CN111767789A - Crowd evacuation method and system based on multi-carrier intelligent guidance - Google Patents
Crowd evacuation method and system based on multi-carrier intelligent guidance Download PDFInfo
- Publication number
- CN111767789A CN111767789A CN202010401446.0A CN202010401446A CN111767789A CN 111767789 A CN111767789 A CN 111767789A CN 202010401446 A CN202010401446 A CN 202010401446A CN 111767789 A CN111767789 A CN 111767789A
- Authority
- CN
- China
- Prior art keywords
- evacuation
- leader
- guidance
- strategy
- path
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000004088 simulation Methods 0.000 claims abstract description 35
- 230000002787 reinforcement Effects 0.000 claims abstract description 34
- 238000012360 testing method Methods 0.000 claims abstract description 21
- 238000012795 verification Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 15
- 230000033001 locomotion Effects 0.000 claims description 48
- 230000009471 action Effects 0.000 claims description 43
- 238000004422 calculation algorithm Methods 0.000 claims description 24
- 238000005457 optimization Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 17
- 230000003993 interaction Effects 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 8
- 238000003064 k means clustering Methods 0.000 claims description 5
- 238000012790 confirmation Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000000007 visual effect Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 238000011160 research Methods 0.000 abstract description 10
- 238000010586 diagram Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Development Economics (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Game Theory and Decision Science (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Educational Administration (AREA)
- Computer Security & Cryptography (AREA)
- Quality & Reliability (AREA)
- Medical Informatics (AREA)
- Operations Research (AREA)
- Computing Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Emergency Lowering Means (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses a crowd evacuation method and a system thereof based on multi-carrier intelligent guidance, wherein the method comprises the following steps: the method comprises the steps of leader evacuation path planning based on reinforcement learning, cooperative guiding strategy based on evacuation identification and a leader and simulation test verification. The system uses the above multi-carrier intelligent guidance crowd evacuation method, and the system comprises: the system comprises a data acquisition and processing module, an intelligent guide strategy generation module and a simulation test verification module. Compared with the evacuation strategy designed based on experience and trial and error in the existing research, the method can more effectively improve the crowd evacuation efficiency. Under the intelligent guide strategy, the configuration of various guide modes of the station has higher flexibility, and the operation cost of the station can be reduced on the premise of ensuring effective safe evacuation.
Description
Technical Field
The invention belongs to the technical field of rail transit management operation, and relates to a crowd evacuation method and system based on multi-carrier intelligent guidance.
Background
The rail transit station is used as a place for collecting and distributing passengers, generally has the characteristics of complex internal structure, relatively closed space, poor connectivity with the outside and the like, and once emergencies such as large passenger flow, fire, terrorist attack and the like occur, the passengers are difficult to evacuate in emergency and rescue, and are easy to cause group death and group injury.
The problem of guiding and evacuating passengers is always the research focus in the field of pedestrian dynamics research, and particularly in places with dense personnel and complex structures, the problem of safe evacuation of pedestrians has attracted high attention. Previous researches have proved the necessity of arranging emergency signs and guides for passenger evacuation, but the distribution schemes for emergency signs and guides are mostly based on experience or trial and error, and although the distribution schemes based on experience or trial and error can improve the passenger evacuation efficiency to a certain extent, the optimality of the distribution schemes cannot be guaranteed. In addition, the existing research mostly focuses on the scheme research of a single guidance mode, and the research of the single guidance mode does not consider the coordination aspect of multiple guidance modes in the existing station. Therefore, a cooperative station passenger guiding method oriented to operation efficiency and safety becomes a current research focus, and how to reasonably set the number and the positions of emergency signs, guides and the like, so that the operation cost is reduced while the overall evacuation efficiency is improved is a problem to be urgently solved.
In many studies considering guidance of crowd evacuation by a guidance member, the guidance member is assumed to be completely familiar with an evacuation environment and guides crowd evacuation in accordance with a predetermined evacuation route. However, in an actual evacuation environment, the evacuation environment and the crowd state are dynamically changed, and a leader cannot adjust an evacuation path according to the dynamically changed environment state, so that a serious congestion phenomenon occurs on an individual predetermined evacuation path, and the overall evacuation efficiency is greatly influenced.
With the continuous development of artificial intelligence technology, the research of multi-agent path planning problem based on reinforcement learning also presents many successful application examples. Compared with other path planning algorithms, the reinforcement learning algorithm has the advantages that when the intelligent agent is not familiar with the environment, the self-learning capability and the self-adaptability of the algorithm can enable the intelligent agent to learn to obtain a path which achieves the target and has the highest accumulated benefit in the interaction process with the environment. The model-free reinforcement learning framework enables the applicable application scenarios of path planning to be wider.
Disclosure of Invention
The invention aims to provide a crowd evacuation method and system based on multi-carrier intelligent guidance, which effectively improve crowd evacuation efficiency.
The technical scheme of the invention is as follows:
a crowd evacuation method based on multi-carrier intelligent guidance comprises a leader evacuation path planning based on reinforcement learning, a collaborative guidance strategy based on evacuation identification and a leader, and simulation test verification.
Preferably, the leader evacuation path planning based on reinforcement learning specifically comprises the following steps:
the method comprises the following steps: extracting a pedestrian position coordinate; performing video processing on a real station video by taking a frame as a unit, extracting the position coordinates of each frame of pedestrian by adopting a video detection and tracking algorithm, and expressing the motion path of each pedestrian reaching an exit position according to a time sequence;
step two: initializing pedestrian groups; initializing and grouping pedestrians in the video by adopting a K-Means clustering algorithm;
step three: normalizing the motion path in the group; calculating the weighted average of the motion paths of all the people in the group as a group common motion path;
step four: leader path planning based on reinforcement learning; and modifying a mapping strategy from the state to the action according to the environment state and the return value learned in the last step by adopting a Q-learning algorithm through interaction between the intelligent agent and the environment, so that the action obtains the maximum accumulated return value from the environment, and finally an optimal motion strategy, namely the optimal motion path of the leader is obtained.
Preferably, the collaborative guiding strategy based on the evacuation identification and the leader is based on the leader evacuation path based on reinforcement learning, and specifically comprises the following steps:
the method comprises the following steps: defining evacuation signs and a leader navigation force model; the evacuation guide marks in the actual subway station are mostly arranged on the surfaces such as walls, and the guide capability of the evacuation guide marks is closely related to the distance and the observation angle of a guided object; similarly, the navigation capacity of the leader is modeled by considering the distance from the guided person, and the action range of the leader is assumed to be a circle;
step two: defining a pedestrian evacuation demand model; pedestrians cannot accurately know a safe and rapid evacuation path due to unfamiliarity with evacuation environments, and need to find a safe exit under the guidance of a leader; generally, a pedestrian always moves to an exit position by following a leader closest to the pedestrian according to the position of the pedestrian with a certain probability;
step three: defining a dynamic model of the leader and the follower; the leader moves according to the evacuation path planned based on the reinforcement learning method, and the movement direction of each time step is known, so that the leader dynamics formula is modified; the follower moves and evacuates along with the leader under the action of the leader, so that a follower dynamics formula is modified;
step four: designing an overall optimization target; in a collaborative relationship in which the evacuation sign and the leader participate in guidance together, dividing the primary and the secondary according to the navigation force; the leader is mainly used as a main guiding means for leading the pedestrians to reach the designated exit, the evacuation sign is used for assisting the pedestrians to find the leader and can provide secondary confirmation of the movement direction for the leader; optimizing the positions and the number of evacuation signs and leaders by considering three parts of signs and human cost, total evacuation time and unsuccessful evacuation penalty terms; and adopting a step-by-step solving method, firstly optimizing the positions and the number of evacuation signs according to the maximum coverage model, secondly, determining the number of the leaders according to the crowd evacuation requirement, comprehensively considering the overall optimization target by combining the positions and the number of the evacuation signs, and solving a feasible solution.
Preferably, the simulation test verification mainly performs experimental verification and validity evaluation on the passenger intelligent guiding strategy; the method specifically comprises the following steps: on the basis of building an actual station physical structure by software, a passenger intelligent guide simulation test platform based on data driving is built; designing a virtual evacuation scene, generating a guidance configuration strategy corresponding to an evacuation path, an evacuation identifier and a leader by using an intelligent guidance strategy generating module, and modeling passengers, evacuation equipment, environment, events and the like; the evacuation effect of the intelligent guidance strategy in various evacuation scenes is previewed through simulation experiments, and the simulation experiment data is subjected to statistical feedback, so that the improvement and optimization of the intelligent guidance strategy generation method are realized.
Preferably, in the fourth step of the leader evacuation path planning based on reinforcement Learning, based on the Q-Learning reinforcement Learning basic framework, the path planning implementation process includes:
the method comprises the following steps: defining a system state set S, discretizing an evacuation scene, and representing the central position state of the people group by using the central point coordinates of each unit; the center point coordinates of the outlet position unit are the target state of the system;
step two: defining an agent action set A, wherein the limited action set A represents the possible motion actions of the agent at the next time step, and the actions can be divided into eight specific directions based on a discretized grid space;
step three: defining a reward return function r, wherein after the agent executes an action each time, the agent can obtain an instant reward and punishment feedback according to the relation with the target state, the feedback reward is a positive value when approaching the target state, and the feedback reward is a negative value when departing from the target state; defining a reward return function r, and considering the distance between the update state of each executed action of the intelligent agent and the outlet, the deviation distance between the update state of each executed action of the intelligent agent and the real video acquisition path and the utilization balance rate of each outlet;
step four: initializing a current state, giving a target state, and initializing a Q matrix;
step five: randomly selecting an action a, calculating feedback rewards and updating the state;
step six: updating the Q value by the following formula;
step seven: updating the policy using a greedy policy;
step eight: and circularly updating until the optimal strategy is output.
Preferably, the position and the number of the evacuation signs are determined by a maximum coverage method in the fourth step of the collaborative guiding strategy based on the evacuation signs and the leader, and the implementation process comprises the following steps:
the method comprises the following steps: determining the total area of the area with the guiding requirement, on one hand, considering the motion path coverage area extracted from the actual video, and on the other hand, considering the evacuation path coverage area generated by a reinforcement learning algorithm;
step two: determining the effective guide range of the single evacuation identification, and simultaneously considering the visual angle and the visual distance of the evacuation identification;
step three: and designing an optimization model, and solving the optimal feasible solution of the placement positions and the number of the evacuation markers on the premise of meeting the maximum coverage of the guidance demand range.
A crowd evacuation system based on multi-carrier intelligent guidance uses any one of the crowd evacuation methods based on multi-carrier intelligent guidance; the system comprises: the system comprises a data acquisition and processing module, an intelligent guide strategy generation module and a simulation test verification module.
Preferably, the data acquisition processing module is used for acquiring and processing the passenger traffic information in the station; the method comprises the steps of collecting passenger motion videos in areas such as a platform, a channel and an exit based on a camera in the station, carrying out image processing on the collected video quantity, extracting passenger position information through a detection tracking algorithm, and calculating passenger motion speed according to video sampling intervals.
Preferably, the intelligent guiding strategy generating module comprises a generating module of a leader evacuation path and an evacuation identifier-leader cooperative guiding strategy generating module; the generation module of the leader evacuation path is used for generating the leader evacuation path, and the evacuation sign-leader cooperative guiding strategy generation module optimizes the evacuation signs and the positions and the number of the leaders on the basis of the leader evacuation path and generates an intelligent guiding strategy corresponding to the current evacuation demand.
Preferably, the simulation test verification module generates a guidance configuration strategy corresponding to the evacuation path, the evacuation sign and the leader by using the intelligent guidance strategy generation module, and models passengers, evacuation equipment, environments, events and the like.
The invention has the beneficial effects that:
(1) the problem of planning the evacuation path of the leader is solved based on reinforcement learning, on one hand, the path planning process has better self-adaptability, and the evacuation path suitable for the dynamically changing environment can be explored and planned in the unknown environment. On the other hand, through video detection and crowd grouping, the state dimension of the reinforcement learning system is effectively reduced, and the calculation speed and the realizability of reinforcement learning training are improved.
(2) On the basis of the intelligent evacuation path generation method, multiple factors such as identification, labor cost and total evacuation time and the collaborative relationship among multiple guidance modes of a station are comprehensively considered, and a collaborative guidance strategy is obtained by designing an optimization model through numerical analysis and calculation. Compared with the evacuation strategy designed based on experience and trial and error in the existing research, the method can more effectively improve the crowd evacuation efficiency. Under the intelligent guide strategy, the configuration of various guide modes of the station has higher flexibility, and the operation cost of the station can be reduced on the premise of ensuring effective safe evacuation.
Drawings
Fig. 1 is a schematic diagram of a crowd evacuation system based on multi-carrier intelligent guidance according to an embodiment of the present invention;
fig. 2 is a flowchart of a crowd evacuation method based on multi-carrier intelligent guidance according to an embodiment of the present invention;
fig. 3 is a schematic frame diagram of a leader evacuation path planning method based on reinforcement learning according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a framework of a cooperative guiding strategy generation method based on evacuation signs and a leader according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a simulation test and verification platform for passenger evacuation according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a simulation process of passenger evacuation according to an embodiment of the present invention.
Detailed Description
The present invention is further described below in conjunction with the following figures and specific examples so that those skilled in the art may better understand the present invention and enable its practice, and the embodiments of the present invention are not limited thereto.
As shown in fig. 1, a crowd evacuation system based on multi-carrier intelligent guidance comprises: the system comprises a data acquisition and processing module, an intelligent guide strategy generation module and a simulation test verification module.
The data acquisition and processing module is mainly used for acquiring and processing the passenger traffic data in the station to form a real evacuation path database. Firstly, shooting and recording of videos are completed, secondly, preprocessing of the videos is completed through an algorithm in a module, and characteristics including passenger position coordinates, movement speed and the like are extracted through a detection tracking algorithm. The finally obtained real path data is used as input data of the intelligent guiding strategy generation module; the intelligent guiding strategy generating module is used for finishing generation of a leader evacuation path and optimization of evacuation indication marks, the number and the positions of the leaders through an internal algorithm under the constraint of real motion data and the actual physical environment of a station, and finally providing a passenger motion model description and guiding strategy setting method for the simulation test verification module; after the simulation test verification module completes the construction of a simulation test platform based on the real station environment, simulation verification is carried out on the guide strategies obtained by the strategy generation module under different simulation initial conditions, the effectiveness of the proposed evacuation guide strategy is verified, and meanwhile, a simulation data result provides support for the optimization and adjustment of the strategy.
As shown in fig. 2, the method for generating the crowd evacuation strategy based on multi-carrier intelligent guidance includes: the method comprises the steps of leader evacuation path planning based on reinforcement learning, cooperative guiding strategy based on evacuation identification and a leader, and simulation test verification.
The leader evacuation path planning based on reinforcement learning comprises four steps:
the method comprises the following steps: and extracting the position coordinates of the pedestrians. And performing video processing on the real station video by taking a frame as a unit, extracting the position coordinates of each frame of pedestrian by adopting a video detection and tracking algorithm, and expressing the movement path of each pedestrian reaching the exit position according to the time sequence.
Step two: and initializing the pedestrian group. And initializing and grouping the pedestrians in the video by adopting a K-Means clustering algorithm.
Step three: the intra-group motion path is normalized. And calculating the weighted average of the motion paths of all the people in the group as the group common motion path.
Step four: and planning the leader path based on reinforcement learning. And modifying a mapping strategy from the state to the action according to the environment state and the return value learned in the last step by adopting a Q-learning algorithm through interaction between the intelligent agent and the environment, so that the action obtains the maximum accumulated return value from the environment, and finally an optimal motion strategy, namely the optimal motion path of the leader is obtained.
The cooperative guiding strategy based on the evacuation identification and the leader comprises four steps:
the method comprises the following steps: an evacuation identity and a leader navigation force model are defined. The actual evacuation guide sign in the subway station is mostly arranged on the surface of a wall and the like, and the guiding capability of the evacuation guide sign is closely related to the distance and the observation angle of a guided object. Similarly, the leader's navigation ability is modeled considering the distance from the guided person, assuming the leader's sphere of action is a circle.
Step two: a pedestrian evacuation demand model is defined. The pedestrian can not accurately know the safe and fast evacuation path due to unfamiliarity with the evacuation environment, and needs to find a safe exit under the guidance of a leader. Generally, the pedestrian always selects to follow the leader closest to the pedestrian to the exit position according to the position with a certain probability.
Step three: a kinetic model of the leader and follower is defined. And the leader moves according to the evacuation path planned by the upper layer based on the reinforcement learning method, and the moving direction of each time step is known, so that the leader dynamics formula is modified. And the followers are driven by the leader to evacuate along with the movement of the leader, so that the follower dynamics formula is modified.
Step four: and designing an overall optimization target. In the cooperative relationship of the evacuation sign and the leader participating in the guidance together, the primary and secondary are divided according to the navigation force. The leader is mainly used as a main guide means for leading the pedestrians to reach the designated exit, the evacuation sign is used for assisting the pedestrians to find the leader, and the sign can provide secondary confirmation of the movement direction for the leader. And optimizing the positions and the number of evacuation signs and the leaders by considering three parts of signs and human cost, total evacuation time and unsuccessful evacuation penalty terms. And adopting a step-by-step solving method, firstly optimizing the positions and the number of evacuation signs according to the maximum coverage model, secondly, determining the number of the leaders according to the crowd evacuation requirement, comprehensively considering the overall optimization target by combining the positions and the number of the evacuation signs, and solving a feasible solution.
And (4) simulation test verification, which is mainly used for carrying out experimental verification and validity evaluation on the passenger intelligent guiding strategy. On the basis of building an actual station physical structure by software, a passenger intelligent guiding simulation test platform based on data driving is built. And designing a virtual evacuation scene, generating a guidance configuration strategy corresponding to an evacuation path, an evacuation identifier and a leader by using an intelligent guidance strategy generating module, and modeling passengers, evacuation equipment, environment, events and the like. The evacuation effect of the intelligent guidance strategy provided by the invention under various evacuation scenes is previewed through simulation experiments, and the simulation experiment data is subjected to statistical feedback, so that the improvement and optimization of the intelligent guidance strategy generation method are realized.
As shown in fig. 3, in the method for generating a crowd evacuation strategy based on multi-carrier intelligent guidance, a leader evacuation path based on reinforcement learning is planned at an upper layer, and the method specifically includes the following steps:
the method comprises the following steps: the method comprises the steps of collecting real moving videos of passengers in the subway station, obtaining higher accuracy and performance by combining a continuous adaptive mean shift (CAMShift) target tracking algorithm with a Kalman filter, extracting crowd parameter information such as position coordinates, moving speed and the like of the passengers in the videos, and forming a moving path database reaching an exit.
Step two: and initializing and grouping the pedestrians in the video by adopting a K-Means clustering algorithm. Firstly, the detection algorithm is usedTo the initial state of the crowd system, n objects S ═ S are included1，S2，S3，...，SnAnd each object has m-dimensional attributes (position, speed magnitude, direction and the like), and the n objects are gathered into specified K class clusters according to the similarity among the objects through a K-Means clustering algorithm. Specifically, k cluster centers { C are initialized1，C2，C3，...，CkAnd k is more than 1 and less than or equal to n, then the Euclidean distance from each object to each cluster center is calculated, each object is divided into all the class clusters with the minimum Euclidean distance, and finally k class clusters { X are obtained1，X2，X3，...，Xk}。
Step three: and calculating the weighted average of the motion paths of all the people in the group as the group common motion path. Each class cluster member initial path may be denoted as Ti 0＝{T1，T2，T3，...，TjI is more than or equal to 1 and less than or equal to k, j is more than or equal to 1 and less than or equal to n, weighting and combining member paths of each class cluster, normalizing a unified motion path of each class cluster,
step four: and modifying a mapping strategy from the state to the action according to the environment state and the return value learned in the last step by adopting a Q-learning algorithm through interaction between the intelligent agent and the environment, so that the action obtains the maximum accumulated return value from the environment, and finally an optimal motion strategy and an optimal motion path of the leader are obtained.
Based on Q-learning reinforcement learning, aiming at an incomplete knowable Markov decision process, an agent with learning ability continuously interacts with the environment in order to realize a moving target, senses the state of the environment, and then takes action and influences the state of the environment. The implementation process of the Q-learning reinforcement-based multi-agent path planning under the reinforcement learning basic framework can be described as the following steps:
the method comprises the following steps: and defining a system state set S, discretizing the evacuation scene, and representing the central position state of the people group by using the central point coordinate of each unit. And the coordinates of the center point of the outlet position unit are the target state of the system.
Step two: defining an agent action set A, wherein the limited action set A represents the possible motion actions taken by the agent at the next time step, and the actions can be divided into eight specific directions based on the discretized grid space,
step three: and defining a reward return function r, wherein after the intelligent agent executes an action each time, the intelligent agent can obtain an instant reward and punishment feedback according to the relation with the target state, the reward is fed back to be a positive value when the intelligent agent is close to the target state, and otherwise, the reward is fed back to be a negative value when the intelligent agent is far from the target state. And defining a reward return function r in the third step, and considering the distance between the update state reached by the agent executing the action each time and the exit, the deviation distance from the real video acquisition path and the utilization balance rate of each exit.
Step four: initializing a current state, giving a target state, and initializing a Q matrix;
step five: randomly selecting an action a, calculating a feedback reward r, and updating the state;
step six: updating the Q value through a formula;
step seven: updating policies with greedy policies
Step eight: and circularly updating until the target state is reached, and outputting the optimal strategy.
The specific parameter steps and parameter descriptions are as follows:
(1) for all S ∈ { S | S1，S2，S3，...，SnH, a ∈ A (S), optionally initializing Q (S, a), given a target state Sgoal；
(2) For each screen, initializing S, and circulating for each step in the screen: selecting a at S using a greedy strategy;
(3) execution of A, observe R, S'
Q(s，a)＝(1-α)Q(s，a)+α{r(s′，a)+λmaxa′∈AQ(s′，a′)-Q(s，a)} (1)
Where s, a represents the current state and action, and s ', a' represents the nextState and action, r (s', a) represents the immediate reward for taking action a in state s, Q (s, a) represents the estimate of path training, maxa′∈AQ (s ', a') represents the optimum path estimation value, α is the learning rate, and λ is the attenuation factor.
(4) Updating a policy pi using a greedy policys←argmaxaQ(s，a)；
(5) Updating is repeated until s is equal to sgoalAnd outputting the optimal strategy pi.
The Q-learning reinforcement learning process is the process by which the agent changes to the final target state by attempting to find the action that yields the greatest benefit in its interaction with the environment. The learning process of finding the optimal strategy can also be seen as a process of finding the maximum value of the reward function. In the invention, the distance d between the update state and the exit, which is reached by the agent in each execution action, is consideredsDeviation distance d from real video capture pathrAnd the utilization balance rate d of each outleteA reward return function r is defined.
r＝w1·ds+w2·dr+w3·de(2)
Wherein, w1，w2And w3Weights of the distance function, path deviation function and exit utilization balance rate function, ds，drAnd deDistance function, path deviation function and exit utilization balance rate function, respectively:
wherein (x)goal，ygoal) Coordinates representing the target position, (x)s，ys) The coordinates representing the current state s.
Wherein (x)s，ys)、(xs′，ys′) And (x)i，yi) Respectively represent the current stateThe state S, the next state S ' and the position coordinates of the intelligent agent in the video data, d represents the position deviation value between the current state S and the next state S ', and d ' represents the position deviation value between the current state S and the motion trail in the real video data of the pedestrian group where the intelligent agent is located.
Wherein
Indicating that the current exit target selected by the agent accumulates the number of people evacuated,
indicating that the cumulative evacuated population for all p exits is summed.
And finishing the evacuation path planning of the leader based on all the steps to obtain the pedestrian simulated intelligent evacuation path of the leader.
As shown in fig. 4, in the method for generating a crowd evacuation policy based on multi-carrier intelligent guidance, a lower layer is based on a cooperative guidance relationship between an evacuation identifier and a leader, and a multi-carrier intelligent guidance policy is generated by comprehensively considering operation cost, evacuation efficiency and overall security in a step-by-step optimization manner, which specifically includes the following steps:
the method comprises the following steps: an evacuation sign and leader navigation force model is defined. The evacuation sign in the embodiment of the invention takes a green rectangular image-text sign fixed on the wall in the station as an example, and the evacuation sign provides reliable evacuation information for passengers. The process of evacuating passengers under the guidance of evacuation signs is divided into three steps: perceptual identification, comprehension identification and decision phase. The size, color, setting height and the like of the evacuation sign are all relevant factors for determining the navigation capability of the evacuation sign. When these relevant factors are determined, the navigation capability of the sign is determined by the perceived probability and the maximum effective evacuation range, and the perceived probability is only determined by the observed distance d of the passengerisAnd an observation angle thetaiThe influence of (c). Defining a perceived probability PsDistance from passenger observationFrom disAnd an observation angle thetaiFunctional relationship f betweens(·)，
Ps＝fs(dis，θi) (6)
The invention assumes that the effective evacuation range of the evacuation sign is a semicircle taking the position as the center of the circle, the radius of the circle represents the visible distance of the evacuation sign, and the visible distance range is [ d ]min，dmax]The effective evacuation range S can be represented by a piecewise function, where S0＝π·dmin 2/2
Defining a variable UsTo characterize whether a certain area is assigned a evacuation identity,
the method considers the cooperative relationship among different guiding modes and divides primary and secondary according to the navigation force. The leader is the main part and can lead the pedestrian to accurately find the designated exit; evacuation signs are secondary, providing timely direction guidance for the pedestrian, and signs may provide secondary confirmation of the direction of movement for the leader. Considering the possibility that the leader deviates from the original evacuation path in the evacuation guidance process, the leader is assumed to continuously correct the deviation of the evacuation path through guidance of the evacuation sign in the guidance evacuation process. The leader is trained and completely rational, moves according to an optimized evacuation path obtained by the upper layer of the strategy, the influence range of the leader is a circular area with a determined radius, and the crowd to be guided to evacuate selects to follow the leader in the group according to grouping conditions until the crowd reaches an exit target position. The invention assumes that the leader evacuation guidance range can cover the area of the group in which the leader evacuation guidance range is positioned, and the probability P that the members in the group are successfully guidedlDistance d from member to leaderilRelated, defining a functional relationship fl(·)。
Pl＝fl(dil) (9)
Defining a variable UlTo characterize whether a region is assigned a leader,
step two: and defining a passenger evacuation demand model, wherein common passengers cannot accurately know a safe and quick evacuation path due to unfamiliarity to an evacuation environment and need to be safely evacuated to an exit under the guidance of a leader. Defining a variable UlmTo characterize whether passenger m is guided by leader 1,
wherein the leader l belongs to {1, 2, 3.,. k }, and the leader m belongs to {1, 2, 3, …, (j-1) · k }.
Step three: defining a leader and a follower dynamic model, wherein the leader moves according to the evacuation path obtained by the strategy layer, meanwhile, the movement deviation is corrected under the action of the evacuation sign, establishing a passenger dynamic model under the influence of the evacuation sign based on the social force model,
wherein m isiIs the quality of the leader or the like,
the acceleration of the leader is represented and,
reflecting the self-driving force of the leader moving according to the established evacuation path, the expected direction of the movement of the leader points to the position of the next motion state in the path,
reflecting the ability of the evacuation sign to communicate evacuation information to the leader,
reflecting the interaction force between people,
reflecting the interaction force between the person and the obstacle.
The follower takes the lead to move, the expected direction of the movement of the follower is consistent with the movement direction of the leader which the follower chooses to follow,
wherein m isiIs the quality of the follower or the like,
indicating the acceleration of the follower, βiTo characterize the degree of willingness of the follower to follow the leader,
reflects the self-driving force of the follower,
reflecting the interaction force between people,
reflecting the interaction force between the person and the obstacle,
reflecting the ability of the leader to communicate evacuation information to the follower.
Step four: and designing an overall optimization target. And optimizing the positions and the number of evacuation signs and the leaders by considering three parts of signs and human cost, total evacuation time and unsuccessful evacuation penalty terms. And (3) adopting a step-by-step solving method, firstly optimizing the positions and the number of evacuation signs according to the maximum coverage model, secondly, determining the number of the leaders according to the crowd evacuation requirement, defining a total optimization target by combining the positions and the number of the evacuation signs, and solving a feasible solution. Optimizing evacuation sign positions and numbers based on the maximum coverage model, defining optimization targets,
min Us(14)
s.t.
∑Us·S≥Ω (15)
∑Us≤Ns(16)
Us＝{0，1} (17)
wherein, Ω represents the total area of the evacuation area to be covered by the evacuation mark, and NsThe upper limit of the number of the evacuation identifications which can be set in the station is represented, and the model aims to reasonably configure the positions of the evacuation identifications under the condition of full coverage of an evacuation area and optimize the number of the identifications.
On the basis of the optimization of the previous step, the overall optimization target is defined by considering the cost, the evacuation time and the safety, namely the unsuccessful evacuation penalty,
min U＝c1∑Us+c2∑Ul+c3∑Ti+c4Φ(T，xi(T)) (18)
s.t.
∑Us·S≥Ω (19)
∑Us≤Ns(20)
∑Ul≤Nl(21)
Us＝{0，1} (22)
Ul＝{0，1} (23)
T＝T0(26)
the objective function represents the selection of an appropriate evacuation strategy to minimize the total cost, taking into account the cost of human labor, the cost of time and the cost of safety, where c1，c2，c3And c4Respectively indicating markIdentifying the weight of cost, manpower cost, time cost and safety cost, wherein the first two costs of the objective function are in direct proportion to the input quantity, the third term is the sum of the evacuation time of all passengers, and each evacuation time of the passengers is determined by a function Ti＝h(xi(t)，vi(t))，t∈[0，T]Determining that the fourth item represents a safety cost and represents that the punishment is not at the specified time T0Interior evacuated passengers, defining a function phi (·)
Wherein d isiIndicating the distance, v, of the passenger from the exitiRepresenting the speed, x, of the passengeri(T) represents the position of the passenger.
In summary, the embodiment of the present invention finally obtains a crowd multi-carrier collaborative guidance method and system based on reinforcement learning, and can generate a guidance evacuation policy pi { U } corresponding to a dynamically changing station evacuation environments，Ul}。
As shown in fig. 5, a simulation test platform is built by taking a typical island platform of a certain station as an example, simulation verification is performed on the guidance strategy obtained by the strategy generation module under different simulation initial conditions, the effectiveness of the proposed evacuation guidance strategy is verified, and meanwhile, a simulation data result provides support for optimization and adjustment of the strategy.
Fig. 6 is a snapshot of a simulation process, and passengers successfully find exits under a cooperative guiding strategy and are orderly and safely evacuated from different exits.
Those of ordinary skill in the art will understand that: the drawings are merely schematic representations of one embodiment, and the flow charts in the drawings are not necessarily required to practice the present invention.
From the above description of the embodiments, it is clear to those skilled in the art that the present invention can be implemented by software plus necessary general hardware platform. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which may be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments or some parts of the embodiments.
Claims (10)
1. A crowd evacuation method based on multi-carrier intelligent guidance comprises a leader evacuation path planning based on reinforcement learning, a collaborative guidance strategy based on evacuation identification and a leader, and simulation test verification.
2. The method for crowd evacuation based on multi-carrier intelligent guidance according to claim 1, wherein the reinforcement learning based leader evacuation path planning specifically comprises the following steps:
the method comprises the following steps: extracting a pedestrian position coordinate; performing video processing on a real station video by taking a frame as a unit, extracting the position coordinates of each frame of pedestrian by adopting a video detection and tracking algorithm, and expressing the motion path of each pedestrian reaching an exit position according to a time sequence;
step two: initializing pedestrian groups; initializing and grouping pedestrians in the video by adopting a K-Means clustering algorithm;
step three: normalizing the motion path in the group; calculating the weighted average of the motion paths of all the people in the group as a group common motion path;
step four: leader path planning based on reinforcement learning; and modifying a mapping strategy from the state to the action according to the environment state and the return value learned in the last step by adopting a Q-learning algorithm through interaction between the intelligent agent and the environment, so that the action obtains the maximum accumulated return value from the environment, and finally an optimal motion strategy, namely the optimal motion path of the leader is obtained.
3. The crowd evacuation method based on multi-carrier intelligent guidance according to claim 1, wherein the collaborative guidance strategy based on evacuation signs and leaders is based on reinforcement learning-based leader evacuation path planning, and specifically comprises the following steps:
the method comprises the following steps: defining evacuation signs and a leader navigation force model; the evacuation guide marks in the actual subway station are mostly arranged on the surfaces such as walls, and the guide capability of the evacuation guide marks is closely related to the distance and the observation angle of a guided object; similarly, the navigation ability of the leader is modeled by considering the distance from the guided person, and the action range of the leader is assumed to be a circle;
step two: defining a pedestrian evacuation demand model; pedestrians cannot accurately know a safe and rapid evacuation path due to unfamiliarity with evacuation environments, and need to find a safe exit under the guidance of a leader; generally, a pedestrian always moves to an exit position by following a leader closest to the pedestrian according to the position of the pedestrian with a certain probability;
step three: defining a dynamic model of the leader and the follower; the leader moves according to the evacuation path planned based on the reinforcement learning method, and the movement direction of each time step is known, so that the leader dynamics formula is modified; the follower moves and evacuates along with the leader under the action of the leader, so that a follower dynamics formula is modified;
step four: designing an overall optimization target; in a collaborative relationship in which the evacuation sign and the leader participate in guidance together, dividing the primary and the secondary according to the navigation force; the leader is mainly used as a main guiding means for leading the pedestrians to reach the designated exit, the evacuation sign is used for assisting the pedestrians to find the leader and can provide secondary confirmation of the movement direction for the leader; optimizing the positions and the number of evacuation signs and leaders by considering three parts of signs and human cost, total evacuation time and unsuccessful evacuation penalty terms; and adopting a step-by-step solving method, firstly optimizing the positions and the number of evacuation signs according to the maximum coverage model, secondly, determining the number of the leaders according to the crowd evacuation requirement, comprehensively considering the overall optimization target by combining the positions and the number of the evacuation signs, and solving a feasible solution.
4. The crowd evacuation method based on multi-carrier intelligent guidance according to claim 1, wherein the simulation test verification mainly performs experimental verification and validity evaluation on passenger intelligent guidance strategies; the method specifically comprises the following steps: on the basis of building an actual station physical structure by software, a passenger intelligent guide simulation test platform based on data driving is built; designing a virtual evacuation scene, generating a guidance configuration strategy corresponding to an evacuation path, an evacuation identifier and a leader by using an intelligent guidance strategy generating module, and modeling passengers, evacuation equipment, environment, events and the like; the evacuation effect of the intelligent guidance strategy in various evacuation scenes is previewed through simulation experiments, and the simulation experiment data is subjected to statistical feedback, so that the improvement and optimization of the intelligent guidance strategy generation method are realized.
5. The crowd evacuation method based on multi-carrier intelligent guidance according to claim 2, wherein the leadership evacuation route planning based on reinforcement Learning in the fourth step is based on a Q-Learning reinforcement Learning basic framework, and the route planning implementation process includes:
the method comprises the following steps: defining a system state set S, discretizing an evacuation scene, and representing the central position state of the people group by using the central point coordinates of each unit; the center point coordinates of the outlet position unit are the target state of the system;
step two: defining an agent action set A, wherein the limited action set A represents the possible motion actions of the agent at the next time step, and the actions can be divided into eight specific directions based on a discretized grid space;
step three: defining a reward return function r, wherein after the agent executes an action each time, the agent can obtain an instant reward and punishment feedback according to the relation with the target state, the feedback reward is a positive value when approaching the target state, and the feedback reward is a negative value when departing from the target state; defining a reward return function r, and considering the distance between the update state of each executed action of the intelligent agent and the outlet, the deviation distance between the update state of each executed action of the intelligent agent and the real video acquisition path and the utilization balance rate of each outlet;
step four: initializing a current state, giving a target state, and initializing a Q matrix;
step five: randomly selecting an action a, calculating feedback rewards and updating the state;
step six: updating the Q value by the following formula;
step seven: updating the policy using a greedy policy;
step eight: and circularly updating until the optimal strategy is output.
6. The crowd evacuation method based on multi-carrier intelligent guidance according to claim 3, wherein in the collaborative guidance strategy based on evacuation signs and leaders, the position and the number of the evacuation signs are determined by a maximum coverage method in step four, and the implementation process comprises:
the method comprises the following steps: determining the total area of the area with the guiding requirement, on one hand, considering the motion path coverage area extracted from the actual video, and on the other hand, considering the evacuation path coverage area generated by a reinforcement learning algorithm;
step two: determining the effective guide range of the single evacuation identification, and simultaneously considering the visual angle and the visual distance of the evacuation identification;
step three: and designing an optimization model, and solving the optimal feasible solution of the placement positions and the number of the evacuation markers on the premise of meeting the maximum coverage of the guidance demand range.
7. A crowd evacuation system based on multi-carrier intelligent guidance uses any one of the crowd evacuation methods based on multi-carrier intelligent guidance; the system comprises: the system comprises a data acquisition and processing module, an intelligent guide strategy generation module and a simulation test verification module.
8. The crowd evacuation system based on multi-carrier intelligent guidance according to claim 7, wherein the data acquisition and processing module is used for acquiring and processing the passenger traffic information in the station; the method comprises the steps of collecting passenger motion videos in areas such as a platform, a channel and an exit based on a camera in the station, carrying out image processing on the collected video quantity, extracting passenger position information through a detection tracking algorithm, and calculating passenger motion speed according to video sampling intervals.
9. The multi-carrier intelligent guidance-based crowd evacuation system of claim 7, wherein the intelligent guidance strategy generating module comprises a leader evacuation path generating module and an evacuation sign-leader cooperative guidance strategy generating module; the generation module of the leader evacuation path is used for generating the leader evacuation path, and the evacuation sign-leader cooperative guiding strategy generation module optimizes the evacuation signs and the positions and the number of the leaders on the basis of the leader evacuation path and generates an intelligent guiding strategy corresponding to the current evacuation demand.
10. The crowd evacuation system based on multi-carrier intelligent guidance according to claim 7, wherein the simulation test verification module generates guidance configuration strategies corresponding to evacuation paths, evacuation signs and leaders by using the intelligent guidance strategy generation module, and models passengers, evacuation equipment, environments, events and the like.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010401446.0A CN111767789B (en) | 2020-05-13 | 2020-05-13 | Crowd evacuation method and system based on multi-carrier intelligent guidance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010401446.0A CN111767789B (en) | 2020-05-13 | 2020-05-13 | Crowd evacuation method and system based on multi-carrier intelligent guidance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111767789A true CN111767789A (en) | 2020-10-13 |
| CN111767789B CN111767789B (en) | 2024-03-26 |
Family
ID=72719151
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010401446.0A Active CN111767789B (en) | 2020-05-13 | 2020-05-13 | Crowd evacuation method and system based on multi-carrier intelligent guidance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111767789B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348285A (en) * | 2020-11-27 | 2021-02-09 | 中国科学院空天信息创新研究院 | Crowd evacuation simulation method in dynamic environment based on deep reinforcement learning |
| CN112862192A (en) * | 2021-02-08 | 2021-05-28 | 青岛理工大学 | Crowd evacuation auxiliary decision-making system based on ant colony algorithm and improved social model |
| CN112884229A (en) * | 2021-02-26 | 2021-06-01 | 中新国际联合研究院 | Large-scale public place people flow guiding path planning method based on differential evolution algorithm |
| CN113935595A (en) * | 2021-09-28 | 2022-01-14 | 北京交通大学 | Urban rail transit road network peak large passenger flow dredging system |
| CN113988408A (en) * | 2021-10-27 | 2022-01-28 | 青岛理工大学 | Multi-objective planning-based artificial guidance method for passenger flow evacuation in subway station |
| CN114792133A (en) * | 2022-06-23 | 2022-07-26 | 中国科学院自动化研究所 | Deep reinforcement learning method and device based on multi-agent cooperation system |
| CN115474172A (en) * | 2022-11-14 | 2022-12-13 | 成都大学 | Indoor dense people stream group pedestrian population evacuation method combined with UWB acquisition |
| CN115527369A (en) * | 2022-09-29 | 2022-12-27 | 北京交通大学 | Large passenger flow early warning and evacuation method under large-area delay condition of airport hub |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103810741A (en) * | 2014-02-19 | 2014-05-21 | 重庆邮电大学 | Underground emergency evacuation virtual crowd simulation method based on multiple intelligent agents |
| KR20140075408A (en) * | 2012-12-11 | 2014-06-19 | 한국전자통신연구원 | Dynamic emergency light system based on swarm robot |
| CN107403049A (en) * | 2017-07-31 | 2017-11-28 | 山东师范大学 | A kind of Q Learning pedestrians evacuation emulation method and system based on artificial neural network |
| CN109101694A (en) * | 2018-07-16 | 2018-12-28 | 山东师范大学 | A kind of the crowd behaviour emulation mode and system of the guidance of safe escape mark |
-
2020
- 2020-05-13 CN CN202010401446.0A patent/CN111767789B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140075408A (en) * | 2012-12-11 | 2014-06-19 | 한국전자통신연구원 | Dynamic emergency light system based on swarm robot |
| CN103810741A (en) * | 2014-02-19 | 2014-05-21 | 重庆邮电大学 | Underground emergency evacuation virtual crowd simulation method based on multiple intelligent agents |
| CN107403049A (en) * | 2017-07-31 | 2017-11-28 | 山东师范大学 | A kind of Q Learning pedestrians evacuation emulation method and system based on artificial neural network |
| CN109101694A (en) * | 2018-07-16 | 2018-12-28 | 山东师范大学 | A kind of the crowd behaviour emulation mode and system of the guidance of safe escape mark |
Non-Patent Citations (1)
| Title |
|---|
| 周敏; 董海荣: "《平行应急疏散系统_基本概念、体系框架及其应用》", 《自动化学报》, pages 1 - 13 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348285B (en) * | 2020-11-27 | 2021-08-10 | 中国科学院空天信息创新研究院 | Crowd evacuation simulation method in dynamic environment based on deep reinforcement learning |
| CN112348285A (en) * | 2020-11-27 | 2021-02-09 | 中国科学院空天信息创新研究院 | Crowd evacuation simulation method in dynamic environment based on deep reinforcement learning |
| CN112862192A (en) * | 2021-02-08 | 2021-05-28 | 青岛理工大学 | Crowd evacuation auxiliary decision-making system based on ant colony algorithm and improved social model |
| CN112884229A (en) * | 2021-02-26 | 2021-06-01 | 中新国际联合研究院 | Large-scale public place people flow guiding path planning method based on differential evolution algorithm |
| CN113935595B (en) * | 2021-09-28 | 2023-07-28 | 北京交通大学 | Urban rail transit road network peak large passenger flow dredging system |
| CN113935595A (en) * | 2021-09-28 | 2022-01-14 | 北京交通大学 | Urban rail transit road network peak large passenger flow dredging system |
| CN113988408A (en) * | 2021-10-27 | 2022-01-28 | 青岛理工大学 | Multi-objective planning-based artificial guidance method for passenger flow evacuation in subway station |
| CN113988408B (en) * | 2021-10-27 | 2024-04-09 | 青岛理工大学 | Manual induction method for passenger flow evacuation in subway station based on multi-objective planning |
| CN114792133A (en) * | 2022-06-23 | 2022-07-26 | 中国科学院自动化研究所 | Deep reinforcement learning method and device based on multi-agent cooperation system |
| CN114792133B (en) * | 2022-06-23 | 2022-09-27 | 中国科学院自动化研究所 | Deep reinforcement learning method and device based on multi-agent cooperation system |
| CN115527369A (en) * | 2022-09-29 | 2022-12-27 | 北京交通大学 | Large passenger flow early warning and evacuation method under large-area delay condition of airport hub |
| CN115474172B (en) * | 2022-11-14 | 2023-01-24 | 成都大学 | Indoor dense people stream group pedestrian population evacuation method combined with UWB (ultra Wide band) acquisition |
| CN115474172A (en) * | 2022-11-14 | 2022-12-13 | 成都大学 | Indoor dense people stream group pedestrian population evacuation method combined with UWB acquisition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111767789B (en) | 2024-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111767789B (en) | Crowd evacuation method and system based on multi-carrier intelligent guidance | |
| CN106127120B (en) | Posture estimation method and device, computer system | |
| CN105892480B (en) | Isomery multiple no-manned plane systematic collaboration, which is examined, beats task self-organizing method | |
| AlMahamid et al. | Autonomous unmanned aerial vehicle navigation using reinforcement learning: A systematic review | |
| Zhang et al. | IoT-based positive emotional contagion for crowd evacuation | |
| CN109543285B (en) | Crowd evacuation simulation method and system integrating data driving and reinforcement learning | |
| Tang et al. | A joint global and local path planning optimization for UAV task scheduling towards crowd air monitoring | |
| CN112215350A (en) | Smart agent control method and device based on reinforcement learning | |
| Ma et al. | An intelligence-based approach for prediction of microscopic pedestrian walking behavior | |
| CN110956684B (en) | Crowd movement evacuation simulation method and system based on residual error network | |
| CN111414681A (en) | In-building evacuation simulation method and system based on shared deep reinforcement learning | |
| CN112862192A (en) | Crowd evacuation auxiliary decision-making system based on ant colony algorithm and improved social model | |
| Leong et al. | An intelligent auto-organizing aerial robotic sensor network system for urban surveillance | |
| Ghosh et al. | A deep ensemble method for multi-agent reinforcement learning: A case study on air traffic control | |
| CN107623590A (en) | The collocation method and system monitored in real time for unmanned aerial vehicle group | |
| Sai et al. | A comprehensive survey on artificial intelligence for unmanned aerial vehicles | |
| Li et al. | When digital twin meets deep reinforcement learning in multi-UAV path planning | |
| CN112121419B (en) | Virtual object control method, device, electronic equipment and storage medium | |
| Xing et al. | Policy distillation with selective input gradient regularization for efficient interpretability | |
| CN112991544A (en) | Group evacuation behavior simulation method based on panoramic image modeling | |
| Zhu et al. | Computer application in game map path-finding based on fuzzy logic dynamic hierarchical ant colony algorithm | |
| Karnchanachari et al. | Towards learning-based planning: The nuPlan benchmark for real-world autonomous driving | |
| Leong et al. | Integrated perception and tactical behaviours in an auto-organizing aerial sensor network | |
| CN114548497A (en) | Crowd movement path planning method and system for realizing scene self-adaption | |
| CN114326826A (en) | Multi-unmanned aerial vehicle formation transformation method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |