An optimization method for evacuation guidance under limited vision conditions

Abstract

The present invention provides an evacuation guidance optimization method under limited vision conditions, which mainly solves the problem of the leader's optimal evacuation strategy under limited vision conditions for pedestrians. The present invention first establishes a motion model for different types of pedestrians based on a social force model, and modifies the calculation method of the expected force of pedestrians following the leader. Then, a grid-based leader initial position optimization scheme is defined, and the optimal initial position of the leader is determined according to the density cost value and distance cost value of the grid. Finally, a density factor and a distance factor are added to construct a dynamic exit selection strategy for the leader. The present method can achieve safe evacuation of pedestrians under limited vision conditions and effectively improve evacuation efficiency.

Description

Technical Field

[0001] The invention belongs to the field of personnel evacuation simulation, and in particular relates to an evacuation guidance optimization method under a limited field of view. Background Art

[0002] With the development of urban construction, more and more people gather in various places. In such places where people gather, once accidents such as fire, toxic gas, earthquake, etc. occur, it will cause immeasurable losses. Therefore, it is of great significance to formulate effective evacuation strategies when these accidents occur. Researchers have carried out a lot of research on the evacuation problem under accidents, and are committed to improving the efficiency of evacuation, thereby reducing casualties and property losses.

[0003] Evacuation models in emergency situations can be roughly divided into microscopic models and macroscopic models. The microscopic model regards pedestrians as independent particles that move under unified motion rules; the macroscopic model regards the crowd as a flowing whole, and then studies the macroscopic motion characteristics of the crowd. Since the microscopic model can better simulate the psychological and behavioral characteristics of pedestrians, the microscopic model is considered to be the most suitable evacuation simulation research model. Common microscopic models include cellular automaton models, social force models, lattice gas models, and agent-based models. The social force model was proposed by Helbing. This model can well simulate the continuous movement of pedestrians in the environment, and can also well reflect the distribution of pedestrians during the evacuation process, such as queuing, arching, etc. At the same time, it is also easy to simulate some psychological factors of pedestrians such as panic and impatience. Therefore, the social force model has a wide range of applications in personnel evacuation.

[0004] In the case of limited vision, it is difficult for pedestrians to evacuate safely on their own due to the lack of information about the location of leaders and exits. The use of exits is an important factor affecting evacuation efficiency. If the number of people choosing exits is unbalanced, it will cause pedestrian congestion in front of the exits, greatly reducing the evacuation efficiency. Existing research has rarely applied limited vision. Establishing an evacuation guidance model under limited vision, configuring the initial position of the leader and setting up a suitable exit selection strategy so that pedestrians can evacuate safely in a short time is of great significance for guiding the evacuation of pedestrians under limited vision. Summary of the invention

[0005] The purpose of the present invention is to provide an evacuation guidance optimization method under limited vision conditions, which enables a leader to guide pedestrians to complete evacuation more safely and improve evacuation efficiency.

[0006] An evacuation guidance optimization method under a limited field of view comprises the following steps:

[0007] Step 1: Construct a social force model that drives pedestrian motion and modify the calculation method of the follower's expected force;

[0008] Step 2: Use the grid-based leader initial position optimization algorithm to obtain the optimal leader initial distribution;

[0009] Step 3: Dynamically select the optimal exit for the leader based on the exit selection strategy of density cost and distance cost;

[0010] Step 4: Check the weight parameters and perform simulation verification on the model.

[0011] Furthermore, the pedestrian motion model and the follower expected force calculation method in step 1 are specifically as follows:

[0012] The movement of pedestrians and leaders is driven by the social force model; pedestrians move under the combined force of three forces, namely the desired force The interaction force between pedestrians i and j and the interaction force between pedestrian i and the wall

[0013] The relationship between the pedestrian's speed and force is:

[0014]

[0015] Where m i For the quality of pedestrians, is the actual walking speed of pedestrians;

[0016] is the self-driving force, representing the pedestrian’s willingness to reach the desired speed:

[0017]

[0018] in is the expected walking speed of pedestrian i, is the direction of the desired velocity, τ i is a constant, representing the time interval of pedestrian reaction time.

[0019] It is the force between pedestrians, including psychological force and contact physical force:

[0020]

[0021] in represents the negative force between pedestrians ij, which makes them move away from each other. is the force exerted on pedestrians when they come into contact with each other, the first term is the collision repulsion and the second term is the tangential friction. i ,B i ,k,κ are constant coefficients, g(x) is a piecewise function, which takes the value x when x≥0 and 0 otherwise. is the unit vector from pedestrian j to pedestrian i, and are the positions of pedestrians i and j respectively. and denote the velocity difference in the tangential and tangential directions, respectively;

[0022] is the relationship between pedestrian i and the wall:

[0023]

[0024] where r i is the radius of pedestrian i, d iw is the distance from the center of pedestrian i to the wall w;

[0025] The construction of the above pedestrian social force model guided by the leader:

[0026]

[0027] Among them, β i is a constant coefficient between zero and one, indicating the degree to which pedestrians trust the guidance of the leader. To represent navigation power, reflects the knowledge gained from leader r;

[0028] The pedestrian's desired direction of motion is a change vector related to the distance between the pedestrian and the leader, that is, when moving away from the leader, the pedestrian moves toward the leader. When approaching the leader, the desired force direction is consistent with the leader. The desired force direction of the guided pedestrian is modified to:

[0029]

[0030] Among them, e r (t) is the expected force direction of the leader, δ(x) is the normalization function used to obtain the unit vector. ω is the coefficient related to the distance between the pedestrian and the leader:

[0031]

[0032] Among them, σ is the radius of the field of view;

[0033] Furthermore, the grid-based leader initial position optimization algorithm in step 2 is specifically as follows:

[0034] The environment is divided into 6×6 squares to calculate the area with the highest density, and then the density cost and distance cost of each square are calculated. Taking the two costs into consideration, the center of the square with the highest cost is the optimized leader position. The square cost is:

[0035]

[0036] Among them, num(i) and num max are the number of pedestrians in the i-th square and the maximum number of pedestrians in the squares, are the horizontal and vertical coordinates of square i and gate j respectively, dis max is the maximum value of the grid distance cost, α is a constant coefficient (0,1) used to adjust the proportion of density cost and distance cost;

[0037] Furthermore, the exit selection strategy based on density cost and distance cost in step 3 is specifically as follows:

[0038] The cost value of each door is established according to the distribution of pedestrians and the distance between the leader and the exit. The leader selects the exit according to the size of the cost value. The cost of the door is:

[0039]

[0040] Among them, n j is the number of people closer to door j in the box centered on the leader, n all is the total number of pedestrians in the box, D rj is the distance between leader r and door j, L is the length of the environment side, and μ is a constant coefficient.

[0041] Furthermore, the step 4 of verifying the weight parameters and model simulation verification is specifically as follows:

[0042] In the simulation, 100 pedestrians are randomly distributed around the room, and the radius of pedestrian vision range is set to 2m. The experimental environment uses a 12m×12m room with two exits on the left and right. Considering the size of the room, 4 leaders are set. The leaders are marked with red dots, and ordinary pedestrians are marked with green dots;

[0043] The parameters of the social force model are consistent with those in existing studies: pedestrian radius r = 0.5 m, expected speed k = 1.2 × 10 4 kg·s -2 ,κ=2.4×10 5 kg·m -1 ·s -1 , A = 2000, B = 0.08m;

[0044] Different α values ​​lead to different leader position distributions, and the number of people that can be successfully guided varies. Simulations are performed under different α values, and the remaining number of people in each simulation is recorded. The α value with the smallest number of remaining people in the room is the optimal value.

[0045] Different μ values ​​will affect the leader's exit choice during the evacuation process, and ultimately affect the overall evacuation time. Simulations are performed under different μ values, and the evacuation time required for each simulation is recorded. Here we specify that if the evacuation is not completed within 60 seconds, the timing will be terminated; the μ value with the shortest evacuation time is the optimal value.

[0046] The beneficial effects of the present invention are:

[0047] The present invention mainly solves the problem of the leader's optimal evacuation strategy under the condition of limited vision of pedestrians. The method first establishes a motion model for different types of pedestrians based on the social force model, and modifies the calculation method of the expected force of pedestrians following the leader. Then, a grid-based leader initial position optimization scheme is defined, and the optimal initial position of the leader is determined according to the density cost value and distance cost value of the grid. Finally, the density factor and distance factor are added to construct the leader's dynamic exit selection strategy. This model can achieve the safe evacuation of pedestrians under limited vision and effectively improve the evacuation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A pedestrian density cost grid map required for the leader position distribution algorithm of the present invention;

[0049] Figure 2 Calculating a gate cost graph for the leader-centric box method in the leader exit selection strategy of the present invention;

[0050] Figure 3 This is a relationship diagram between different values ​​of α and the number of people remaining in the room of the present invention;

[0051] Figure 4 This is a relationship diagram between different values ​​of μ and evacuation time of the present invention;

[0052] FIG5( a ) is a diagram of the pedestrian following leader method of the present invention before modification;

[0053] FIG5( b ) is a modified diagram of the pedestrian following leader method of the present invention;

[0054] FIG6( a ) is a distribution diagram of the leader position uniform distribution method of the present invention;

[0055] FIG6( b ) is a distribution diagram of the maximum coverage method of leader position distribution of the present invention;

[0056] FIG6( c ) is a distribution diagram of the leader position distribution K-means clustering method of the present invention;

[0057] FIG. 7( a ) is a diagram of a simulation process of pedestrian evacuation in a uniformly distributed manner according to the present invention;

[0058] FIG7( b ) is a diagram showing a simulation process of pedestrian evacuation under the maximum coverage method distribution of the present invention;

[0059] FIG7( c ) is a diagram of the simulation process of pedestrian evacuation under K-means clustering distribution of the present invention;

[0060] FIG7( d ) is a diagram showing a simulation process of pedestrian evacuation under the distribution of the present invention;

[0061] Figure 8 A line graph showing the change of the number of people remaining in the room over time under different distributions of the present invention;

[0062] FIG9( a ) is a diagram showing a simulation process of pedestrian evacuation according to the nearest exit strategy of the present invention;

[0063] FIG9( b ) is a diagram showing a simulation process of pedestrian evacuation according to the exit selection strategy of the present invention;

[0064] Fig.10 It is a line graph showing the change of the number of successful people over time under different exit selection strategies of the present invention. DETAILED DESCRIPTION

[0065] The present invention is further described below in conjunction with the accompanying drawings.

[0066] An evacuation guidance optimization method under a limited field of view comprises the following steps:

[0067] Step 1: Construct a social force model that drives pedestrian motion and modify the calculation method of the follower's expected force;

[0068] The social force model was proposed by Helbing. The movement of pedestrians and leaders is driven by the social force model. Pedestrians move under the combined force of three forces, namely, the expected force The interaction force between pedestrians i and j, and the interaction force between pedestrian i and the wall, According to Newton's second law, the relationship between the pedestrian's velocity and force is expressed as:

[0069]

[0070] Where m i For the quality of pedestrians, is the actual walking speed of pedestrians;

[0071] In (1) the first item is the self-driving force, which represents the pedestrian’s willingness to reach the desired speed and is expressed as follows:

[0072]

[0073] in is the walking speed of pedestrian i, is the direction of the desired velocity, τ iis a constant, representing the time interval of pedestrian reaction time;

[0074] The second term in (1) is the force between pedestrians, including psychological force and contact physical force, expressed as follows:

[0075]

[0076] in represents the negative force between pedestrians ij, which makes them move away from each other. is the force exerted on pedestrians when they come into contact with each other, the first term is the collision repulsion and the second term is the tangential friction. i ,B i ,k,κ are constant coefficients, g(x) is a piecewise function, which takes the value x when x≥0 and 0 otherwise. is the unit vector from pedestrian j to pedestrian i, and are the positions of pedestrians i and j respectively. and denote the velocity difference in the tangential and tangential directions, respectively;

[0077] The third term in (1) represents the relationship between pedestrian i and the wall, and The form is similar and is expressed as follows:

[0078]

[0079] where r i is the radius of pedestrian i, d iw is the distance from the center of pedestrian i to the wall w;

[0080] According to the different information obtained within the field of vision of pedestrians in the environment, we assume that pedestrians have four movement situations;

[0081] a. If you can see the exit directly within your field of vision, move towards it;

[0082] b. If you cannot see the exit within your field of vision but can see the leader, move according to the information provided by the leader;

[0083] c. If there is neither an exit nor a leader within the field of vision, the pedestrians will be influenced by other pedestrians and form a herd phenomenon;

[0084] d. If there are no pedestrians, exits or leaders in the field of vision, the person moves according to his / her will until he / she hits a wall, thus generating “wall-following behavior”;

[0085] The above leader-guided pedestrian social force model is constructed as follows:

[0086]

[0087] where β i is a constant coefficient between zero and one, indicating the degree to which pedestrians trust the guidance of the leader. To represent navigation power, it reflects the knowledge gained from leader r;

[0088] The way other pedestrians construct social forces is consistent with (1) and (2), except that the direction of the expected force is different;

[0089] The pedestrian's expected movement direction is no longer directly pointing to the leader, but a changing vector related to the distance between the leader and the pedestrian, that is, when moving away from the leader, the pedestrian moves toward the leader, and when approaching the leader, the expected force direction is consistent with the leader, which is more in line with the actual situation. The expected force direction of the guided pedestrian is modified as follows:

[0090]

[0091] where e r (t) is the expected force direction of the leader, δ(x) is the normalization function used to obtain the unit vector. ω is the coefficient related to the distance between the pedestrian and the leader, expressed as:

[0092]

[0093] Among them, σ is the radius of the field of view, and ω changes with the distance between the pedestrian and the leader. When the pedestrian is far away from the leader, the larger ω is, the pedestrian moves toward the leader; when the pedestrian is close to the leader, the smaller ω is, the pedestrian moves in the direction indicated by the leader.

[0094] Step 2: Use the grid-based leader initial position optimization algorithm to obtain the optimal leader initial distribution;

[0095] The environment is divided into 6X6 squares to calculate the maximum density area, such as Figure 1 As shown, the density cost and distance cost of each square are calculated. Taking the two costs into consideration, the center of the square with the largest cost is the optimized leader position. The square cost is calculated as follows:

[0096]

[0097] Among them, num(i) and num max are the number of pedestrians in the i-th square and the maximum number of pedestrians in the squares, are the horizontal and vertical coordinates of square i and gate j respectively, dis maxis the maximum value of the grid distance cost. Step 3: The exit selection strategy based on density cost and distance cost dynamically selects the optimal exit for the leader;

[0098] The cost value of each door is established according to the distribution of pedestrians and the distance between the leader and the exit. The leader selects the exit according to the size of the cost value. The cost value of the door is expressed as:

[0099]

[0100] Among them, n j is the number of people closer to door j in the box centered on the leader, n all is the total number of pedestrians in the box, D rj is the distance between leader r and door j, L is the length of the environment, μ is a constant coefficient, and a box is used here to calculate the distribution of pedestrians around the leader, such as Figure 2 As shown;

[0101] Step 4: Check the weight parameters and simulate the model;

[0102] In the simulation, 100 pedestrians are randomly distributed around the room, and the radius of pedestrian vision range is set to 2m. The experimental environment uses a 12m×12m room with two exits on the left and right. Considering the size of the room, 4 leaders are set. The leaders are marked with red dots, and ordinary pedestrians are marked with green dots;

[0103] The parameters of the social force model are consistent with those in existing studies: pedestrian radius r = 0.5 m, expected speed k = 1.2 × 10 4 kg·s -2 ,κ=2.4×10 5 kg·m -1 ·s -1 , A = 2000, B = 0.08m. Different values ​​of α result in different leader position distributions, and the number of people that can be successfully guided varies. Simulations are performed under different values ​​of α, and the remaining number of people in each simulation is recorded;

[0104] The α value with the smallest number of remaining people in the room is the optimal value; Figure 3 As shown, when the value of α is between 0.5 and 0.7, the position allocation of the leader is optimal. The present invention selects α = 0.5;

[0105] Different μ values ​​will affect the leader's exit selection during the evacuation process, and ultimately affect the overall evacuation time. Simulations are performed under different μ values, and the evacuation time required for each simulation is recorded. Here, we specify that if the evacuation is not completed within 60 seconds, the timing will be terminated;

[0106] The μ value with the shortest evacuation time is the optimal value. Figure 4 Evacuation time under different values. It can be found that when μ is 0.5, the evacuation efficiency is the highest. Therefore, μ=0.5;

[0107] The guided evacuation behavior simulation settings are as follows:

[0108] Pedestrians are randomly distributed in the scene, and leaders are also randomly distributed. The number of pedestrians is 100, and the number of leaders is set to 4. The pedestrian's field of view is 2m, that is, when the distance between the pedestrian and the leader is less than or equal to 2m, the pedestrian can see the leader and get the guidance of the leader. Compare the movement before and after changing the pedestrian following method;

[0109] As shown in Figure 5(a), it can be seen that when the traditional following method is adopted, the leader is obviously surrounded. The modified following mode solves this problem. As shown in Figure 5(b), the pedestrian moves closer to the leader when moving away from the leader. When closer to the leader, the pedestrian moves in the same direction as the leader;

[0110] The simulation settings for the impact of the leader's initial position distribution on evacuation are as follows:

[0111] The commonly used methods for leader position distribution include uniform distribution, maximum coverage method, and K-means clustering. Figure 6(a)-Figure 6(c) Under the above parameter settings, the evacuation guidance situation under the optimized leader initial position distribution and the three commonly used distributions is compared.

[0112] from Figure 7(a)-Figure 7(c) It can be seen that due to the limited field of view of pedestrians, some pedestrians still cannot be guided even with the three commonly used leader distributions. The proposed method takes into account the density and the distance to the exit, and the optimized leader position ensures that all pedestrians are guided, as shown in Figure 7(d). The density factor ensures that as many pedestrians as possible are covered. The distance factor ensures that the leader can guide more pedestrians in the process of walking towards the exit;

[0113] In order to further verify the effectiveness of the proposed method, Figure 8 The relationship between the remaining number of people and the evacuation time was simulated. In order to ensure the reliability of the simulation, the average calculation of the results of ten simulations was performed. It can be seen that the number of pedestrians remaining in the room at 30 seconds is less than that of other methods, and almost all pedestrians are evacuated within 40 seconds;

[0114] The simulation settings for the impact of leader exit selection on evacuation are as follows:

[0115] In existing research, leaders usually use the nearest exit strategy to select their targets. In order to better demonstrate the advantages of the dynamic exit selection strategy, pedestrians are initially unevenly distributed, with slightly more pedestrians in the right half. The method of the present invention is compared with the nearest exit strategy, and the time required for the two methods to complete the evacuation is recorded. In order to ensure the effectiveness of the simulation, the average results of ten simulations are taken.

[0116] In order to verify the effectiveness of the proposed leader position selection strategy, the two leader position selection strategies are compared after the initial leader position is calculated using the proposed leader position optimization method. As can be seen from Figure 7(a), the leader selects the nearest exit, which causes more pedestrians to choose the right exit, causing serious congestion. In the optimal exit selection strategy, the leader dynamically selects the exit according to the pedestrian distribution. As can be seen from Figure 9(b), the proposed strategy is effective for the leader to guide some pedestrians in a smaller range.

[0117] In summary, in order to fully guide pedestrians under limited vision, a leader initial position allocation method that comprehensively considers density and distance factors is proposed. In view of the uneven distribution of pedestrian exits during the evacuation process, an optimized exit selection strategy is proposed, in which the leader selects the exit in real time based on the distance to the exit and the density of pedestrian distribution. A series of simulation experiments prove the effectiveness of the evacuation model in this paper under limited vision.

[0118] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An evacuation guidance optimization method under limited vision conditions, characterized in that: The following steps are involved: Step 1: Construct a social force model that drives pedestrian motion and modify the calculation method of the follower's expected force; The movement of pedestrians and leaders is driven by the social force model; pedestrians move under the combined force of three forces, namely the desired force The interaction force between pedestrians i and j and the interaction force between pedestrian i and the wall The relationship between the pedestrian's speed and force is: Where m i For the quality of pedestrians, is the actual walking speed of pedestrians; is the self-driving force, representing the pedestrian’s willingness to reach the desired speed: in is the expected walking speed of pedestrian i, is the direction of the desired velocity, τ i is a constant, representing the time interval of pedestrian reaction time; It is the force between pedestrians, including psychological force and contact physical force: in represents the negative force between pedestrians ij, which makes them move away from each other; is the force exerted on pedestrians when they come into contact with each other, the first term is the collision repulsion and the second term is the tangential friction; A i ,B i , k, κ are constant coefficients, g(x) is a piecewise function, which takes the value x when x ≥ 0, otherwise it takes the value 0; is the unit vector from pedestrian j to pedestrian i, and are the positions of pedestrians i and j respectively; and denote the velocity difference in the tangential and tangential directions, respectively; is the relationship between pedestrian i and the wall: where r i is the radius of pedestrian i, d iw is the distance from the center of pedestrian i to the wall w; The construction of the above pedestrian social force model guided by the leader: Among them, β i is a constant coefficient between zero and one, indicating the extent to which pedestrians trust the guidance of the leader; To represent navigation power, it reflects the knowledge gained from leader r; The pedestrian's desired direction of motion is a change vector related to the distance between the pedestrian and the leader, that is, when moving away from the leader, the pedestrian moves toward the leader. When approaching the leader, the desired force direction is consistent with the leader. The desired force direction of the guided pedestrian is modified to: Among them, e r (t) is the expected force direction of the leader, δ(x) is the normalization function used to obtain the unit vector; ω is the coefficient related to the distance between the pedestrian and the leader: Among them, σ is the radius of the field of view; Step 2: Use the grid-based leader initial position optimization algorithm to obtain the optimal leader initial distribution; Step 3: Dynamically select the optimal exit for the leader based on the exit selection strategy of density cost and distance cost; Step 4: Check the weight parameters and perform simulation verification on the model.

2. The evacuation guidance optimization method under limited vision conditions according to claim 1, characterized in that: The grid-based leader initial position optimization algorithm in step 2 is specifically as follows: The environment is divided into 6×6 squares to calculate the area with the highest density, and then the density cost and distance cost of each square are calculated. Taking the two costs into consideration, the center of the square with the highest cost is the optimized leader position; the square cost is: Among them, num(i) and num max are the number of pedestrians in the i-th square and the maximum number of pedestrians in the squares, are the horizontal and vertical coordinates of square i and gate j respectively, dis max is the maximum value of the grid distance cost, and α is a constant coefficient (0,1) used to adjust the proportion of density cost and distance cost.

3. The evacuation guidance optimization method under limited vision conditions according to claim 1, characterized in that: The exit selection strategy based on density cost and distance cost in step 3 is specifically as follows: The cost value of each door is established according to the distribution of pedestrians and the distance between the leader and the exit. The leader selects the exit according to the size of the cost value. The cost of the door is: Among them, n j is the number of people closer to door j in the box centered on the leader, n all is the total number of pedestrians in the box, D rj is the distance between leader r and door j, L is the length of the environment side, and μ is a constant coefficient.

4. The evacuation guidance optimization method under limited vision conditions according to claim 1, characterized in that: The step 4 of checking weight parameters and model simulation verification is specifically as follows: In the simulation, 100 pedestrians are randomly distributed around the room, and the radius of pedestrian vision range is set to 2m. The experimental environment uses a 12m×12m room with two exits on the left and right. Considering the size of the room, 4 leaders are set; the leaders are marked as red dots, and ordinary pedestrians are marked as green dots. The parameters of the social force model are consistent with those in existing studies: pedestrian radius r = 0.5 m, expected speed k = 1.2 × 10 4 kg·s -2 ,κ=2.4×10 5 kg·m -1 ·s -1 , A = 2000, B = 0.08m; Different α values ​​lead to different leader position distributions, and the number of people that can be successfully guided varies. Simulations are performed under different α values, and the remaining number of people in each simulation is recorded. The α value with the smallest number of remaining people in the room is the optimal value. Different μ values ​​will affect the leader's exit choice during the evacuation process, and ultimately affect the overall evacuation time. Simulations are performed under different μ values, and the evacuation time required for each simulation is recorded. Here we specify that if the evacuation is not completed within 60 seconds, the timing will be terminated; the μ value with the shortest evacuation time is the optimal value.

Images