Pedestrian path prediction method based on cellular automaton model

Abstract

The invention discloses a pedestrian path prediction method based on a cellular automaton model, and relates to the technical field of path prediction, and the method comprises the following steps: calculating a static field, a dynamic field and a rejection field in a pedestrian motion process; calculating the transition probability of each pedestrian moving to unoccupied neighbor cells according to the static field, the dynamic field and the rejection field; judging whether a plurality of pedestrians enter one cell or not based on the transition probability; judging whether the pedestrian entering the cell has the maximum transition probability or not; and based on the current position of the pedestrian, repeating the above steps until the pedestrian reaches a set cut-off distance, and completing pedestrian path prediction based on the cellular automaton model. The method solves the problem that the existing method does not consider the influence of the surrounding environment and the crowd on the pedestrian path, so that the pedestrian flow cannot be accurately simulated.

Description

Technical Field

[0001] The present invention relates to the technical field of path prediction, and in particular to a pedestrian path prediction method based on a cellular automaton model. Background Art

[0002] Crowd behavior and self-organized walking models can be divided into two categories: (1) based on the macroscopic characteristics of pedestrian flow and the relationship between pedestrian walking speed and the surrounding environment, a macroscopic model is constructed; (2) based on the relationship between human walking speed and the surrounding environment, a microscopic model is constructed. When we study the impact of human vibration on crowd loads, macroscopic models are usually used to describe the movement trajectory of the crowd, but they often ignore the social characteristics of humans. Humans are complex creatures that are influenced by the surrounding environment and the crowd when making decisions. However, existing methods do not consider the impact of the surrounding environment and the crowd on pedestrian paths and cannot accurately simulate pedestrian flow. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides a pedestrian path prediction method based on a cellular automaton model, which solves the problem that the prior method does not consider the impact of the surrounding environment and the crowd on the pedestrian path, thereby failing to accurately simulate the pedestrian flow.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a pedestrian path prediction method based on a cellular automaton model, comprising the following steps:

[0005] S1: Calculate the static field, dynamic field and repulsive field during pedestrian movement;

[0006] S2: Calculate the transition probability of each pedestrian moving to an unoccupied neighbor cell based on the static field, dynamic field, and repulsion field;

[0007] S3: judging whether there are multiple pedestrians entering a cell based on the transition probability, if yes, proceeding to step S4, otherwise the pedestrian is allowed to move and proceeding to step S5;

[0008] S4: Determine whether the pedestrian entering the cell has the maximum transfer probability. If so, the pedestrian is allowed to move and the process proceeds to step S5. Otherwise, the pedestrian remains at the original position and the process proceeds to step S5.

[0009] S5: Based on the current position of the pedestrian, repeat the above steps S1-S4 until the pedestrian reaches the set cutoff distance, completing the pedestrian path prediction based on the cellular automaton model.

[0010] Furthermore, the calculation of the static field in S1 includes the following sub-steps:

[0011] S11: Set the exit unit, assign the value of the cell at the exit unit to 1, and assign the value of other cells to 100;

[0012] S12: Based on the assigned cells, add 1 to the horizontally adjacent cells and add 1.4 to the diagonally adjacent cells;

[0013] S13: Calculate the static field S based on the assignment result ij , the formula is:

[0014]

[0015] S ij =min{[(ii okx ) 2 +(jj oky ) 2 ] 12}

[0016] Among them, (i, j) is the unit cell position, Γ is the unit set of space available for people to walk, is the initial static field, (i okx ,j oky ) is the location of the kth exit.

[0017] Furthermore, the dynamic field in S1 The calculation formula is:

[0018]

[0019] Among them, α is the diffusion probability of the cell dynamic field strength, β is the attenuation probability of the cell dynamic field strength, is the dynamic field of cell (i, j) at the previous time step, and are the dynamic fields of the neighbors of cell (i, j) in the previous time step.

[0020] Furthermore, the repulsive field E in S1 kij The calculation formula is:

[0021]

[0022] Among them, i kx is the x-coordinate of the kth pedestrian, i lx is the x coordinate of the lth pedestrian, j ky is the y coordinate of the kth pedestrian, j ly is the y coordinate of the lth pedestrian, and n is the number of pedestrians within the set range around the kth pedestrian.

[0023] Furthermore, the transition probability P of each pedestrian in S2 moving to an unoccupied neighbor cell is ijThe calculation formula is:

[0024]

[0025] k S ∈[0,+∞]

[0026] k D ∈[0,+∞]

[0027] k E ∈[0,+∞]

[0028] Where N is the normalization factor, η ij and ij are the parameters describing the cell state, k S , k D and k E are the static field parameter, dynamic field parameter and repulsive field parameter respectively, S ij , and E kij They are static field, dynamic field and repulsive field respectively.

[0029] Furthermore, the calculation formula of the normalization factor N is:

[0030]

[0031] Furthermore, the parameter η describing the cell state ij and ij The calculation formula is:

[0032]

[0033] The beneficial effects of the present invention are as follows: the present invention adopts cellular automata, and by using cellular automata technology, it can effectively capture the complex self-regulation mechanism of pedestrians in different environments, thereby reflecting the complex dynamic mechanism of macroscopic, microscopic and self-regulation of pedestrians in real time. Compared with traditional continuous simulation technology, the simulation effect of cellular automata technology is significantly improved.

[0034] The present invention combines the dynamic field, the static field and the repulsive field to simulate the influence of the surrounding environment and the crowd on the pedestrian path, thereby being closer to the real situation and simulating the pedestrian flow more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of a pedestrian path prediction method based on a cellular automaton model.

[0036] Figure 2 A map of the bridge model.

[0037] Figure 3 This is a simulation diagram of the bridge working condition.

[0038] Figure 4 This is a graph showing the changes in the shortest motion time for various working conditions when the number of pedestrians N changes.

[0039] Figure 5 This is the motion position change diagram after repulsive field optimization.

[0040] Figure 6 This is a simulation diagram of the working conditions before and after the repulsion field optimization.

[0041] Figure 7 Pedestrian trajectory map. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0043] like Figure 1 As shown, a pedestrian path prediction method based on a cellular automaton model includes the following steps:

[0044] S1: Calculate the static field, dynamic field and repulsive field during pedestrian movement;

[0045] S2: Calculate the transition probability of each pedestrian moving to an unoccupied neighbor cell based on the static field, dynamic field, and repulsion field;

[0046] S3: judging whether there are multiple pedestrians entering a cell based on the transition probability, if yes, proceeding to step S4, otherwise the pedestrian is allowed to move and proceeding to step S5;

[0047] S4: Determine whether the pedestrian entering the cell has the maximum transfer probability. If so, the pedestrian is allowed to move and the process proceeds to step S5. Otherwise, the pedestrian remains at the original position and the process proceeds to step S5.

[0048] S5: Based on the current position of the pedestrian, repeat the above steps S1-S4 until the pedestrian reaches the set cutoff distance, completing the pedestrian path prediction based on the cellular automaton model.

[0049] The static field describes the behavior of pedestrians in an automatic cellular machine model, where pedestrians choose the shortest path to move to the exit. For a certain cellular automaton model, the static floor field is fixed and does not change over time, nor does it change due to the presence or movement of pedestrians. The static field can be used to calibrate the degree of attraction of an area in space to cells, and generally reaches a maximum value at an emergency exit or a bridge exit. The value of the static field is related to the distance of the grid from the exit. Generally, the closer to the exit, the larger the field value.

[0050] The calculation of the static field in S1 includes the following steps:

[0051] S11: Set the exit unit, assign the value of the cell at the exit unit to 1, and assign the value of other cells to 100;

[0052] S12: Based on the assigned cells, add 1 to the horizontally adjacent cells and add 1.4 to the diagonally adjacent cells;

[0053] S13: Calculate the static field S based on the assignment result ij , the formula is:

[0054]

[0055] S ij =min{[(ii okx ) 2 +(jj oky ) 2 ] 12}

[0056] Among them, (i, j) is the unit cell position, Γ is the unit set of space available for people to walk, is the initial static field, (i okx ,j oky ) is the location of the kth exit.

[0057] The dynamic field can be seen as a mechanism to simulate human behavior. It originates from the pheromones released by ant colonies when foraging. These pheromones can stimulate a dynamic field, that is, when pedestrians leave, they will release bosons, leaving their own movement tracks in space, and these tracks will continue to change over time. Among them, the dynamic field is a virtual pedestrian trajectory. This is used to model the mutual attraction between particles and can be used to establish a model of mutual attraction between particles. The dynamic field will appear in the following three stages during the movement of pedestrians:

[0058] (1) At t = 0, all pedestrians are not moving, and the dynamic field value D of all points (i, j) is zero, that is, D ij =0.

[0059] (2) Once a person walks over a unit cell (i, j), the dynamic field at that location increases by 1, i.e., D ij =D ij +1.

[0060] (3) As time changes, the dynamic field will diffuse or decay, which can be expressed by the following formula calculate.

[0061] The dynamic field of S1 The calculation formula is:

[0062]

[0063] Wherein, α is the diffusion probability of the cell dynamic field strength, β is the attenuation probability of the cell dynamic field strength, in this embodiment, α=β=0.2, is the dynamic field of cell (i, j) at the previous time step, and are the dynamic fields of the neighbors of cell (i, j) in the previous time step.

[0064] In previous studies on automatic cellular machine models, only the effects of static and dynamic fields on pedestrian movement paths are usually considered. That is, only the interaction between pedestrians and exits and the attraction of pedestrian trajectories to other pedestrians are considered, while the mutual repulsion between pedestrians is ignored. Under high-density traffic conditions, people will try to keep a certain distance from each other to avoid contact. Pedestrians generally try to minimize their interactions to avoid the discomfort and inefficiency caused by contact. This scheme proposes a repulsive floor field to describe the mutual repulsion between pedestrians. The repulsive field changes with the position of the pedestrian and the relative position with other pedestrians. Each pedestrian has a separate repulsive field at each different position.

[0065] The repulsion field for a certain pedestrian is calculated based on a certain distance index between each grid point and other pedestrians, so the field value increases in the direction away from other pedestrians, and the closer to other pedestrians, the smaller the field value. When the distance between pedestrians is too large, the mutual repulsion becomes weaker and the line of sight is blocked by other pedestrians, and the interaction between pedestrians will become weaker. In this embodiment, the interaction between pedestrians within 1.5m apart constitutes the repulsion field. Assuming that there are n pedestrians within 1.5m around the kth pedestrian, the final value of the repulsion floor field of the kth pedestrian is calculated as follows:

[0066] The repulsive field E in S1 kij The calculation formula is:

[0067]

[0068] Among them, i kx is the x-coordinate of the kth pedestrian, i lx is the x coordinate of the lth pedestrian, j ky is the y coordinate of the kth pedestrian, j ly is the y coordinate of the lth pedestrian, and n is the number of pedestrians within the set range around the kth pedestrian.

[0069] The transition probability P of each pedestrian in S2 moving to an unoccupied neighbor cell ij The calculation formula is:

[0070]

[0071] k S ∈[0,+∞]

[0072] k D ∈[0,+∞]

[0073] k E ∈[0,+∞]

[0074] Where N is the normalization factor, η ij and ij are the parameters describing the cell state, k S , k D and k E are the static field parameter, dynamic field parameter and repulsive field parameter respectively, S ij , and E kij They are static field, dynamic field and repulsive field respectively.

[0075] The calculation formula of the normalization factor N is:

[0076]

[0077] The parameter η describing the cell state ij and ij The calculation formula is:

[0078]

[0079] The pedestrian movement rules based on the cellular automaton model established by the present invention are:

[0080] (1) The basis for adjusting and modifying the dynamic field value is the diffusion and attenuation rules;

[0081] (2) Each pedestrian has a separate repulsion field, which is determined by its relative position to the surrounding pedestrians;

[0082] (3) The transition probability P of each pedestrian moving to an unoccupied neighbor cell (i, j) is calculated through the three floor fields: static field, dynamic field and repulsion field. ij , this probability determines the next movement direction of the pedestrian. S , dynamic field parameter k D and the repulsive field parameter k E Three sensitivity parameters weight the floor field;

[0083] (4) Each cell representing a pedestrian has a transition probability P calculated in the previous step. ij Select a target cell to transfer;

[0084] (5) When multiple pedestrians try to move to the same target unit, conflicts will occur. This scheme uses a probabilistic method to resolve such conflicts. Movement probability P ij Larger pedestrians are allowed to move, and those who are allowed to move will continue to perform their steps;

[0085] (6) The dynamic field value increases with the cell motion, and the repulsive field value changes continuously.

[0086] In one embodiment of the present invention, a pedestrian path prediction method based on a cellular automaton model established in this solution is simulated.

[0087] The simulated bridge size is 4×26m, with a total of 10×65 cellular automaton grid cells. Each pedestrian is represented by a cell in the cellular automaton, the pedestrian's walking speed is 1.440m / s, and the time step is 0.2778s. The static field parameter k S is 3, the repulsive field parameter k D =0.2, dynamic field parameter k E = 0.3, based on experience, this scheme takes the dynamic field diffusion and attenuation parameters α = β = 0.2. The bridge model and bridge working condition simulation are as follows Figure 2 and Figure 3 The following is an analysis of the impact of the number of pedestrians on the pedestrian flow model, the impact of adding social force optimization on the pedestrian flow model, and the pedestrian movement trajectory.

[0088] The impact of the number of pedestrians on the pedestrian flow model:

[0089] According to experience, the number of pedestrians has a very important influence on the movement time. This paper sets the model parameters as follows: static field parameter k S =3.0, dynamic field parameter k D =0.3, when the distance between pedestrians is less than 1.5m, the repulsion field parameter k E =1.0, when the distance between pedestrians is greater than 1.5m, the repulsion field parameter k E = 0, dynamic field diffusion and attenuation parameters α = β = 0.2. The number of pedestrians on the bridge is N = 10, 20, 30, 40, 50, 60, and the passing states are one-way passing, two-way passing, and emergency evacuation.

[0090] according to Figure 4It can be clearly seen that there is a close relationship between the initial number of pedestrians and the movement time, especially in the case of one-way or two-way movement. When the number of people in the pedestrian bridge increases, the movement time will also increase significantly. When the number of pedestrians increases from 10 to 40, the movement time in the evacuation state increases from 9.72 to 14.16s, and the movement time of pedestrians has a relatively large change. As the initial number of people continues to increase, the growth rate of the movement time slows down significantly. This is because the exit of the pedestrian bridge is not blocked. Even after the number of people reaches a certain level, there will be no congestion. In addition, the pace and distribution of pedestrians will also have an important impact on the movement time.

[0091] It can also be seen that in both one-way and two-way conditions, the movement time of pedestrians is basically proportional to the initial number of people. This is because the simulation of these two conditions requires continuous entry of pedestrians from the entrance, and the time interval for pedestrians to enter the cellular automaton map is 5 time steps, so the movement time of pedestrians mainly depends on the time when pedestrians enter the cellular automaton map. Studies have shown that when the number of pedestrians is small, the time for two-way walking is shorter than that for one-way walking, but as the number of pedestrians increases, the time for two-way walking will increase significantly. This is because the congestion on the bridge becomes more severe, and pedestrians walking in both directions may bump into each other, causing congestion on the bridge, making the time for two-way walking longer. According to the chart, the activity duration of pedestrians is proportional to the number of their participants. In addition, the starting and ending points of pedestrians will also have an important impact on the overall duration and efficiency of their activities.

[0092] The impact of adding social force optimization on the pedestrian flow model:

[0093] In previous simulation experiments, only the interaction between the two types of floor fields, static field and dynamic field, is usually considered, and the interaction between pedestrians is ignored. According to existing research, pedestrians always try to maintain a certain distance from each other, so this scheme adds a real-time repulsive floor field that is different for each pedestrian. When the distance between pedestrians is too large, the interaction is relatively weak due to the obstruction of vision and the long distance. Therefore, the mutual repulsive force generated when the distance from the pedestrian is less than 1.5m is used to form the repulsive floor field. After the optimization of the repulsive field, the position of the pedestrian during movement has a very obvious change, such as Figure 5 As shown, it is obviously more consistent with the movement of pedestrians in real situations.

[0094] The number of pedestrians N is set to 30, 40, 50, and 60 respectively. The number of pedestrians gathered at a certain time is counted for the optimized model and the model before optimization in three working modes: evacuation state, one-way movement state, and two-way movement state.

[0095] The fixed parameters set are as follows: static field parameter kS =3.0, fixed dynamic field parameter k D = 0.3, and the dynamic field diffusion and attenuation parameters α = β = 0.2 are taken according to experience. The repulsive field sensitivity parameter k E Take 0 and 1.0 respectively, the simulation results are as follows Figure 6 shown.

[0096] Figure 6 The number of pedestrians gathered after 5 time steps under three working conditions, free evacuation, one-way movement and two-way movement, with different initial numbers before and after the addition of the repulsion field optimization was compared. It can be seen that after the addition of the repulsion field optimization, the number of pedestrians gathered during the movement process was greatly reduced, so it is more in line with the actual situation.

[0097] Pedestrian movement trajectory:

[0098] Combined with the pedestrian flow model of the pedestrian bridge, the spatial coordinates of pedestrians at each time step on the pedestrian bridge are recorded respectively, and the movement trajectory of pedestrians can be obtained by connecting them. The following takes the trajectory of the pedestrian flow model of the pedestrian bridge with initial number of people N = 10 and 20 as examples. The static field parameter k S =3.0, fixed dynamic field parameter k D = 0.3, and the dynamic field diffusion and attenuation parameters α = β = 0.2 are taken according to experience. The repulsive field sensitivity parameter is 1, and the pedestrian trajectory diagram is as follows Figure 7 shown.

[0099] according to Figure 7 ,The pedestrian flow model of cellular automata can accurately simulate the movement trajectory of pedestrians, and can fully consider the influence of the surrounding environment and the crowd, so as to provide the optimal route at the exit. This model can more accurately reflect the behavioral characteristics of pedestrians, thus providing a more accurate movement trajectory.

[0100] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the invention.

Claims

1. A pedestrian path prediction method based on a cellular automaton model, characterized in that: The following steps are involved: S1: Calculate the static field, dynamic field and repulsive field during pedestrian movement; S2: Calculate the transition probability of each pedestrian moving to an unoccupied neighbor cell based on the static field, dynamic field, and repulsion field; S3: judging whether there are multiple pedestrians entering a cell based on the transition probability, if yes, proceeding to step S4, otherwise the pedestrian is allowed to move and proceeding to step S5; S4: Determine whether the pedestrian entering the cell has the maximum transfer probability. If so, the pedestrian is allowed to move and the process proceeds to step S5. Otherwise, the pedestrian remains at the original position and the process proceeds to step S5. S5: Based on the current position of the pedestrian, repeat the above steps S1-S4 until the pedestrian reaches the set cutoff distance, completing the pedestrian path prediction based on the cellular automaton model.

2. A pedestrian path prediction method based on a cellular automaton model according to claim 1, characterized in that: The calculation of the static field in S1 includes the following steps: S11: Set the exit unit, assign the value of the cell at the exit unit to 1, and assign the value of other cells to 100; S12: Based on the assigned cells, add 1 to the horizontally adjacent cells and add 1.4 to the diagonally adjacent cells; S13: Calculate the static field S based on the assignment result ij , the formula is: S ij =min{[(ii okx ) 2 +(jj oky ) 2 ] 12 } Among them, (i, j) is the unit cell position, Γ is the unit set of space available for people to walk, is the initial static field, (i okx ,j oky ) is the location of the kth exit.

3. The pedestrian path prediction method based on the cellular automaton model according to claim 1 is characterized in that: The dynamic field of S1 The calculation formula is: Among them, α is the diffusion probability of the cell dynamic field strength, β is the attenuation probability of the cell dynamic field strength, is the dynamic field of cell (i, j) at the previous time step, and are the dynamic fields of the neighbors of cell (i, j) in the previous time step.

4. The pedestrian path prediction method based on cellular automaton model according to claim 1, characterized in that: The repulsive field E in S1 kij The calculation formula is: Among them, i kx is the x-coordinate of the kth pedestrian, i lx is the x coordinate of the lth pedestrian, j ky is the y coordinate of the kth pedestrian, j ly is the y coordinate of the lth pedestrian, and n is the number of pedestrians within the set range around the kth pedestrian.

5. The pedestrian path prediction method based on cellular automaton model according to claim 1, characterized in that: The transition probability P of each pedestrian in S2 moving to an unoccupied neighbor cell ij The calculation formula is: k S ∈[0,+∞] k D ∈[0,+∞] k E ∈[0,+∞] Where N is the normalization factor, η ij and ij are the parameters describing the cell state, k S , k D and k E are the static field parameter, dynamic field parameter and repulsive field parameter respectively, S ij , and E kij They are static field, dynamic field and repulsive field respectively.

6. The pedestrian path prediction method based on cellular automaton model according to claim 5, characterized in that: The calculation formula of the normalization factor N is:

7. The pedestrian path prediction method based on cellular automaton model according to claim 5, characterized in that: The parameter η describing the cell state ij and ij The calculation formula is:

Images