An indoor pedestrian flow evacuation control method, system, electronic device and medium

Abstract

The application discloses an indoor pedestrian flow evacuation control method and system, electronic equipment and medium, and relates to the field of pedestrian evacuation. The method first constructs a state equation and a measurement equation of an exit pedestrian density, a BP neural network for identifying a state function and an observation function, and performs offline iterative training. The state function and the observation function are solved according to the weight values of each layer in the BP neural network. The pedestrian flow evacuation density is predicted according to the solved function, and an error online neural network for correcting the predicted value is constructed. When there is an abnormal data state, the sum of the pedestrian flow evacuation density prediction value and the prediction error value is calculated as a pedestrian flow evacuation density prediction improvement value, and indoor pedestrian flow evacuation control is performed according to the improvement value. The method can ensure the accuracy of the pedestrian flow evacuation density prediction under the condition that the sensor performance is reduced or damaged, thereby improving the pedestrian flow evacuation efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pedestrian evacuation, in particular to an indoor pedestrian flow evacuation control method and system, an electronic device and a medium. BACKGROUND

[0002] With the continuous development of society, the public places such as entertainment facilities and large shopping malls in cities have increased dramatically. The above-mentioned public places often have a high concentration of people. After disasters such as fire and earthquake, stampede events are likely to occur. If the personnel can be guided to evacuate effectively in the event of an emergency, the casualties will be greatly reduced. Therefore, crowd evacuation has become a hot research topic.

[0003] Compared with the current research status at home and abroad, path finding algorithm, cellular automaton model and reinforcement learning control strategy are widely used in the research of pedestrian evacuation. However, the premise of excellent performance of the above-mentioned algorithm in the process of crowd evacuation is the accurate measurement of sensors (such as cameras). If there are uncontrollable factors such as smoke around the sensor at this time, it may cause the sensor to measure the distribution of pedestrians inaccurately, thereby affecting the working efficiency of the evacuation algorithm. In actual application, the reduction of the efficiency of the evacuation algorithm may cause irreparable loss of life and property. SUMMARY

[0004] The purpose of the present application is to provide an indoor pedestrian flow evacuation control method, system, electronic device and medium to improve the efficiency of pedestrian flow evacuation under the condition of sensor performance degradation or damage.

[0005] To achieve the above-mentioned purpose, the present application provides the following scheme:

[0006] In one aspect, the present application provides an indoor pedestrian flow evacuation control method, comprising:

[0007] Modeling the evolution rule of pedestrian density in an indoor evacuation scene, constructing a state equation and a measurement equation of exit pedestrian density;

[0008] Constructing a BP neural network for identifying state functions and observation functions in the state equation and the measurement equation; the BP neural network comprises an input layer, a hidden layer and an output layer;

[0009] Offline iterative training of the BP neural network, and obtaining the weight values of each layer in the BP neural network after training;

[0010] Solving the state functions and observation functions in the state equation and the measurement equation according to the weight values of each layer in the BP neural network;

[0011] According to the state function and the observation function, a pedestrian flow evacuation density prediction value is obtained.

[0012] An error online neural network for correcting the pedestrian flow evacuation density prediction value is constructed.

[0013] According to the state of the sensor in the indoor evacuation scene, it is determined whether there is a data state anomaly.

[0014] If there is no data state anomaly, indoor pedestrian flow evacuation control is directly performed according to the pedestrian flow evacuation density prediction value, and the error online neural network is continuously trained.

[0015] If there is a data state anomaly, a prediction error value is calculated according to the currently trained error online neural network, and a sum of the pedestrian flow evacuation density prediction value and the prediction error value is calculated as a pedestrian flow evacuation density prediction improvement value.

[0016] According to the pedestrian flow evacuation density prediction improvement value, indoor pedestrian flow evacuation control is performed.

[0017] Optionally, the pedestrian density evolution rule in the indoor evacuation scene is modeled, and a state equation and a measurement equation of the exit pedestrian density are constructed, specifically including:

[0018] The pedestrian density evolution rule in the indoor evacuation scene is modeled, and a state equation X(k) = f(X(k-1)) + q(k-1) and a measurement equation Z(k) = h(X(k)) + r(k) of the exit pedestrian density are constructed; wherein the indoor evacuation scene has multiple exits, and each exit is equipped with an evacuation guide and a sensor; X(k) ∈ R n×n and X(k-1) represent state variables at k time and k-1 time, respectively; Z(k) ∈ R m×n represents an observation variable at k time; f(·) represents a nonlinear state function; h(·) represents a nonlinear observation function; q(k-1) ∈ R n×n represents process noise; r(k) ∈ R m×n represents measurement noise.

[0019] Optionally, the BP neural network for identifying the state function and the observation function in the state equation and the measurement equation is constructed, specifically including:

[0020] A BP neural network for identifying the nonlinear state function f(·) in the state equation is constructed, wherein the BP neural network takes X(k) as input, Z(k) as output, a Sigmoid function as an activation function, and a function expression is Z(k) = h(X(k)) = W2 T sigmoid(W1 TX(k)) ; wherein W1 is the weight value from the input layer to the hidden layer, and W2 is the weight value from the hidden layer to the output layer;

[0021] a BP neural network for identifying a nonlinear observation function h(·) in the measurement equation is constructed, the BP neural network taking X(k-1) as input, X(k) as output, and a Sigmoid function as an activation function, and the function expression is X(k) = f(X(k-1)) = W2 T sigmoid(W1 T X(k-1)).

[0022] Optionally, the BP neural network is iteratively trained offline, and the weight values of each layer in the BP neural network are obtained after the training is completed, and the training specifically comprises:

[0023] The sample data for training the BP neural network are obtained based on a pedestrian evacuation simulation experiment, the sample data are divided into a training set and a validation set in a ratio of 7:3, the SGD is used as an optimizer to update the weight values, the MSE is used as a loss function, a learning rate and a training number are set to perform iterative training, and the weight values of each layer in the BP neural network are obtained after the training is completed.

[0024] Optionally, the error online neural network for correcting the pedestrian flow evacuation density prediction value is constructed, and the construction specifically comprises:

[0025] The error online neural network for correcting the pedestrian flow evacuation density prediction value is constructed based on the BP neural network; the input of the error online neural network is the pedestrian flow evacuation density prediction value, and the output is an error between the pedestrian flow evacuation density prediction value and a real value of the pedestrian flow evacuation density, i.e., a prediction error value.

[0026] Optionally, whether there is a data state abnormality is determined according to a sensor state in the indoor evacuation scene, and the determination specifically comprises:

[0027] If the sensor in the indoor evacuation scene has a performance decline or a damaged state, it is determined that there is a data state abnormality.

[0028] If the sensor in the indoor evacuation scene does not have a performance decline or a damaged state, it is determined that there is no data state abnormality.

[0029] Optionally, indoor pedestrian flow evacuation control is performed according to the improved pedestrian flow evacuation density prediction value, and the control specifically comprises:

[0030] The improved pedestrian flow evacuation density prediction value of the i th exit at the k th moment is substituted into the formula i (k) to perform indoor pedestrian flow evacuation control; wherein ​u (k) represents the target density of the pedestrian flow at the i th exit at the k th moment; u i (k) represents the guidance coefficient of the i th exit at the k th moment; when u i (k) is 1, the guidance of opening the i th exit, when u i (k) is 0, the guidance of closing the i th exit.

[0031] In another aspect, the present application also provides an indoor pedestrian flow evacuation control system, comprising:

[0032] A pedestrian density equation modeling module is configured to model the pedestrian density evolution rule in an indoor evacuation scenario, and to construct a state equation and a measurement equation of the exit pedestrian density;

[0033] A BP neural network construction module is configured to construct a BP neural network for identifying state functions and observation functions in the state equation and the measurement equation; the BP neural network comprises an input layer, a hidden layer and an output layer;

[0034] An offline network iterative training module is configured to perform offline iterative training on the BP neural network, and to obtain the weight values of each layer in the BP neural network after the training is completed;

[0035] A state and observation function solving module is configured to solve the state functions and the observation functions in the state equation and the measurement equation according to the weight values of each layer in the BP neural network;

[0036] A pedestrian flow evacuation density prediction module is configured to perform prediction of the pedestrian flow evacuation density according to the solved state functions and observation functions, and to obtain a pedestrian flow evacuation density prediction value;

[0037] An error online neural network construction module is configured to construct an error online neural network for correcting the pedestrian flow evacuation density prediction value;

[0038] A data state abnormality judgment module is configured to judge whether there is a data state abnormality according to the sensor state in the indoor evacuation scenario;

[0039] An error online neural network training module is configured to, if there is no data state abnormality, directly perform indoor pedestrian flow evacuation control according to the pedestrian flow evacuation density prediction value, and continuously train the error online neural network;

[0040] A density prediction improvement value calculation module is configured to, if there is a data state abnormality, calculate a prediction error value according to the currently trained error online neural network, and calculate the sum of the pedestrian flow evacuation density prediction value and the prediction error value as a pedestrian flow evacuation density prediction improvement value;

[0041] The indoor pedestrian flow evacuation control module is used for indoor pedestrian flow evacuation control according to the pedestrian flow evacuation density prediction improved value.

[0042] In another aspect, the present application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the indoor pedestrian flow evacuation control method when executing the computer program.

[0043] In another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executable to implement the indoor pedestrian flow evacuation control method.

[0044] According to the specific embodiments of the present application, the following technical effects are achieved.

[0045] The present application provides an indoor pedestrian flow evacuation control method, system, electronic device and medium, the method comprising: modeling the pedestrian density evolution rule in the indoor evacuation scene, constructing the state equation and measurement equation of the exit pedestrian density; constructing a BP neural network for identifying the state function and observation function in the state equation and measurement equation; the BP neural network comprises an input layer, a hidden layer and an output layer; the BP neural network is iteratively trained offline, and the weight values of each layer in the BP neural network are obtained after training; the state function and observation function in the state equation and measurement equation are solved according to the weight values of each layer in the BP neural network; the pedestrian flow evacuation density is predicted according to the solved state function and observation function, and a pedestrian flow evacuation density prediction value is obtained; an error online neural network is constructed for correcting the pedestrian flow evacuation density prediction value; it is judged whether there is data state abnormality according to the sensor state in the indoor evacuation scene; if there is no data state abnormality, indoor pedestrian flow evacuation control is directly performed according to the pedestrian flow evacuation density prediction value, and the error online neural network is continuously trained; if there is data state abnormality, a prediction error value is calculated according to the currently trained error online neural network, and the sum of the pedestrian flow evacuation density prediction value and the prediction error value is calculated as a pedestrian flow evacuation density prediction improved value; indoor pedestrian flow evacuation control is performed according to the pedestrian flow evacuation density prediction improved value. The method of the present application can still ensure the accuracy of pedestrian flow evacuation density prediction under the condition of sensor performance degradation or damage, thereby improving the pedestrian flow evacuation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0047] Figure 1 Flow chart of the indoor pedestrian flow evacuation control method of the present application;

[0048] Figure 2 Technical roadmap of the indoor pedestrian flow evacuation control method of the present application;

[0049] Figure 3 Indoor evacuation scene schematic diagram provided by the embodiments of the present application;

[0050] Figure 4 BP neural network structure schematic diagram of nonlinear observation function identification provided by the embodiments of the present application;

[0051] Figure 5 BP neural network structure schematic diagram of nonlinear state function identification provided by the embodiments of the present application;

[0052] Figure 6 Structure schematic diagram of error online neural network provided by the embodiments of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] The purpose of the present application is to provide an indoor pedestrian flow evacuation control method, system, electronic device and medium, which realizes the prediction of the pedestrian flow evacuation density in the field of view of the sensor (such as a camera) on the basis of the simulation of the pedestrian flow evacuation density control, so as to solve the problem that the pedestrian flow evacuation density control cannot work normally after the performance of the sensor is degraded or even damaged (such as the camera is damaged due to fire) during the evacuation process, and improve the pedestrian flow evacuation efficiency under the condition that the performance of the sensor is degraded or damaged.

[0055] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0056] Figure 1A flow chart of an indoor pedestrian flow evacuation control method of the present application; Figure 2 A technical roadmap of an indoor pedestrian flow evacuation control method of the present application. Referring to Figure 1 and Figure 2 The indoor pedestrian flow evacuation control method of the present application comprises the following steps:

[0057] Step 1: modeling the pedestrian density evolution rule under the indoor evacuation scenario, constructing the state equation and measurement equation of the exit pedestrian density.

[0058] Figure 3 An indoor evacuation scenario schematic diagram provided by an embodiment of the present application. Referring to Figure 3 The evacuation scenario targeted by the present application is an indoor evacuation scenario, such as a large shopping mall, an amusement park, a library, a museum, etc., wherein the indoor evacuation scenario has multiple exits, and each exit is equipped with an evacuation guide and a sensor; the sensor is usually a camera. The indoor evacuation scenario is subjected to grid processing, and a space comprising N×N grids can be formed. As shown in the embodiment Figure 3 has 4 exits, and each exit is equipped with an evacuation guide. Through the camera located at the exit, the evacuation guide can obtain the pedestrian density of the exit and adjust the guide signal (open or close), thereby realizing the control of the entire pedestrian flow evacuation guide process.

[0059] The present application models the pedestrian density evolution rule under the indoor evacuation scenario, and constructs the state equation and measurement equation of the exit pedestrian density as follows.

[0060] The state equation of the exit pedestrian density is:

[0061] X(k) = f(X(k-1)) + q(k-1) (1)

[0062] The measurement equation of the exit pedestrian density is:

[0063] Z(k) = h(X(k)) + r(k) (2)

[0064] wherein X(k) ∈ R n×n and X(k-1) ∈ R n×n represent the state variables at time k and time k-1, respectively; Z(k) ∈ R m×n represents the observation variable at time k; f(·) represents a nonlinear state function; h(·) represents a nonlinear observation function; q(k-1) ∈ R n×n represents the process noise; r(k) ∈ R m×n represents the measurement noise.

[0065] wherein r(k) and q(k-1) are independent of each other, have a mean of 0 and satisfy the following equation:

[0066] E[r(k)] = 0, E[q(k-1)] = 0 (3)

[0067] E[r(k)r T (k)] = R(k), E[q(k-1)q T (k-1)] = Q(k-1) (4)

[0068] wherein E[.] denotes expectation; R(k) and Q(k-1) denote the covariance matrix of r(k) and q(k-1) respectively, and both are Gaussian white noise. The value of Q(k-1) is The value of R(k) is 0.1.

[0069] In the embodiment shown in Figure 3 In the embodiment shown in

[0070]

[0071] Z(k) = (the fourth sensor detects the flow density) (6)

[0072] Since the state function f(·) and the observation function h(·) in the above equation cannot be directly solved, the BP neural network is used for identification in the present application, and then the nonlinear state function f(·) and the nonlinear observation function h(·) are obtained.

[0073] Step 2: Construct the BP neural network for identifying the state function and the observation function in the state equation and the measurement equation.

[0074] For the embodiment shown in Figure 3 In the state equation (1) and the measurement equation (2) constructed in step 1, the state variable X(k-1) at k-1 time is R 3×1 , the state variable X(k) at k time is R 3×1 , and the observation variable Z(k) at k time is R 1×1 Since the input and output of the two functions f(·) and h(·) are different, two BP neural networks are respectively constructed for identifying the nonlinear state function f(·) and the nonlinear observation function h(·). Both of the two BP neural networks include an input layer, an intermediate layer (usually a hidden layer) and an output layer.

[0075] The BP neural network structure for identifying the nonlinear observation function h(·) is as shown in Figure 4As shown, dense_1_input:ImputLayer, dense_1:Dense, and dense_2:Dense represent the input layer, intermediate layer, and output layer of the BP neural network, respectively; input and output represent the input and output of a certain layer of the BP neural network, respectively. Figure 4 In this model, the pedestrian flow density at the 1st, 2nd, and 3rd gates at time k is used as input, so the input layer is 3-dimensional. "None" represents the number of samples required for input, thus the input layer can be written as (None, 3). The input layer transforms (None, 3) into (None, 7) based on the connection weights between the input layer and the intermediate layers. Similarly, the intermediate layers obtain the output layer's output (None, 1), which is the pedestrian flow density at the 4th gate at time k, through the connection weights. In other words, the BP neural network used for identifying the nonlinear observation function h(·) in this invention uses X(k) as input, Z(k) as output, and the Sigmoid function as the activation function of the BP neural network. Therefore, the function expression of this BP neural network is:

[0076] Z(k)=h(X(k))=W2 T sigmoid(W1 T X(k)) (7)

[0077] Where W1 represents the weights from the input layer to the hidden layer, and W2 represents the weights from the hidden layer to the output layer; the Sigmoid function nonlinearizes the linear weighted calculation results of each neuron in the neural network, giving the neural network nonlinear mapping capabilities. Its function expression is:

[0078] The BP neural network structure used for identifying nonlinear state functions f(·) is as follows: Figure 5 As shown, this BP neural network takes the pedestrian flow density at the 1st, 2nd, and 3rd gates at time k-1 as input and the pedestrian flow density at the 1st, 2nd, and 3rd gates at time k as output. Therefore... Figure 5 The BP neural network structure shown is similar to Figure 4 The difference lies in the output layer dense_2:Dense of this BP neural network, where the output dimension is 3-dimensional. That is, the input layer transforms (None, 3) into (None, 7) based on the connection weights between the input layer and the intermediate layers. Similarly, the intermediate layers obtain the output of the output layer as (None, 3) through the connection weights, representing the flow density of the 1st, 2nd, and 3rd gates at time k. In other words, the BP neural network used in this invention for identifying the nonlinear state function f(·) uses X(k-1) as input, X(k) as output, and the Sigmoid function as the activation function of the BP neural network. Therefore, the function expression of this BP neural network is:

[0079] X(k)=f(X(k-1))=W2T sigmoid(W1 T X(k-1)) (8)

[0080] where W1 is the weight value from input layer to hidden layer, and W2 is the weight value from hidden layer to output layer.

[0081] Step 3: offline iterative training is performed on the BP neural network, and the weight values of each layer in the BP neural network are obtained after the training is completed.

[0082] On the basis of the BP neural network constructed in step 2, sample data used for training the BP neural network is obtained based on a pedestrian evacuation simulation experiment, the sample data is divided into a training set and a validation set in a ratio of 7:3, SGD is used as an optimizer to update the weight values, MSE is used as a loss function, a learning rate and a training number are set for iterative training, and the weight values W1 and W2 of each layer in the BP neural network are obtained after the training is completed.

[0083] Specifically, a pedestrian evacuation simulation experiment based on an existing indoor multi-outlet pedestrian flow simulation method is performed, the pedestrian evacuation simulation experiment records the pedestrian flow density data measured by four doors at each time during the pedestrian evacuation process, 1000 simulation results are randomly selected and saved, and sample data used for training the BP neural network of the application is obtained. The sample data is divided into a training set and a validation set in a ratio of 7:3, SGD is used as an optimizer to update the weight values, MSE is used as a loss function, a learning rate and a training number are set for iterative training, and the weight values of each layer are obtained after the training of the BP neural network is completed.

[0084] where the stochastic gradient descent strategy (SGD) adjusts the weight values of the BP neural network in the negative gradient direction of the target gradient, and for a certain output node:

[0085]

[0086] where η represents a learning rate and is 0.01; E k is a mean square error MSE; XW ij represents the connection weight value between the i-th layer and the j-th layer of the BP neural network, and ΔW ij is the update amount of the connection weight value XW ij during the training process.

[0087] The connection weight value and the threshold value between the hidden layer and the input layer need to be adjusted as follows:

[0088]

[0089] The calculated connection weight value XW ij is a scalar; W 12W1 is the connection weight value between the input layer and the intermediate layer, and is taken as W1 in formula (7) and (8) after the training is completed. 23 W2 is the connection weight value between the intermediate layer and the output layer, and is taken as W2 in formula (7) and (8) after the training is completed. Both W1 and W2 are in the form of a matrix.

[0090] In addition, in order to make the actual error and the prediction error of the network as small as possible, it is necessary to iteratively adjust the weight values of each layer of the BP neural network according to the stochastic gradient descent strategy, so that the value of the loss function is continuously reduced.

[0091] Step 4: Solve the state function and the observation function in the state equation and the measurement equation according to the weight values of each layer in the BP neural network.

[0092] Substitute the trained weight values W1 and W2 of each layer into the transfer function of the corresponding BP neural network , that is, formula (7) and (8) for the state function and the observation function respectively, so as to obtain the nonlinear state function f(·) and the nonlinear observation function h(·). In the formula, W1 is the connection weight value from the input layer to the hidden layer, W2 is the connection weight value from the hidden layer to the output layer, input is the input of the BP neural network, and output is the output of the BP neural network.

[0093] After f(·) and h(·) are obtained, the real-time prediction of the pedestrian flow density of each exit at each time can be performed according to formula (1) and formula (2), and the predicted value of the pedestrian flow density of the corresponding exit is obtained.

[0094] Step 5: Perform the prediction of the pedestrian flow density according to the solved state function and the observation function, and obtain the predicted value of the pedestrian flow density.

[0095] The prediction of the pedestrian flow density is performed according to the solved nonlinear state function f(·) and the nonlinear observation function h(·), and the prediction process is as follows:

[0096] 1) Calculate the prior estimate X(k|k-1); specifically, substitute the state variable X(k-1) at time k-1 into formula (11) to calculate the prior estimate X(k|k-1) of the state variable X(k) at time k:

[0097] X(k|k-1)=f(X(k-1)) (11)

[0098] 2) Predict the predicted value Z4(k) of the pedestrian flow density of the fourth exit at time k according to X(k|k-1):

[0099] Z4(k)=H(k)X(k|k-1) (12)

[0100] where H(k) is the Jacobian matrix of h(X(k|k-1)).

[0101] 3) Calculate the covariance matrix P(k|k-1):

[0102] P(k|k-1) = F(k|k-1)P(k-1)F T (k,k-1) + Q(k-1) (13)

[0103] where F(k|k-1) is the Jacobian matrix of f(X(k-1)); P(k-1) represents the covariance matrix at time k-1.

[0104] 4) Calculate the gain K(k) for reflecting the confidence of the predicted value and the observed value:

[0105] K(k) = P(k|k-1)H T (k) [H(k)P(k|k-1)H T (k) + R(k)] -1 (14)

[0106] 5) Combine the gain K(k) to calculate the posterior estimate X(k), that is, to correct X(k|k-1):

[0107] X(k) = X(k|k-1) + K(k)[Z(k) - h(X(k|k-1))] (15)

[0108] 6) Calculate the covariance matrix P(k):

[0109] P(k) = [I - K(k)H(k)]P(k|k-1) (16)

[0110] where I is the unit matrix; F(k|k-1) is the Jacobian matrix of f(X(k-1)); H(k) is the Jacobian matrix of h(X(k|k-1)). According to formulas (11) to (16), the pedestrian flow density at each time can be recursively predicted, for example, the predicted value of the pedestrian flow density Z4(k) of the fourth exit at time k and the predicted value of the pedestrian flow density X(k) of the first, second, and third exits at time k.

[0111] Step 6: Construct an error online neural network for correcting the predicted value of the pedestrian flow density.

[0112] Step 5 can recursively predict the flow density of each exit at each time, but the prediction error may be amplified over time as the recursion continues over time. On the other hand, if the sensor in the indoor evacuation scene has a performance decline or damage state, the flow density predicted according to the sensor data will have a significant error. To solve this problem, the present application constructs an error online neural network for correcting the predicted value of the pedestrian flow evacuation density based on a BP neural network. The network structure diagram of the error online neural network for correcting the predicted value is shown in Figure 6 . Figure 6 In the dense_1_input:ImputLayer, dense_1:Dense and dense_2:Dense, the input layer, the intermediate layer and the output layer of the BP neural network are respectively represented; the input and the output represent the input and the output of a certain layer of the BP neural network. Figure 6 In the dense_1_input:ImputLayer, dense_1:Dense and dense_2:Dense, the input layer, the intermediate layer and the output layer of the BP neural network are respectively represented; the input and the output represent the input and the output of a certain layer of the BP neural network. 真 In the dense_1_input:ImputLayer, dense_1:Dense and dense_2:Dense, the input layer, the intermediate layer and the output layer of the BP neural network are respectively represented; the input and the output represent the input and the output of a certain layer of the BP neural network. 真 (k) between the predicted value of the pedestrian flow evacuation density and the real value of the pedestrian flow evacuation density, i.e. the prediction error value ΔZ(k). The real value of the pedestrian flow evacuation density Z 真 (k) at the k time is the flow density data of the fourth door at the k time recorded in the pedestrian evacuation process when the pedestrian evacuation simulation experiment is running. When there is no data state abnormality, Z 真 (k) = Z i (k) = Z i (k) = Z

[0114] Step 7: judging whether there is a data state abnormality according to the state of the sensor in the indoor evacuation scene.

[0115] Existing indoor multi-exit pedestrian flow simulation methods require high accuracy in the pedestrian density information acquired by sensors to function properly. However, in real-world evacuation scenarios, such as emergency fire evacuations, sensor performance may degrade due to smoke obstruction or be damaged by fire. In these situations, cameras collecting pedestrian evacuation density data are prone to inaccurate data collection. These problems prevent existing methods from functioning continuously, thus affecting evacuation efficiency. To address this issue, this invention proposes performing predictive correction on the pedestrian density data acquired by cameras to improve data accuracy. Then, indoor pedestrian flow evacuation control is performed based on the corrected predicted pedestrian evacuation density value, thereby reducing the impact of data errors on the guidance and control system.

[0116] Specifically, if the sensors in the indoor evacuation scenario are in a state of performance degradation or damage, then it is determined that there is an abnormal data status; if the sensors in the indoor evacuation scenario are not in a state of performance degradation or damage, then it is determined that there is no abnormal data status.

[0117] Step 8: If there are no abnormal data conditions, directly control the indoor pedestrian flow evacuation based on the predicted pedestrian flow evacuation density, and continuously train the online error neural network.

[0118] If the data collected by the sensors is currently within the normal range, the online error neural network is put into training mode and continuously trained. At this point, indoor pedestrian evacuation control can be directly performed based on the predicted pedestrian flow density. The control algorithm is as follows:

[0119]

[0120] However, if it is determined that there are no abnormal data conditions, the predicted pedestrian evacuation density Z at the i-th exit at time k can be used. i (k) (in this embodiment of the invention, Z4(k)) is used as ρ i (k), substitute into formula (17) for indoor pedestrian evacuation control. Where ρ i aim This represents the pedestrian evacuation target density at the i-th exit at time k in an indoor evacuation scenario; u i (k) represents the guiding effect coefficient of the i-th exit at time k; when u i (k) is 1 when the i-th exit is opened, and when u i When (k) is 0, the guidance function of the i-th exit is closed. That is, when the evacuation guidance personnel at the i-th exit obtain the pedestrian density ρ at time k, the guidance function of the exit is closed. i (k) is less than or equal to the set target density At that time, the guiding coefficient u of the corresponding gatei (k) will be adjusted to 1, thereby opening the guiding effect. Conversely, the guiding effect coefficient u i (k) will be adjusted to 0, thereby closing the guiding effect.

[0121] Step 9: If there is a data state anomaly, calculate the predicted error value according to the currently trained error online neural network, and calculate the sum of the pedestrian flow evacuation density prediction value and the predicted error value as the pedestrian flow evacuation density prediction improvement value.

[0122] If the current data collected by the sensor is in an abnormal state, the error online neural network enters the use state at this time, and the predicted error value ΔZ(k) is calculated according to the currently trained error online neural network. The improved result is obtained by combining the network prediction value and the predicted error value obtained by the error online neural network. That is, the sum of the pedestrian flow evacuation density prediction value Z4(k) and the predicted error value ΔZ(k) is calculated as the pedestrian flow evacuation density prediction improvement value Z 改 (k) according to the following formula (18).

[0123] Z 改 (k) = Z4(k) + ΔZ(k) (18)

[0124] Step 10: Perform indoor pedestrian flow evacuation control according to the pedestrian flow evacuation density prediction improvement value.

[0125] When it is judged that there is a data state anomaly, the pedestrian flow evacuation density prediction improvement value Z 改 (k) of the i th exit at time k is taken as ρ i (k), and indoor pedestrian flow evacuation control is performed according to the above formula (17). In the embodiment of the present application, when u4(k) is 1, the guiding effect of the fourth exit is opened, and when u4(k) is 0, the guiding effect of the fourth exit is closed. The results show that after the pedestrian density data obtained by the camera is corrected by the method of the present application, the effect of indoor pedestrian flow evacuation control is significantly improved.

[0126] It can be seen that the method of the application firstly models the pedestrian density evolution rule under the indoor evacuation scene, constructs the nonlinear state function and the nonlinear observation function expression of the exit pedestrian density; secondly, the parameters of the nonlinear state function and the nonlinear observation function are solved by using the BP neural network identification, so as to solve the problem that the function parameters cannot be directly solved by derivation; thirdly, the prediction value is optimized and adjusted in real time by the error online neural network, and the prediction accuracy is further improved; finally, the real-time prediction improvement value is used for pedestrian flow evacuation control, and the evacuation efficiency under the condition that the sensor performance is reduced or damaged is improved. The method of the application realizes the prediction of the pedestrian evacuation density in the field of view of the sensor (such as a camera) on the basis of the existing pedestrian flow evacuation density control simulation, thereby solving the problem that the pedestrian flow evacuation density control cannot work normally after the sensor performance is reduced or damaged (such as the camera is damaged due to fire) in the evacuation process.

[0127] Based on the method provided by the application, the application further provides an indoor pedestrian flow evacuation control system, comprising:

[0128] A pedestrian density equation modeling module is configured to model the pedestrian density evolution rule under the indoor evacuation scene, and construct the state equation and the measurement equation of the exit pedestrian density;

[0129] A BP neural network construction module is configured to construct a BP neural network for identifying the state function and the observation function in the state equation and the measurement equation; the BP neural network comprises an input layer, a hidden layer and an output layer;

[0130] An offline network iterative training module is configured to perform offline iterative training on the BP neural network, and obtain the weight values of each layer in the BP neural network after the training is completed;

[0131] A state and observation function solving module is configured to solve the state function and the observation function in the state equation and the measurement equation according to the weight values of each layer in the BP neural network;

[0132] A pedestrian flow evacuation density prediction module is configured to predict the pedestrian flow evacuation density according to the solved state function and observation function, and obtain a pedestrian flow evacuation density prediction value;

[0133] An error online neural network construction module is configured to construct an error online neural network for correcting the pedestrian flow evacuation density prediction value;

[0134] A data state abnormality judgment module is configured to judge whether there is a data state abnormality according to the sensor state in the indoor evacuation scene;

[0135] an error online neural network training module, configured to, if there is no data state abnormality, directly perform indoor pedestrian flow evacuation control according to the pedestrian flow evacuation density prediction value, and continuously train the error online neural network;

[0136] a density prediction improvement value calculation module, configured to, if there is data state abnormality, calculate a prediction error value according to the currently trained error online neural network, and calculate a sum of the pedestrian flow evacuation density prediction value and the prediction error value as a pedestrian flow evacuation density prediction improvement value;

[0137] an indoor pedestrian flow evacuation control module, configured to perform indoor pedestrian flow evacuation control according to the pedestrian flow evacuation density prediction improvement value.

[0138] Overall, the method and system of the present application use a modeling method to predict and correct the pedestrian density data obtained by the camera, thereby improving the accuracy of the data. Then, indoor pedestrian flow evacuation control is performed according to the corrected data, thereby reducing the influence of data errors on the guidance control system, and thereby improving the evacuation efficiency under the condition of performance degradation or damage of the sensor. The present application solves the problem that the pedestrian flow evacuation system model is complex and difficult to solve directly by using the method of neural network identification to identify the nonlinear state function and the nonlinear observation function training set. The present application also uses an offline method for system identification. Since less identification data is obtained in real time in the experiment, offline identification can make full use of historical data, and can effectively improve the identification training efficiency. The present application also uses an error online neural network to correct the prediction data. If only the model is continuously propagated, the error will be continuously amplified with the accumulation of time. To solve this problem, the present application uses a neural network to online correct the prediction value data output by the model, thereby improving the robustness of the method of the present application in practical application.

[0139] Further, the present application also provides an electronic device, which can include a processor, a communication interface, a memory and a communication bus. Wherein the processor, the communication interface and the memory complete the communication among each other through the communication bus. The processor can call the computer program in the memory to execute the indoor pedestrian flow evacuation control method.

[0140] In addition, the computer program in the memory described above is realized in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, and various media that can store program codes.

[0141] Further, the present application also provides a non-transitory computer readable storage medium, having stored thereon a computer program, which, when executed, can implement the indoor pedestrian flow evacuation control method.

[0142] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the system disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0143] The principles and implementation manners of the present application are described by using specific examples in the present application. The above embodiment description is only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will be changed. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for controlling the evacuation of pedestrians indoors, characterized in that, include: The evolution rule of pedestrian density in an indoor evacuation scenario is modeled, and the state equation X(k)=f(X(k-1))+q(k-1) and the measurement equation Z(k)=h(X(k))+r(k) are constructed. The indoor evacuation scenario has multiple exits, and each exit is equipped with evacuation guides and sensors. X(k) and X(k-1) represent the state variables at time k and time k-1, respectively. Z(k) represents the observation variable at time k. f(·) represents the nonlinear state function. h(·) represents the nonlinear observation function. q(k-1) represents the process noise. r(k) represents the measurement noise. A backpropagation (BP) neural network is constructed to identify the state function and observation function in the state equation and measurement equation; the BP neural network includes an input layer, a hidden layer, and an output layer; The construction of the BP neural network for identifying the state function and observation function in the state equation and measurement equation specifically includes: A backpropagation (BP) neural network is constructed to identify the nonlinear state function f(·) in the state equation. The BP neural network takes X(k) as input, Z(k) as output, and the sigmoid function as activation function, with the function expression Z(k)=h(X(k))=W2. T sigmoid(W1 T X(k)); where W1 is the weight from the input layer to the hidden layer, and W2 is the weight from the hidden layer to the output layer; A backpropagation (BP) neural network is constructed to identify the nonlinear observation function h(·) in the measurement equation. The BP neural network takes X(k-1) as input, X(k) as output, and the sigmoid function as activation function. The function expression is X(k)=f(X(k-1))=W2. T sigmoid(W1 T X(k-1)); The BP neural network is trained offline iteratively, and the weights of each layer in the BP neural network are obtained after the training is completed. Based on the weights of each layer in the BP neural network, the state function and observation function in the state equation and measurement equation are solved; Based on the solved state function and observation function, the pedestrian flow evacuation density is predicted, and the predicted value of the pedestrian flow evacuation density is obtained. An online error neural network is constructed to correct the predicted value of pedestrian evacuation density; the input of the online error neural network is the predicted value of pedestrian evacuation density, and the output is the error between the predicted value of pedestrian evacuation density and the actual value of pedestrian evacuation density, i.e., the prediction error value. Determine whether there are any abnormal data states based on the sensor status in the indoor evacuation scenario; If there are no abnormal data conditions, the indoor pedestrian evacuation control is directly carried out based on the predicted pedestrian evacuation density, and the online error neural network is continuously trained. If there is an abnormal data status, the prediction error value is calculated based on the currently trained online error neural network, and the sum of the predicted pedestrian flow evacuation density value and the prediction error value is calculated as the improved value of the pedestrian flow evacuation density prediction. Indoor pedestrian evacuation control is carried out based on the improved predicted value of pedestrian evacuation density.

2. The indoor pedestrian flow evacuation control method according to claim 1, characterized in that, The offline iterative training of the BP neural network, and the acquisition of the weights of each layer in the BP neural network after training, specifically includes: Sample data for training the BP neural network was obtained based on pedestrian evacuation simulation experiments. The sample data was divided into training and validation sets in a 7:3 ratio. SGD was used as the optimizer to update the weights, and MSE was used as the loss function. The learning rate and number of training iterations were set for iterative training. After training, the weights of each layer in the BP neural network were obtained.

3. The indoor pedestrian flow evacuation control method according to claim 2, characterized in that, The step of determining whether there is an abnormal data status based on the sensor status in the indoor evacuation scenario specifically includes: If the sensors in the indoor evacuation scenario exhibit performance degradation or damage, then an abnormal data status is determined. If the sensors in the indoor evacuation scenario do not exhibit performance degradation or damage, then it is determined that there are no abnormal data states.

4. The indoor pedestrian flow evacuation control method according to claim 3, characterized in that, The indoor pedestrian evacuation control based on the predicted improved value of pedestrian evacuation density specifically includes: The improved value ρ for predicting pedestrian evacuation density at the i-th exit at time k. i (k) Substitute into the formula Indoor pedestrian flow evacuation control; among which This represents the pedestrian evacuation target density at the i-th exit at time k in an indoor evacuation scenario; u i (k) represents the guiding effect coefficient of the i-th exit at time k; when u i (k) is 1 when the i-th exit is opened, and when u i (k) is 0, which is the guiding function to close the i-th exit.

5. An indoor pedestrian flow evacuation control system, characterized in that, The method for controlling the evacuation of pedestrians as described in any one of claims 1 to 4; the indoor pedestrian evacuation control system includes: The pedestrian density equation modeling module is used to model the evolution rules of pedestrian density in indoor evacuation scenarios and construct the state equation and measurement equation of pedestrian density at the exit. A BP neural network construction module is used to construct a BP neural network for identifying the state function and observation function in the state equation and measurement equation; the BP neural network includes an input layer, a hidden layer, and an output layer; The offline network iterative training module is used to perform offline iterative training on the BP neural network, and obtain the weights of each layer in the BP neural network after training. The state and observation function solving module is used to solve for the state function and observation function in the state equation and measurement equation based on the weights of each layer in the BP neural network. The pedestrian flow evacuation density prediction module is used to predict the pedestrian flow evacuation density based on the solved state function and observation function, and obtain the predicted value of the pedestrian flow evacuation density. An online error neural network construction module is used to construct an online error neural network for correcting the predicted value of pedestrian evacuation density; The data status anomaly judgment module is used to determine whether there is a data status anomaly based on the sensor status in the indoor evacuation scenario. An online error neural network training module is used to directly control indoor pedestrian evacuation based on the predicted pedestrian evacuation density if there are no abnormal data conditions, and to continuously train the online error neural network. The density prediction improvement value calculation module is used to calculate the prediction error value based on the currently trained error online neural network if there is an abnormal data status, and to calculate the sum of the predicted pedestrian flow evacuation density value and the prediction error value as the improved pedestrian flow evacuation density prediction value. An indoor pedestrian flow evacuation control module is used to control indoor pedestrian flow evacuation based on the predicted improved value of pedestrian flow evacuation density.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the indoor pedestrian flow evacuation control method as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the indoor pedestrian flow evacuation control method as described in any one of claims 1 to 4.

Images