerprint points; p(RSSI) represents the total probability of RSSI occurring, and its formula is:

[0051]

[0052] C(RSSI1,RSSI2,…,RSSIM) represents the number of specified RSSI vectors observed at the fingerprint point; K represents the number of fingerprint point observation epochs.

[0053] Substituting the total probability formula back into the Bayesian posterior estimation formula, the posterior conditional probability is calculated. Using a Bayesian weighted location estimation formula with multiple fingerprint points can calculate the location of the estimated point in a relatively short time. Let the location of the estimated point be p.

[0054]

[0055] In the formula: (x,y) represents the two-dimensional coordinates of the estimated position point; (x i ,y i ) represents the coordinates of the i-th fingerprint point; ω i The weighted weight representing the i-th fingerprint point is the probability of the Bayesian posterior condition; K represents the number of neighboring points.

[0056] like Figure 3 As shown, Wi-Fi is used as a sensor in a rodent model. The localization model consists of three main parts: a Wi-Fi fingerprint, a pose cell network, and an experience map. The Wi-Fi fingerprint acquires the Wi-Fi signal strength of the environment and is called a Wi-Fi signal strength template. This fingerprint information is used to identify familiar environments. When newly input Wi-Fi signal strength information is matched against an existing Wi-Fi signal strength template using the algorithm described above, the activity factors of the pose cell network are activated. The combination of these two factors can largely prevent false matches and generate a more accurate experience map.

[0057] As is known from common technical knowledge, this invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. Therefore, the disclosed embodiments described above are merely illustrative in all respects and are not the only ones. All modifications within the scope of this invention or its equivalents are included in this invention.

Claims

1. A SLAM method based on a rodent model and WIFI fingerprint, characterized in that, The method includes the following steps: Step 1) Using a competitive attractor network, head orientation cells and position cells are modeled as pose sensing cells. After processing by the pose cells, the robot's self-centered and non-centered information stimuli affect the activity of the pose cells. Step 2) Use the wireless signal network WIFI as a sensor in the rodent model. In the offline stage, store the sensing snapshot of the WIFI signal strength in the environment to form a WIFI strength template. Step 3) In the online phase, the newly input WIFI strength is matched with the WIFI strength in the offline phase. The active factors of the pose cell network are activated and corrected. The combination of the two can largely prevent the occurrence of mismatch and generate a more accurate experience map.

2. The SLAM method based on a rodent model and WIFI fingerprint as described in claim 1, characterized in that: In step 1), the dynamics of the attractor competition network constantly manipulate the activities within the pose-sensing cell network. Its internal dynamic processes can be divided into excitation update, global inhibition of pose-sensing cells, and normalization of pose-sensing cell activity.

3. The SLAM method based on a rodent model and WIFI fingerprint as described in claim 2, characterized in that: In step 2), the WIFI fingerprint acquires relevant environmental information and interacts with the pose sensing cells. By establishing a connection between the WIFI strength fingerprint and the pose sensing cells, the energy from the activated WIFI fingerprint is injected into the pose sensing cells. The WIFI fingerprint is then connected to the available robot positions, and the updated connection strength... It can then be expressed as In the formula: R i Q represents the signal strength at a specific location in a Wi-Fi fingerprint. xyθ To sense the activity level of cells based on their position.

4. The SLAM method based on a rodent model and WIFI fingerprint as described in claim 3, characterized in that: In step 2), a location area map is drawn, a series of test reference points are planned, and the RSSI values ​​of each reference point from different APs are measured sequentially by a WIFI strength receiving device. That is, the average value of multiple measurements is taken as the WIFI signal strength of the AP at that reference point, and recorded in a certain format in the location fingerprint database, which is also called the location fingerprint map.

5. The SLAM method based on a rodent model and WIFI fingerprint as described in claim 4, characterized in that: In step 3), each experience has an activity level, determined by the proximity of the energy peak in the pose-aware cell and the WIFI fingerprint to each experience-related unit. Each experience has a corresponding active region in the pose-aware cell and the WIFI fingerprint. When the energy peak is within these active regions, the stimulus is immediately activated. These regions are continuous within the pose-aware cell, but discontinuous in the corresponding regions of the WIFI fingerprint. Each experience e i From the experience of activity level E i WIFI signal strength R i The decision, And i ={And i ,R i } An experienced energy level E xyθ and the total energy level E of the i-th experience i The following formula can be used to calculate: In the formula: x pc y pc and θ pc x represents the coordinates of the cell with the highest activity posture; i y i θ i The coordinates of the pose-aware cells associated with this experience; r a θ is the regional constant of the (x,y) plane; a R is the region constant in the θ dimension. curr R represents the current Wi-Fi signal strength. i The Wi-Fi signal strength associated with experience i.

6. The SLAM method based on a rodent model and WIFI fingerprint as described in claim 4, characterized in that: In step 3), the online phase uses a Bayesian algorithm to match WIFI fingerprint information, which corrects the activation level of pose cells and generates an experience map. The Bayesian algorithm online positioning is divided into two phases: the selection of access points (APs) adopts the mutual information minimization strategy and the Bayesian algorithm is used for location estimation.

Images