CN109727519A - A kind of large-scale virtual crowd analogue system - Google Patents
A kind of large-scale virtual crowd analogue system Download PDFInfo
- Publication number
- CN109727519A CN109727519A CN201811504286.1A CN201811504286A CN109727519A CN 109727519 A CN109727519 A CN 109727519A CN 201811504286 A CN201811504286 A CN 201811504286A CN 109727519 A CN109727519 A CN 109727519A
- Authority
- CN
- China
- Prior art keywords
- crowd
- module
- behavior
- triangulation network
- analogue system
- 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.)
- Pending
Links
Abstract
The invention discloses a kind of large-scale virtual crowd analogue systems, module is realized in crowd behaviour model emulation module, ambient process module and behavior, wherein: crowd behaviour model emulation module, for creating crowd's sensor model of region perception and virtual sense organ composition, external information is obtained for crowd, behaviour decision making is carried out and technical support is provided;Ambient process module renders navigation grid for carrying out calculation system to static environment model;Behavior realizes that module exports behavior by logic judgment for handling the information input of crowd.The present invention provides a kind of large-scale virtual crowd analogue system, when it is applied after by the emergency drilling simulation training system of virtual crowds simulation technical application to subway, high-speed rail platform, pass through the accumulation of application experience and going deep into for research, the application of the technology extends also to escape truck design and emergency plan optimization of the public field building such as high-speed rail, subway etc., has huge Social benefit and economic benefit.
Description
Technical field
The present invention relates to simulation/emulation technologies, and in particular to a kind of large-scale virtual crowd analogue system.
Background technique
With the fast development of public transport cause, the track vehicles such as high-speed rail, subway have been increasingly becoming people and have gone out
Capable important selection, while in recent years, the safety of public place and emergency processing by social extensive concern, therefore high-speed rail,
The administrative departments at different levels of iron also gradually establish the contingency management body of the public places such as high-speed rail station, the subway station more standardized
System.But due to its emergency processing flow be related to each department such as scheduling, station, service on buses or trains, driver, passenger traffic and work post and 110,
119 equal social emergency management and rescue mechanisms, emergency management and rescue tissue difficulty is big, and co-ordination is complicated, therefore, in order to guarantee to face burst
When event occurs, each work post/department can accomplish calm reply, coordinate tacit agreement, daily emergency management and rescue training and each work post
Between coordinated training be just particularly important.
Because being related to department and the factors such as personnel are more, and tissue difficulty is big, and place limits, at present on the whole, emergency drilling
Still lack safety, economy, efficient emergency preplan implementation rehearsal means, ways and means, the rehearsal of emergency preplan is in mostly
The modes such as theory teaching examination, oral enquirement, desktop maneuvers carry out, and effect is limited.
In recent years, with the fast development of computer virtual simulation and VR technology, computer virtual simulation technology is also gradually
The various aspects such as design, manufacture, operation, the training of field of track traffic are applied to, such as buildings model BIM application, train simulation
Pilot instrument, in particular, Train Operation Simulator device has been widely applied in the training of driver and conductor.Virtual crowds simulation
And it visualizes technology and has been gradually applied to public place due to having feature that is intuitive, safe, can interacting, repeat
Design rationalization analysis with decision in, such as skyscraper, commercial center, conference and exhibition center's heavy construction design optimization,
But high-speed rail, subway station emergency processing in temporarily without mature technology and case.At present virtual crowds simulation technology there is also
The problems such as algorithm is complicated, memory consumption is big, there is also biggish rooms for promotion in terms of real-time, group's vivid effect.This class
Concrete application of the topic for public places emergency processings such as subway station, high-speed rail stations, the key of research large-scale crowd emulation
Technology, including crowd's information Perception system, group behavior simulation model, magnanimity crowd's Real-time Rendering Technology based on GPU etc., and
Trial will be in large scale emulation technical application to subway, the emergency drilling analogue system of high-speed rail.
Summary of the invention
The present invention is in view of the above-mentioned problems, provide a kind of large-scale virtual crowd analogue system, by visual human when application
Group's manufactaring to subway, high-speed rail platform emergency drilling simulation training system in after, by the accumulation of application experience and
Research is goed deep into, and the application of the technology extends also to the escape truck design and emergency of the public places such as high-speed rail, subway building
Scheme optimization etc. has huge Social benefit and economic benefit.
The present invention is achieved through the following technical solutions:
A kind of large-scale virtual crowd analogue system, including crowd behaviour model emulation module, ambient process module and row
To realize module, in which:
The ambient process module renders navigation grid, creates ring for carrying out calculation system to static environment model
Border model;
The crowd behaviour model emulation module, the crowd for creating region perception and virtual sense organ composition perceive mould
Type, crowd's sensor model obtain external information from the environmental model, realize that module carries out behaviour decision making and mentions for the behavior
For information input;
The behavior realizes that module passes through logic for handling the information input of the crowd behaviour model emulation module
Judge output behavior.
Further, a kind of large-scale virtual crowd analogue system, the crowd behaviour model emulation module include synthesis
Perceptron, signal filter and decision behavior module, synthesis perceptron obtain all information in environment, the letter for crowd
Breath filter is used to be filtered to obtain to all information information needed, the decision behavior module be used for by information needed into
Row behaviour decision making.
Further, a kind of large-scale virtual crowd analogue system, the synthesis perceptron include environmental area perception mould
Block and sensory perception's module.
Further, a kind of large-scale virtual crowd analogue system, the environmental area sensing module include intelligent body, penetrate
Line transmitter and collision body with specific label, the position for needing to carry out region perception is arranged in the collision body, described to penetrate
Line transmitter is arranged in the location of intelligent body.
Further, a kind of large-scale virtual crowd analogue system, the mistake of the ambient process module rendering navigation grid
Journey are as follows: establish one and contain the triangulation network of all areas of feasible solutions of environment space, obtain being walked by what optimum triangular shape formed
The region triangulation network is settled accounts by the node to the triangulation network, can obtain global optimum path with faster speed.
Further, a kind of large-scale virtual crowd analogue system, the establishment process of the triangulation network are as follows: define a packet
Enclose the initial polygon of all discrete datas;For the initial polygon, an initial triangulation network is established;It will be in discrete point set
First point P insertion initial triangulation in, find out in the triangulation network include P point triangle T, three vertex phases of P and T
Even, a new triangulation network is generated;Triangles optimization is carried out using empty circumscribed circle characteristic, obtains the newest triangulation network;In repetition
Step is stated, until all the points into discrete point set are all gradually added in the triangulation network;Delete the correlation comprising initial bounding box vertex
Triangle forms the final triangulation network.
Further, a kind of large-scale virtual crowd analogue system, obtains the walkable region being made of optimum triangular shape
The detailed process of the triangulation network are as follows: by merging convex polygon to the triangulation network, deleting extra side, ignore small surface grids etc.
Reason, so that convex polygon quantity, number of edges amount navigation grid as few as possible are obtained, to reduce the calculation amount sum number during pathfinding
According to memory space, merges mainly comprising the processes of for chimb and divided edge triangles since triangular mesh;Merge convex polygon
Shape and adjacent triangle guarantee that the triangle after merging remains as convex polygons;Repeat above-mentioned two step, until second step not
It can continue to execute.
Further, a kind of large-scale virtual crowd analogue system, the behavior realize that module is exported by logic judgment
Behavior include route searching and dynamic collision prevention, behaviour decision making and execution.
Further, a kind of large-scale virtual crowd analogue system, the behavior realize that module carries out route searching and passes through
Navigation grid route searching method, for navigation grid route searching method by continuous estimated path, finding can be such that the shortest point in path makees
For next node, until arriving at terminal.
Compared with prior art, the present invention having the following advantages and benefits:
1, by virtual crowds simulation technical application to subway, the emergency drilling emulation training of high-speed rail platform when the present invention applies
In system, by the accumulation of application experience and going deep into for research, the application of the technology extends also to the public field such as high-speed rail, subway
Escape truck design and emergency plan optimization for being built etc., have huge Social benefit and economic benefit.
2, the present invention by virtual crowds simulation and visualizes technical application to high-speed rail, the station emergency processing of subway
In, the problems such as it is more to solve high-speed rail, subway department and personnel, and tissue difficulty is big, and place limits, provides a kind of emergency drilling still
Lack safety, economy, efficient emergency preplan implementation rehearsal means, ways and means.
3, when can guarantee that facing emergency event occurs when present invention application, each work post/department can accomplish calm answer
It is right, coordinate tacit agreement, the coordinated training between daily emergency management and rescue training and each work post ensures the person and property peace
Entirely.
Detailed description of the invention
Attached drawing described herein is used to provide to further understand the embodiment of the present invention, constitutes one of the application
Point, do not constitute the restriction to the embodiment of the present invention.In the accompanying drawings:
Fig. 1 is the workflow schematic diagram that perceptron is synthesized in the present invention;
Fig. 2 is the simulated experiment figure of environmental area sensing module in the present invention;
Fig. 3 is sensory perception's modular structure schematic illustration in the present invention;
Fig. 4 is the simulated experiment figure of sensory perception's module in the present invention;
Fig. 5 is the structural schematic diagram of ambient process module intermediate cam net in the present invention;
Fig. 6 is the triangle decomposition point insertion process schematic diagram of ambient process module intermediate cam net in the present invention;
Fig. 7 is the triangulation navigation grid schematic diagram of ambient process module intermediate cam net in the present invention;
Fig. 8 be in the present invention ambient process module close to optimal navigation grid schematic diagram;
Fig. 9 is that navigation grid renders schematic diagram in the present invention;
Module realizing route smoothing processing schematic diagram is realized in Figure 10 behavior of the present invention;
Figure 11 is that RVO algorithm schematic diagram in module is realized in behavior of the present invention;
Figure 12 is the collision prevention schematic diagram that different weights and equal weight in module are realized in behavior of the present invention;
Figure 13 is passenger of the present invention simulation result schematic diagram out of the station.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below with reference to embodiment and attached drawing, to this
Invention is described in further detail, and exemplary embodiment of the invention and its explanation for explaining only the invention, are not made
For limitation of the invention.
Embodiment 1
A kind of large-scale virtual crowd analogue system, including crowd behaviour model emulation module, ambient process module and row
To realize module, in which: the ambient process module renders Navigation Network for carrying out calculation system to static environment model
Lattice create environmental model;The crowd behaviour model emulation module, for creating the crowd of region perception and virtual sense organ composition
Sensor model, crowd's sensor model obtain external information from the environmental model, realize that module carries out behavior for the behavior
Decision provides information input;Module is realized in the behavior, for handling the information input of the crowd behaviour model emulation module,
Behavior is exported by logic judgment.
The crowd behaviour model emulation module includes synthesis perceptron, signal filter and decision behavior module, synthesis
Perceptron obtains all information in environment for crowd, and the signal filter is for being filtered all information to obtain institute
Information is needed, the decision behavior module is used to carry out behaviour decision making by information needed.The synthesis perceptron includes ambient zone
Domain sensing module and sensory perception's module.The environmental area sensing module includes intelligent body, ray emitter and has specific
The position for needing to carry out region perception is arranged in the collision body of label, the collision body, and the ray emitter is arranged in intelligence
The location of body.
The ambient process module renders navigation grid for carrying out calculation system to static environment model;The ring
The process of border processing module rendering navigation grid are as follows: establish one and contain the triangulation network of all areas of feasible solutions of environment space, obtain
To the walkable region triangulation network being made of optimum triangular shape, settled accounts by the node to the triangulation network, it can be with faster speed
Obtain global optimum path.The establishment process of the triangulation network are as follows: define the initial polygon for surrounding all discrete datas;
For the initial polygon, an initial triangulation network is established;First point P in discrete point set is inserted into initial triangulation
In, the triangle T in the triangulation network comprising P point is found out, P is connected with three vertex of T, generates a new triangulation network;It utilizes
Empty circumscribed circle characteristic carries out triangles optimization, obtains the newest triangulation network;It repeats the above steps, until all into discrete point set
Point is all gradually added in the triangulation network;The associated triangle comprising initial bounding box vertex is deleted, the final triangulation network is formed.It obtains
By the detailed process for the walkable region triangulation network that optimum triangular shape forms are as follows: by the triangulation network merge convex polygon,
Extra side is deleted, ignores small surface grids etc. reason, so that convex polygon quantity, number of edges amount navigation grid as few as possible are obtained,
To reduce the calculation amount and data space during pathfinding, merge mainly comprising the processes of from triangular mesh for chimb
Edge triangles start to divide;Merge convex polygon and adjacent triangle, guarantees that the triangle after merging remains as convex polygons;
Above-mentioned two step is repeated, until second step cannot continue to execute.
The behavior realizes that module exports behavior by logic judgment for handling the information input of crowd.The behavior
Realization module includes route searching and dynamic collision prevention, behaviour decision making and execution by the behavior that logic judgment exports.The behavior
Realize that module carries out route searching by navigation grid route searching method, navigation grid route searching method is by constantly estimating road
Diameter, finding can make the shortest point in path as next node, until arriving at terminal.
Embodiment 2
Based on embodiment 1, simulation process of the present invention are as follows:
The first step, crowd's sense of crowd behaviour model emulation, the mainly perception of creation creation region and virtual sense organ composition
Perception model obtains external information for crowd, carries out behaviour decision making and provides technical support.When crowd is in virtual environment, how to feel
Know the environmental information of surrounding, such as barrier, pedestrian, elevator, fire scene, and the information perceived is made accordingly
Feedback.
Perception is intelligent body to the cognition of environment and provides itself behavior of behaviour decision making condition.In general, crowd feels
Know that needs provide following three functions:
1, region sensing capability.Crowd judges the ring locating for itself by the perception identification to object each in environment
Border.It is in square, in supermarket or in clothes shop etc., certain behaviors of crowd can only occur in spy such as the crowd in market
In fixed environment.Real-life crowd can judge locating environment by cognition of oneself etc., and real in this way
Existing perception of the crowd in virtual environment needs to be related to many contents of artificial intelligence technology, and Rule of judgment is complicated, and system is realized
It is difficult.Unlike reality scene, in emulation platform, the information of static virtual environment is known.Therefore we can be with
Static environment is pre-processed, the environmental area needed for every kind provides specific label, and intelligent body is only needed in a certain area
When domain, the information in the area label need to be only read, so that it may obtain information needed.
2, surrounding objects are perceived.To the ability that the object of surrounding is perceived, for example the pedestrian of surrounding is perceived, to keep away
Exempt to collide, the chair for perceiving surrounding can go to sit down automatically.Perception of the crowd to surrounding objects, be realize dynamic collision prevention,
One of the premise of careful Behavior modeling.In the relatively high system of validity and complexity, perception of the crowd to surrounding objects
Very more Behavior modelings may be implemented.Such as in virtual traffic system, crowd perceives the position of fire, to flee from fire hair
Raw point;Passenger on vehicle invites guests to be seated for the elderly, when can wait road when pedestrian goes across the road without vehicle after etc..
3, decision itself behavior.The final purpose of crowd's perception is for the behavior of decision itself, by perception information
Processing, determine the behavior of next step.In the present system, crowd behaviour is driven by perception and purpose, is perceived for decision
Influence mainly there are two aspect: change final purpose and stage sexual behaviour decision condition be provided.
By perceiving the analysis of demand to crowd above, devise herein a kind of by environmental area perception and sensory perception's group
At synthesis perceptron.As shown in Figure 1, crowd by synthesis perceptron obtain environment in all information, to these information into
Information needed is obtained by filtration in row, and information needed is used for behaviour decision making.Its workflow is as follows:
Region sensing module is collided by ray and is realized.It is had first to need the part for carrying out region perception to add in environment
The collision body of specific label, intelligent body are obtained locating for itself with the location of itself divergent-ray by the collision body of return
The relevant information of environment.
It is illustrated in figure 2 the experiment of region perception, the circle on the left side is intelligent body, sequentially enters region 1, region 2, area
Domain 3, region 4.
As shown in figure 3, sensory perception's module is synthesized by vision and the sense of hearing, it is mainly responsible for the people of perception surrounding dynamic motion
And object relevant to behavior triggering, such as the chair in scene.Vision is made of a fan-shaped region, and the sense of hearing is then by one
Circular region composition.Crowd can have the generation of sound during movement, talk etc. in reality scene, and crowd can lead to
Cross the people that surrounding is perceived to the judgement of sound.It is calculated to simplify, it is considered herein that auditory system can ignore barrier, inspection
All people in audibility range are measured, and there will be the effects of a barrier obstruction for vision system.Barrier obstruction passes through several
What algorithm is realized, for determining visual human to packet respectively using classical visibility processing method in object within the vision
Whether the sight and other object bounding box rectangles for enclosing eight vertex of box have intersection point, carry out the hiding relation of decision space.By vision
Perceptron with sense of hearing synthesis is as shown, wherein R2Optimal viewing angle distance is indicated, since the sight of people in actual life is ratio
It is farther away, but for the thing of distant place, perception information will not influence the behaviour decision making of people mostly, therefore herein set the distance
It is set to 8m.A indicates the angular field of view of people, and according to ergonomic theory, a is about 110 degree.R1Indicate audibility range, the present invention
It is set as 3 meters.
The experimental result of the sensor model is as shown in Figure 4.Main body A therein random walk in the scene is only both regarded in A
Feel that the people in range again in A audibility range can become black.
When intelligent body is when clear, all objects in synthesis sensing range can be freely perceived, and
When there are barrier, the visual perception of intelligent body is affected, and can not perceive the object by wall block vision, but the sense of hearing
Perception is not affected then, still can perceive the object being obscured by an obstacle but within the scope of Auditory Perception.
Second step, ambient process module are mainly responsible for and carry out calculation system to constant static environment model, that is, navigate
Grid rendering.Navigation grid generation is a kind of common terrain information processing technique, in the present invention for obtaining by optimal three
The walkable region triangulation network of angular composition is settled accounts by the node to the triangulation network, can obtain intelligent body with faster speed
Global optimum path (no dynamic barrier).
Geographic information processing in virtual scene for a long time, is all the hot issue in scene simulation, is to have path
The basic function of the artificial intelligence body indispensability of search and dynamic collision prevention function.Processing of the early stage for ground environment, generally by it
It is divided into square or hexagonal mesh, optimal path is calculated by A* algorithm.A* algorithm has been proved to most effective road
Path search algorithm can not optimize to obtain time, the lower algorithm of space complexity again.But by map partitioning at square or
The mode of six warp mesh has very big advantage, but there is also very big drawbacks.This grid processing method is simple, is easy to
It realizes, map lesser for area is used very much, but in the big map of area, it may be desirable to which the grid of ten thousand orders of magnitude causes on road
It will appear a large amount of calculating during path search, the storage of gridding information can also occupy a large amount of memories, be difficult to use in real-time calculating
In.It solves the problems, such as this key, exactly reduces the quantity of cell in grid, navigation grid can complete this well
Business.
In virtual environment, we can assume that intelligent body can walk road, floor, the wall that cannot be walked, desk etc.,
With the region that can be walked in simple, the rough geometric figure definition space for covering surface region that is spacious, can walking about,
This geometric figure is referred to as navigation grid (navigationmesh).
The foundation of navigation grid needs to defer to some geometric graphics rules.As composed by three lines in space
Region centainly in a plane, therefore navigation grid generally by triangle sets at.Navigation grid is defined in environment and can be gone
The surface information walked, when carrying out pathfinding calculating, all feasible points necessarily all are in navigation grid, therefore avoid ring automatically
Static-obstacle thing in border.Meanwhile three dimensional environmental space can be calculated and converted by the calculating of feasible height, the feasible gradient
For two-dimensional surface calculating, huge expense in three-dimensional computations is avoided.
The key for establishing navigation grid is to establish one to contain the triangulation network of all areas of feasible solutions of environment space.According to
Knowledge in Cartography and Geography Information System, the triangulation network being made of irregular triangle show non-in various geometrical characteristics
Normal good property can maximize the appearance for avoiding long and narrow triangle, therefore present invention selection is established initially using the triangulation network
Navigation grid.
The triangulation network passes through triangulation by the point in scene, side, convex shape changeable information and handles to obtain.According to the triangulation network
Empty circumscribed circle characteristic, i.e. the circumscribed circle of arbitrary triangle do not include other points in addition to the Atria vertex, can be very
Easily judge whether a triangulation network is the triangulation network.As shown in figure 5, two kinds of triangulation networks that tetra- discrete points of A, B, C, D are formed,
The left side is the common triangulation network, the triangulation network on the right.
It for any number of discrete point, is divided anyway, the obtained triangulation network is the same.The present invention adopts
Triangulation is carried out with incremental inserting algorithm, as shown in fig. 6, the algorithm is divided into the following steps:
1, the initial polygon for surrounding all discrete datas is defined;
2, for the initial polygon, an initial triangulation network is established;
3, by first point P insertion initial triangulation in discrete point set, the triangle in the triangulation network comprising P point is found out
T is connected P with three vertex of T, generates a new triangulation network;
4, triangles optimization is carried out using empty circumscribed circle characteristic, obtains the newest triangulation network;
5, it repeats the above steps, until all the points into discrete point set are all gradually added in the triangulation network;
6, the associated triangle comprising initial bounding box vertex is deleted, the final triangulation network is formed.
By carrying out triangulation to discrete point, environmental geography information is handled, available complex environment is located
Triangle gridding after reason obtains as shown in fig. 7, carrying out triangulation to information such as the discrete of subway platform, side, convex polygons
The triangulation network, wherein black portions indicate barrier.
By carrying out convex polygon merging, deleting extra side, ignore small surface grids etc. reason to triangle navigation grid, from
And convex polygon quantity, number of edges amount navigation grid as few as possible are obtained, it is deposited with reducing calculation amount during pathfinding and data
Store up space.Merge mainly comprising the processes of for chimb
1, it is divided edge triangles since triangular mesh;
2, merge convex polygon and adjacent triangle, guarantee that the triangle after merging remains as convex polygons;
3, above-mentioned two step is repeated, until second step cannot continue to execute.
Optimize the triangulation network by this method, obtains convex polygon quantity, the less grid of number of edges amount, as shown in Figure 8.Storage
Each convex polygon information (apex coordinate, adjacent polygon), the information (endpoint, place convex polygon) of each edge, for pathfinding
It uses.
By above-mentioned treatment process, three-dimensional scenic can be rendered to the navigation grid being made of convex polygon plane.It will
The convex polygon of navigation grid, the information on side are stored in MeshData, for obtaining in real time when passenger's pathfinding.The present invention is tested
In subway scene carry out grid rendering, stair are replaced using the close plane of gradient, and rendering result is as shown in Figure 9.
Module is realized in third step, behavior, is responsible for the information input of processing crowd, exports behavior by logic judgment, including
Route searching and dynamic collision prevention, behaviour decision making and execution.
One of the basic capacity that intelligent body in virtual environment has is exactly searching for and avoid collision automatically dynamically for path
Barrier.Algorithm about route searching has had many years, at present mainly by static state search and two kinds of strategies of News Search.
Static path planning is the space that will can not be moved in environment as static state, obtains the path for avoiding all static-obstacle things, so
Carry out the processing of avoiding obstacles again when encountering dynamic barrier afterwards.Active path planning is then according to dynamic object in environment
The change of position, real-time perfoming optimum path planning, for example dynamic is carried out using the Real-time Traffic Information obtained in auto navigation
Path planning.Since active path planning is influenced by other moving object positions, real-time perfoming route searching is needed to calculate,
When the dynamic object in environment is more, this method can occupy huge computing cost, therefore current crowd simulation model
In mostly all by the way of static path planning.Path search algorithm is calculated based on the node in A* algorithm, only to map
Node has different processing modes.For lesser map, map is simply generally divided into the small lattice being made of square, intelligence
Body occupies one or more grid, and moving step sizes are a lattice.For large-scale map, path search algorithm is broadly divided into path grid
Lattice method (path lattice) and navigation grid (NavMesh) generate two kinds.The former is put in map using visible dots structured approach
Key point is set, line between key point, in searching route, the shortest route for finding out initial position and terminal is shortest path
Diameter.But this method has significant limitation, and placement problem including path node, frequently generates low quality at representation method problem
Movement routine problem etc., and when navigation spots are more, square to increase again, the time of algorithm is multiple for wiring quantity between points
Miscellaneous degree and space complexity are N2.Navigation grid is then a convex polygon geometry, and all in storage 3D scene " can walk table
The information in face ", even large area scene can also be completely represented with less grid, this is a kind of very powerful road AI
Path search method and the present invention are the method that pedestrian's pathfinding uses.
A* algorithm is a kind of heuristic search algorithm for combining depth-first search and breadth first search, which exists
There is high efficiency in static path search, have proven to best algorithm, be difficult to be promoted by advanced optimizing
Speed.
The essence of A* algorithm is exactly to pass through continuous estimated path, and finding can make the shortest point in path as next node, directly
To arrival terminal.A* algorithm calculates the expression of the evaluation function f (n) in path are as follows:
F (n)=g (n)+h (n)
Wherein g (n) indicates search A node to the distance of starting point, and middle h (n) expression search A node, should to the distance of terminal
Value is the estimated value for ignoring barrier, is all the barrier ignored between the point and terminal when calculating the estimated value, and calculating should
Point arrives the linear distance of terminal, or calculates the sum of each coordinate difference of the point and terminal.Terminal is calculated simultaneously with starting point, it can be with
Avoid the calculating of a large amount of invalid nodes.During calculating g (n), the walking difficulty coefficient of node can be set, it is such as flat
Path coefficients is smaller, and muddy path coefficient is larger, to keep path computation result more reasonable.
Specific search process is from the off, successively to calculate f (n) value of each feasible neighbors, and f (n) value is minimum
Node as present node, be placed on node and close in list, remaining neighbors, which is placed on, to be drawn up in table.Present node is settled accounts every time
F (n) value, and the smallest point of f (n) value is found out in table as present node, until for terminal or drawing up table to present node from drawing up
For empty (path can not be found).
The core of A* algorithm is exactly constantly to calculate f (n) value of each node, in the navigation grid formed with convex polygon
In, node is the midpoint on each side, the advantage of doing so is that can walk in pathfinding all along side to avoid intelligent body.With
For triangle, each triangle has tri- nodes of A, B, C, it is assumed that previous node is A point, then the evaluation function f (n) of B point are as follows:
F (n)=g (n)+h (n)
G (n)=g (n-1)+k (n)
Wherein g (n-1) the i.e. g value of node A, k (n) are A point at a distance from B point, and h (n) is equal to B point to the distance of terminal.
By calculating all nodes passed through, it is sequentially connected from starting point, as optimal path.
The optimal path obtained in this way can exist curved due to being advanced along the midpoint of convex polygon
The problem of folding, zigzag, path is unsmooth.Smooth path is obtained using the method for line-of-sight detection, i.e., whenever intelligent body reaches one
When a approach point, just look forward remaining approach point, creates one from current point to the straight line of the point of approach below, utilizes computer
Knowledge in graphics, can easily judge whether the line intersects with solid side very much, if do not intersected with solid side, then it is assumed that
The point is visible point, if intersecting with solid side, then it is assumed that is changed the time as invisible point.Detection range is changed the time farthest visible transit point
Line can skip these visible approach points, so that several sections of broken line distances of front become straight line, then from changing the time out
Begin to carry out visible point detection, until terminal is visible point, finally obtains a very smooth path.As shown in Figure 10,
For path obtained after smoothing processing.
Pedestrian automatically can avoid touching with static-obstacle thing in virtual environment in such a way that navigation grid generates
It hits, also needs to carry out the processing of dynamic barrier collision prevention when encountering dynamic barrier (such as other intelligent bodies).The present invention makes
The collision prevention to dynamic disorder is realized, while for the row under different conditions as shown in figure 11 with opposite barrier (RVO) algorithm
People designs different collision prevention weight factors, to realize the interspersed detour characteristic of pedestrian.
Assuming that two pedestrians A and B in scene, are reduced to a cylindrical body for pedestrian's approximation, radius is respectively RA、RB,
Position is respectivelyRespectively withWithSpeed movement.Judge whether the two pedestrians can occur during the motion
Collision, according to the principle of relative motion, in the coordinate system that pedestrian's B speed is always 0, then the phase of pedestrian A in this coordinate system
To speedAre as follows:
The speed of pedestrian BAre as follows:
Regard pedestrian A as particle that a radius is 0, the impact radius of pedestrian B is correspondingly puffed to RBR, value size
Are as follows:
RBR=RA+RB
Make tangent line, the size of tangent line and the angle α of AB line from A point are as follows:
A spatial dimension C is obtained, when the speed of related movement of pedestrian AWhen in the spatial dimension, no matter pedestrian A
It is how far apart with pedestrian B, it is moved with current speed, is eventually collided.Velocity variable is setObtain speed change
The relative velocity of pedestrian A afterwardsIts size are as follows:
WhenWhen not in spatial dimension C, then pedestrian A and pedestrian B is moved with the speed after changing, and will not be touched
It hits.Meet the set D of the speed knots modification avoided collision0, in the range of:
Wherein λ isWith the angle of AB line, value are as follows:
Pedestrian can realize avoiding for collision with the smallest speed knots modification, pass through judgementWith the position of AB line
Obtain the minimum speed as shown in figure 11 variableIts direction is vertical with tangent line, value size are as follows:
When in environment, there are when multiple pedestrians, the variable-ratio for each pedestrian successively can be found out according to the method described above
Value range is spent, finds out the intersection in these ranges, then be averaged to intersection, is changed as final speed.It is calculated
It after speed knots modification, needs for the variable to be assigned to each pedestrian, distributes how many knots modifications to each pedestrian and weighed by its collision prevention
Weight is determined.In reality, pedestrian takes when carrying out collision prevention under different attribute (old man, child), different situations (walking stops)
Collision strategy it is different, if young man always actively gives way when encountering old man, being parked in the people in original place, to be always to maintain speed constant
(staying on), for this situation, set forth herein the speed knots modification methods of salary distribution of different collision prevention weight ratios.Assuming that the power of A
Weight isThen after speed changes, A and B are relative to the speed in global coordinate systemWithIt is respectively as follows:
WhenWhen, it is the collision prevention of equal weight ratio, is the collision prevention of different weight ratios when being not equal to 0.5.It shows as
Mutually evacuation and one-sided active detours.As shown in figure 12.
4th step, obtains Simulation results.The present invention is tested by taking the southern station of Line of Nanjing Subway S1 line Nanjing as an example, should
Scene is an an island platform, and one passenger of generation in platform entrance every 0.4 second, the simulation that passenger enters station is as shown in figure 13.
Above-described specific embodiment has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that being not intended to limit the present invention the foregoing is merely a specific embodiment of the invention
Protection scope, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all include
Within protection scope of the present invention.
Claims (9)
1. a kind of large-scale virtual crowd analogue system, which is characterized in that including crowd behaviour model emulation module, environmental treatment
Module is realized in module and behavior, in which:
The ambient process module renders navigation grid, creates environment mould for carrying out calculation system to static environment model
Type;
The crowd behaviour model emulation module, for creating crowd's sensor model of region perception and virtual sense organ composition, people
Group's sensor model obtains external information from the environmental model, realizes that module carries out behaviour decision making and provides information for the behavior
Input;
The behavior realizes that module passes through logic judgment for handling the information input of the crowd behaviour model emulation module
Output behavior.
2. a kind of large-scale virtual crowd analogue system according to claim 1, which is characterized in that the crowd behaviour mould
Type emulation module includes synthesis perceptron, signal filter and decision behavior module, and synthesis perceptron obtains environment for crowd
In all information, the signal filter is for being filtered all information to obtain information needed, the decision behavior mould
Block is used to carry out behaviour decision making by information needed.
3. a kind of large-scale virtual crowd analogue system according to claim 2, which is characterized in that the synthesis perceptron
Including environmental area sensing module and sensory perception's module.
4. a kind of large-scale virtual crowd analogue system according to claim 3, which is characterized in that the environmental area sense
Know that module includes intelligent body, ray emitter and the collision body with specific label, the collision body setting is needing to carry out area
The position of domain perception, the ray emitter are arranged in the location of intelligent body.
5. a kind of large-scale virtual crowd analogue system according to claim 1, which is characterized in that the environmental treatment mould
The process of block rendering navigation grid are as follows: establish one and contain the triangulation network of all areas of feasible solutions of environment space, obtain by optimal
Triangle sets at the walkable region triangulation network, settled accounts, can be obtained with faster speed global by node to the triangulation network
Optimal path.
6. a kind of large-scale virtual crowd analogue system according to claim 5, which is characterized in that the triangulation network is built
Vertical process are as follows: define the initial polygon for surrounding all discrete datas;For the initial polygon, establish one it is initial
The triangulation network;By in first point P insertion initial triangulation in discrete point set, the triangle T in the triangulation network comprising P point is found out,
P is connected with three vertex of T, generates a new triangulation network;Triangles optimization is carried out using empty circumscribed circle characteristic, is obtained most
The new triangulation network;It repeats the above steps, until all the points into discrete point set are all gradually added in the triangulation network;It deletes comprising initial
The associated triangle on bounding box vertex forms the final triangulation network.
7. a kind of large-scale virtual crowd analogue system according to claim 5, which is characterized in that obtain by optimum triangular
The detailed process of the walkable region triangulation network of shape composition are as follows: by the triangulation network merge convex polygon, delete extra side,
Ignore small surface grids etc. reason, so that convex polygon quantity, number of edges amount navigation grid as few as possible are obtained, to reduce pathfinding
Calculation amount and data space in journey, the edge triangles from triangular mesh that mainly comprise the processes of for merging chimb are opened
Begin to divide;Merge convex polygon and adjacent triangle, guarantees that the triangle after merging remains as convex polygons;It repeats above-mentioned two
Step, until second step cannot continue to execute.
8. a kind of large-scale virtual crowd analogue system according to claim 1, which is characterized in that mould is realized in the behavior
Block includes route searching and dynamic collision prevention, behaviour decision making and execution by the behavior that logic judgment exports.
9. a kind of large-scale virtual crowd analogue system according to claim 8, which is characterized in that mould is realized in the behavior
Block carries out route searching by navigation grid route searching method, and navigation grid route searching method is found by continuous estimated path
The shortest point in path can be made as next node, until arriving at terminal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811504286.1A CN109727519A (en) | 2018-12-10 | 2018-12-10 | A kind of large-scale virtual crowd analogue system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811504286.1A CN109727519A (en) | 2018-12-10 | 2018-12-10 | A kind of large-scale virtual crowd analogue system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109727519A true CN109727519A (en) | 2019-05-07 |
Family
ID=66295913
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811504286.1A Pending CN109727519A (en) | 2018-12-10 | 2018-12-10 | A kind of large-scale virtual crowd analogue system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109727519A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110428688A (en) * | 2019-08-09 | 2019-11-08 | 郑州铁路职业技术学院 | A kind of rail traffic vehicles maintenance simulation training system based on virtual scene |
CN110917624A (en) * | 2019-12-06 | 2020-03-27 | 珠海海鸟科技有限公司 | Path determination method, device and equipment |
US11880945B2 (en) | 2020-07-25 | 2024-01-23 | Silver Spoon Animation Inc. | System and method for populating a virtual crowd in real time using augmented and virtual reality |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1889044A (en) * | 2006-07-19 | 2007-01-03 | 中国科学院计算技术研究所 | Virtual human movement simulation frame |
CN101777093A (en) * | 2009-12-25 | 2010-07-14 | 电子科技大学 | Large-scale virtual crowd routing method |
CN102467588A (en) * | 2010-11-10 | 2012-05-23 | 上海日浦信息技术有限公司 | Pedestrian behavior simulation method in microscopic traffic simulation |
CN102708581A (en) * | 2011-03-28 | 2012-10-03 | 上海日浦信息技术有限公司 | Virtual crowd motion simulation framework |
CN104008562A (en) * | 2014-06-06 | 2014-08-27 | 东南大学 | User-oriented planned virtual crowd simulation framework |
CN105205866A (en) * | 2015-08-30 | 2015-12-30 | 浙江中测新图地理信息技术有限公司 | Dense-point-cloud-based rapid construction method of urban three-dimensional model |
CN105468801A (en) * | 2014-09-09 | 2016-04-06 | 中国科学院深圳先进技术研究院 | Simulation method and system for crowd evacuation in public place |
CN107423480A (en) * | 2017-05-19 | 2017-12-01 | 山东师范大学 | A kind of crowd movement's emulation mode and system |
KR101821592B1 (en) * | 2016-11-16 | 2018-01-25 | 그리드스페이스(주) | Walking simulator system using virtual reality and method for calculating pedestrian position thereof |
CN108153945A (en) * | 2017-12-11 | 2018-06-12 | 中山大学 | A kind of crowd evacuation emulation model based on multi-Agent |
CN108376198A (en) * | 2018-02-27 | 2018-08-07 | 山东师范大学 | A kind of crowd simulation method and system based on virtual reality |
CN108389259A (en) * | 2018-03-13 | 2018-08-10 | 武汉大学 | A kind of random center focused image mesh tune method and system |
CN108434741A (en) * | 2018-03-15 | 2018-08-24 | 网易(杭州)网络有限公司 | Control method for movement in virtual reality and device |
CN108447125A (en) * | 2018-02-28 | 2018-08-24 | 哈尔滨工程大学 | A kind of Delaunay triangulation network rapid generation based on multibeam bathymetric data |
CN108959667A (en) * | 2017-05-18 | 2018-12-07 | 株式会社日立制作所 | Pedestrian's following behavior emulation mode and pedestrian's following behavior simulator |
-
2018
- 2018-12-10 CN CN201811504286.1A patent/CN109727519A/en active Pending
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1889044A (en) * | 2006-07-19 | 2007-01-03 | 中国科学院计算技术研究所 | Virtual human movement simulation frame |
CN101777093A (en) * | 2009-12-25 | 2010-07-14 | 电子科技大学 | Large-scale virtual crowd routing method |
CN102467588A (en) * | 2010-11-10 | 2012-05-23 | 上海日浦信息技术有限公司 | Pedestrian behavior simulation method in microscopic traffic simulation |
CN102708581A (en) * | 2011-03-28 | 2012-10-03 | 上海日浦信息技术有限公司 | Virtual crowd motion simulation framework |
CN104008562A (en) * | 2014-06-06 | 2014-08-27 | 东南大学 | User-oriented planned virtual crowd simulation framework |
CN105468801A (en) * | 2014-09-09 | 2016-04-06 | 中国科学院深圳先进技术研究院 | Simulation method and system for crowd evacuation in public place |
CN105205866A (en) * | 2015-08-30 | 2015-12-30 | 浙江中测新图地理信息技术有限公司 | Dense-point-cloud-based rapid construction method of urban three-dimensional model |
KR101821592B1 (en) * | 2016-11-16 | 2018-01-25 | 그리드스페이스(주) | Walking simulator system using virtual reality and method for calculating pedestrian position thereof |
CN108959667A (en) * | 2017-05-18 | 2018-12-07 | 株式会社日立制作所 | Pedestrian's following behavior emulation mode and pedestrian's following behavior simulator |
CN107423480A (en) * | 2017-05-19 | 2017-12-01 | 山东师范大学 | A kind of crowd movement's emulation mode and system |
CN108153945A (en) * | 2017-12-11 | 2018-06-12 | 中山大学 | A kind of crowd evacuation emulation model based on multi-Agent |
CN108376198A (en) * | 2018-02-27 | 2018-08-07 | 山东师范大学 | A kind of crowd simulation method and system based on virtual reality |
CN108447125A (en) * | 2018-02-28 | 2018-08-24 | 哈尔滨工程大学 | A kind of Delaunay triangulation network rapid generation based on multibeam bathymetric data |
CN108389259A (en) * | 2018-03-13 | 2018-08-10 | 武汉大学 | A kind of random center focused image mesh tune method and system |
CN108434741A (en) * | 2018-03-15 | 2018-08-24 | 网易(杭州)网络有限公司 | Control method for movement in virtual reality and device |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110428688A (en) * | 2019-08-09 | 2019-11-08 | 郑州铁路职业技术学院 | A kind of rail traffic vehicles maintenance simulation training system based on virtual scene |
CN110428688B (en) * | 2019-08-09 | 2021-03-02 | 郑州铁路职业技术学院 | Rail transit vehicle overhauls real standard system of simulation based on virtual scene |
CN110917624A (en) * | 2019-12-06 | 2020-03-27 | 珠海海鸟科技有限公司 | Path determination method, device and equipment |
CN110917624B (en) * | 2019-12-06 | 2024-01-23 | 珠海海鸟科技有限公司 | Path determination method, device and equipment |
US11880945B2 (en) | 2020-07-25 | 2024-01-23 | Silver Spoon Animation Inc. | System and method for populating a virtual crowd in real time using augmented and virtual reality |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Loscos et al. | Intuitive crowd behavior in dense urban environments using local laws | |
JP3370555B2 (en) | Pedestrian information provision system | |
Meijers et al. | 3D geoinformation indoors: structuring for evacuation | |
Jiang et al. | Continuum crowd simulation in complex environments | |
CN109727519A (en) | A kind of large-scale virtual crowd analogue system | |
Sarmady et al. | Modeling groups of pedestrians in least effort crowd movements using cellular automata | |
Zhang et al. | Modeling and analyzing 3D complex building interiors for effective evacuation simulations | |
Kim et al. | Modeling handicapped pedestrians considering physical characteristics using cellular automaton | |
Aschwanden et al. | Empiric design evaluation in urban planning | |
Chang et al. | Legible simplification of textured urban models | |
Bouvier et al. | Crowd simulation in immersive space management | |
Stojanovski et al. | Typo-morphology and environmental perception of urban space | |
Benedikt et al. | Isovists and the metrics of architectural space | |
Yang et al. | A multi-agent and GIS based simulation for emergency evacuation in park and public square | |
Rodgers | Graph drawing techniques for geographic visualization | |
Lin et al. | Bird eye analysis of skyline views from an elevated recreation trail—The case of the Fu Forest Trail, Fuzhou, China | |
Wang et al. | Incorporating topography in a cellular automata model to simulate residents evacuation in a mountain area in China | |
París et al. | A new autonomous agent approach for the simulation of pedestrians in urban environments | |
Xu et al. | Infrastructures for research on multimodal moving objects | |
Lou et al. | The research on aggregate analysis of the regulatory detailed planning by considering the traffic capacity | |
Bandini et al. | Multi-agent approach to localization problems: The case of multilayered multi-agent situated system | |
Aschwanden et al. | Evaluation of 3D city models using automatic placed urban agents | |
Song et al. | A grid-based graph data model for pedestrian route analysis in a micro-spatial environment | |
Berwaldt et al. | Procedural generation of favela layouts on arbitrary terrains | |
Bartie et al. | Referring expressions in location based services: The case of the ‘opposite’relation |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190507 |