Observations of Passenger Flow and Verification of a Crowd Dynamics Model

This paper reports observations of passenger flow in the Wuchang railway station in Wuhan, China during the Chinese Traditional Spring Festival in 2006. The data collected are used to verify a crowd dynamics model previously developed. The crowd dynamics model is based on simulating the global movement of each individual under the influence of the surrounding crowd, and the good agreement between the predictions and observations validates the prediction model. The crowd dynamics model suggests that the crowd movement speed is dominated by two factors： the front-back inter-person effect, and the pedestrian＇s self-motive. The first effect gives logarithmic relationship between the crowd speed and crowd density. The second factor depends on the individual motive driven with which people try to divorce themselves from the control of the crowd movement. The prediction model are helpful to guide the design of public traffic systems for effective crowd dispersal.

Visual Motion Segmentation in Crowd Videos Based on Spatial-Angular Stacked Sparse Autoencoders

Visual motion segmentation(VMS)is an important and key part of many intelligent crowd systems.It can be used to figure out the flow behavior through a crowd and to spot unusual life-threatening incidents like crowd stampedes and crashes,which pose a serious risk to public safety and have resulted in numerous fatalities over the past few decades.Trajectory clustering has become one of the most popular methods in VMS.However,complex data,such as a large number of samples and parameters,makes it difficult for trajectory clustering to work well with accurate motion segmentation results.This study introduces a spatial-angular stacked sparse autoencoder model(SA-SSAE)with l2-regularization and softmax,a powerful deep learning method for visual motion segmentation to cluster similar motion patterns that belong to the same cluster.The proposed model can extract meaningful high-level features using only spatial-angular features obtained from refined tracklets(a.k.a‘trajectories’).We adopt l2-regularization and sparsity regularization,which can learn sparse representations of features,to guarantee the sparsity of the autoencoders.We employ the softmax layer to map the data points into accurate cluster representations.One of the best advantages of the SA-SSAE framework is it can manage VMS even when individuals move around randomly.This framework helps cluster the motion patterns effectively with higher accuracy.We put forward a new dataset with itsmanual ground truth,including 21 crowd videos.Experiments conducted on two crowd benchmarks demonstrate that the proposed model can more accurately group trajectories than the traditional clustering approaches used in previous studies.The proposed SA-SSAE framework achieved a 0.11 improvement in accuracy and a 0.13 improvement in the F-measure compared with the best current method using the CUHK dataset.

Numerical Simulation of the Flow of Crowds at the Jamarat Bridge during the Annual Hajj Event

The huge number of pilgrims to the holy Mecca in the Hajj needs high awareness of crowd safety management. The stoning of the Jamarat, which is one of the rituals of the Hajj, undergoes the most dangerous crowd movements where fatal accidents occurred. This work investigates some problems related with the crowd dynamics when stoning the Jamarat pillars and gives some solutions. The main idea of this research is to suppose that the crowd dynamics is assimilated to fluid movement under certain conditions. Numerical simulation using a computational fluid dynamics program is used to solve Navier-Stokes equations governing the mechanics of homogeneous and incompressible fluid in a domain similar to the Jamarat Bridge from the entrance to the middle Jamarah. Some solutions are proposed inspired by the flow solutions to better manage crowd movements in the Jamarat Bridge and eventually in other similar dynamics events like sporting events.

Modeling Passengers Boarding in Aircraft Using Cellular Automata

Aircraft are profitable to their owners as long as they are in the air transporting passengers to their destinations;therefore it is vital to minimize as much as possible their preparation time on the ground.In this paper we simulate different boarding strategies with the help of a model based on cellular automata parallel computational tool,attempting to find the most efficient way to deliver each passenger to her/his assigned seat.Two seat arrangements are used,a small one based on Airbus A320/ Boeing 737 and a larger one based on Airbus A380/ Boeing777-300.A wide variety of parameters,including time delay for luggage storing,the frequency by which the passengers enter the plane,different walking speeds of passengers depending on sex,age and height,and the possibility of walking past their seat,are simulated in order to achieve realistic results,as well as monitor their effects on boarding time.The simulation results indicate that the boarding time can be significantly reduced by the simple grouping and prioritizing of passengers.In accordance with previous papers and the examined strategies,the outside-in and reverse pyramid boarding methods outperform all the others for both the small and large airplane seat layout.In the latter,the examined strategies are introduced for first time in an analogous way to the initial small seat arrangement of Airbus A320/ Boeing737 aircraft family.Moreover,since in real world scenarios,the compliance of all the passengers to the suggested group division and boarding strategy cannot be guaranteed,further simulations were conducted.It is clear that as the number of passengers disregarding the priority of the boarding groups increases,the time needed for the boarding to complete tends towards that of the random boarding strategy,thus minimizing the possible advantages gained by the proposed boarding strategies.

AB065. Pedestrian modeling using the least-action principle

Background:In this work,we present a theoretical and experimental study of the natural movement of pedestrians when passing through a limited and known area of a shopping center.The modeling problem for the motion of a single pedestrian is complex and extensive;therefore,we focus on the need to design models taking into account mechanistic aspects of human locomotion.The theoretical study used mean values of pedestrian characteristics,e.g.,density,velocity,and many obstacles.We propose a human pedestrian trajectory model by using the least-action principle,and we compared it against experimental results.The experimental study is conducted in a Living Lab inside a shopping center using infrared cameras.For this experiment,we collected highly accurate trajectories allowing us to quantify pedestrian crowd dynamics.The tests included 20 runs distributed over five days with up to 25 test persons.Additionally,to gain a better understanding of subjects’trajectories,we simulated a background of different pathway scenarios and compared it with real trajectories.Our theoretical framework takes the minimum error between previously simulated and real point pathways to predict future points on the subject trajectory.Methods:This paper explores paths of 25 pedestrians along a known area.After obtaining the trajectory and their points of origin,we evaluated the speed with the objective to calculate the kinetic force of the pedestrian.In the present model,we assume that the principle of least action holds and using this concept we can obtain the potential force.Once all the forces of pedestrian movement are known,then we calculate the adjustment of the parameters employed in the equations of the social force model.Results:It is possible to reproduce observed results for real pedestrian movement by using the Principle of Least Action.In the first scenario,we focused on a pedestrian walking without obstacles.Using the actual trajectories of the experiment we obtained the necessary information and applied it to the Social Force Model.Our simulations were clearly able to reproduce the actual observed average trajectories for the free obstacle walking conditions.Conclusions:When a scenario does not represent free walking(obstacles,constraints),the potential energy and the kinetic energy are modified.Note that when the trajectory is real,the action is assumed to equal zero.That is the value of the potential energy changes in each interaction with a new obstacle.However,the value of the action remains.It is shown here that we can clearly reproduce some scenarios and calibrate the model according to different situations.Using different values of potential energy,we can obtain the values of the actual pathway.Nevertheless,as a significant extension concerning this model,it would be desirable to simulate cellular automata that could learn the situation and improve the approximation model to predict the real trajectories with more accuracy.

Non-monotonic behavior of jam probability and stretchedexponential distribution in pedestrian counterflow

We adopt a floor field cellular automata model to study the statistical properties of bidirectional pedestrian flow movingin a straight corridor. We introduce a game-theoretic framework to deal with the conflict of multiple pedestrians tryingto move to the same target location. By means of computer simulations, we show that the complementary cumulative distributionof the time interval between two consecutive pedestrians leaving the corridor can be fitted by a stretched exponentialdistribution, and surprisingly, the statistical properties of the two types of pedestrian flows are affected differently by theflow ratio, i.e., the ratio of the pedestrians walking toward different directions. We also find that the jam probability exhibitsa non-monotonic behavior with the flow ratio, where the worst performance arises at an intermediate flow ratio of around0.2. Our simulation results are consistent with some empirical observations, which suggest that the peculiar characteristicsof the pedestrians may attributed to the anticipation mechanism of collision avoidance.