Experimental and Real World Applications of Agent-Based Pedestrian Group Modeling Giuseppe Vizzari, Lorenza Manenti Kazumichi Ohtsuka, Kenichiro Shimura Complex Systems and Artificial Intelligence research center Research Center for Advanced Science & Technology Università degli Studi di Milano–Bicocca, Milano, Italy The University of Tokyo, Japan {vizzari,manenti}@disco.unimib.it tukacyf@mail.ecc.u-tokyo.ac.jp, shimura@tokai.t.u-tokyo.ac.jp Abstract—The simulation of pedestrian dynamics is a consoli- relevance of human behaviour, and especially of the move- dated area of application for agent-based models: successful case ments of pedestrians, in built environment in normal and studies can be found in the literature and off-the-shelf simulators extraordinary situations, and its implications for the activities are commonly employed by decision makers and consultancy companies. These models, however, generally do not consider of architects, designers and urban planners are apparent (see, the explicit representation of pedestrians aggregations (groups), e.g., [1]), especially considering dramatic episodes such as the related occurring relationships and their dynamics. This terrorist attacks, riots and fires, but also due to the growing work is aimed at discussing the relevance and significance of issues in facing the organization and management of public this research effort with respect to the need of empirical data events (ceremonies, races, carnivals, concerts, parties/social about the implication of the presence of groups of pedestrians in different situations (e.g. changing density, spatial configurations gatherings, and so on) and in designing naturally crowded of the environment). The paper describes an agent-based model places (e.g. stations, arenas, airports). Computational models encapsulating in the pedestrian’s behavioural specification effects for the simulation of crowds are thus growingly investigated representing both traditional individual motivations (i.e. tendency in the scientific context, and these efforts led to the real- to stay away from other pedestrians while moving towards the ization of commercial off-the-shelf simulators often adopted goal) and a simplified account of influences related to the presence of groups in the crowd. The model is tested in a simple scenario by firms and decision makers1 . Models and simulators have to evaluate the implications of some modeling choices and the shown their usefulness in supporting architectural designers presence of groups in the simulated scenario. Moreover, the model and urban planners in their decisions by creating the possibility is applied in a real world scenario characterized by the presence to envision the behavior of crowds of pedestrians in specific of organized groups as an instrument for crowd management. actual environments and planned designs, to elaborate what- Results are discussed and compared to experimental observations and to data available in the literature. if scenarios and evaluate their decisions with reference to specific metrics and criteria. Despite the substantial amount of research efforts this area is still quite lively and we are far I. I NTRODUCTION from a complete understanding of the complex phenomena Agent–based approaches to the simulation of complex sys- related to crowds of pedestrians in the environment: one tems represent a relatively recent but extremely successful of the least studied and understood aspects of crowds of application area of concepts, abstractions, models defined pedestrians is represented by the implications of the presence in the area of autonomous agents and multi–agent systems of groups [2]. In particular, little work in the direction of (MAS). Agent-based models have been adopted to model modeling and simulating relatively large groups within a crowd complex systems in very different contexts, ranging from of pedestrians encompassing some form of validation (either social and economical simulation to logistics optimization, quantitative or qualitative) against real data can be found in from biological systems to traffic. Large groups and crowds the literature. of pedestrians represent a typical example of complex system: The main aim of this work is to present motivations, the overall behavior of the system can only be defined in fundamental research questions and directions, and results terms of the actions of the individuals that compose it, and of an agent–based modeling and simulation approach to the the decisions of the individuals are influenced by the previous multidisciplinary investigation of the complex dynamics that actions of other pedestrians sharing the same space. Sometimes characterize aggregations of pedestrians and crowds. In par- the interaction patterns are competitive, since pedestrians may ticular, in this paper we will present an agent–based model have conflicting goals (i.e. they might wish to occupy the same of pedestrians considering groups as a first–class abstraction spot of the shared environment), but collaborative patterns influencing the behaviour of its members and, in turn, of the can also be identified (e.g. leave room to people getting off whole system. The model has been tested (i) in a schematic a subway train before getting on). The overall system is situation that has also been analyzed by means of field characterized by self-organization mechanisms and emergent phenomena. 1 see http://www.evacmod.net/?q=node/5 for a significant although not nec- Despite the complexity of the studied phenomenon, the essarily comprehensive list of simulation platforms. experiments to characterize the implications of groups in the fundamental proxemical [5] concepts like the tendency of a overall pedestrian dynamics and (ii) in a real world scenario pedestrian to stay away from other ones while moving towards in which pedestrians were organized in large groups for sake his/her goal. Proxemics essentially represents a fundamental of crowd management. assumption of most modeling approaches, although very few The paper breaks down as follows: the following section authors actually mention this anthropological theory [6], [7]. will set the present work in the state of the art of pedestrian CA based approaches can be roughly classified in ad-hoc and crowd modeling and simulation, with specific reference approaches for specific situations (like the case of bidirectional to recent works focusing on the modeling and implications of flows at intersections described in [8]) and general approaches, groups. Section III will introduce the model that was adopted whose main representative is the floor-field approach [9], in an experimental scenario, described in section IV, and in in which the cells are endowed with a discretized gradient a real world scenario, described in section V. The scenarios guiding pedestrians towards potential destinations. will be described and the achieved results will be discussed. While particle and CA based approaches are mostly aimed Conclusions and future developments will end the paper. at generating quantitative results about pedestrian and crowd This work is set in the context of the Crystals project2 , movement, agent based approaches are sometimes aimed at a joint research effort between the Complex Systems and the generation of effective visualizations of believable crowd Artificial Intelligence research center of the University of dynamics, and therefore the above approaches do not neces- Milano–Bicocca, the Centre of Research Excellence in Hajj sarily share the same notion of realism and validation. Works and Omrah and the Research Center for Advanced Science like [10] and [11] essentially extend CA approaches, separat- and Technology of the University of Tokyo. The main focus ing the pedestrians from the environment, but they essentially of the project is on the adoption of an agent-based pedestrian adopt similar methodologies. Other approaches like [12], [13] and crowd modeling approach to investigate meaningful rela- are more aimed at generating visually effective and believable tionships between the contributions of anthropology, cultural pedestrians and crowds in virtual worlds. Other approaches, characteristics and existing results on the research on crowd like [14], employ cognitive agent models for different goals, dynamics, and how the presence of heterogeneous groups but they are not generally aimed at making predictions about influence emergent dynamics in the context of the Hajj and pedestrian movement for sake of decision support. Omrah. The implications of particular relationships among A small number of recent works represent a relevant effort pedestrians in a crowd are generally not considered or treated towards the modeling of groups, respectively in particle- in a very simplistic way by current approaches. In the specific based [15], [16] (extending the social force model), in CA- context of the Hajj, the yearly pilgrimage to Mecca that based [17] (with ad-hoc approaches) and in agent-based ap- involves over 2 millions of people coming from over 150 proaches [18], [19], [20], [21] (introducing specific behavioral countries, the presence of groups (possibly characterized by an rules for managing group oriented behaviors): in all these internal structure) and the cultural differences among pedes- approaches, groups are modeled by means of additional con- trians represent two fundamental features of the reference tributions to the overall pedestrian behaviour representing the scenario. Studying implications of these basic features is the tendency to stay close to other group members. However, main aim of the Crystals project. the above approaches only mostly deal with small groups in relatively low density conditions; those dealing with relatively II. R ELATED W ORKS large groups (tens of pedestrians) were not validated against A comprehensive but compact overview of the different real data. The last point is a crucial and critical element of this approaches and models for the simulation of pedestrian and kind of research effort: computational models represent a way crowd dynamics is not easily defined: scientific interdisci- to formally and precisely define a computable form of theory plinary workshops and conferences are in fact specifically de- of pedestrian and crowd dynamics. However, these theories voted to this topic (see, e.g., the proceedings of the first edition must be validated employing field data, acquired by means of the International Conference on Pedestrian and Evacuation of experiments and observations of the modeled phenomena, Dynamics [3] and consider that this event has reached the fifth before the models can actually be used for sake of prediction. edition in 2010). A possible schema to classify the different approaches is based on the way pedestrians are represented III. GA-PED M ODEL and managed. From this perspective, pedestrian models can be roughly classified into three main categories that respectively We will now briefly introduce a model based on simple consider pedestrians as particles subject to forces, particular reactive situated agents based on some fundamental features states of cells in which the environment is subdivided in of CA approaches to pedestrian and crowd modeling and Cellular Automata (CA) approaches, or autonomous agents simulation, with specific reference to the representation and acting and interacting in an environment. management of the simulated environment and pedestrians; in The most successful particle based approach is represented particular, the adopted approach is discrete both in space and by the social force model [4], which implicitly comprises in time. The present description of the model is simplified and reduced for sake of space, reporting only a basic description 2 http://www.csai.disco.unimib.it/CSAI/CRYSTALS/ of the elements required to understand its basic mechanisms; Fig. 1. Schematic representation of a simple scenario: a 2.5 by 10 m corridor, with exits on the short ends and two sets of 25 pedestrians. The discretization of 50 cm and the floor field directing towards the right end is shown on the right. an extended version of the model description can be found in according to their perception of the environment and their goal, [22]. but their action is actually triggered by the simulation engine and they are not thus provided with a thread of control of A. Environment their own. More precisely, the simulation turn activates every The environment in which the simulation takes place is a pedestrian once in every turn, adopting a random order in lattice of cells, each representing a portion of the simulated the agent selection: this agent activation strategy, also called environment and comprising information about its current shuffled sequential updating [25], is characterized by the fact state, both in terms of physical occupation by an obstacle that conflicts between pedestrians are prevented. or by a pedestrian, and in terms of additional information, Each pedestrian is provided with a simple set of attributes: for instance describing its distance from a reference point or pedestrian = hpedID, groupIDi with pedID being an identi- point of interest in the environment and/or its desirability for fier for each pedestrian and groupID (possibly null, in case pedestrians following a certain path in the environment. of individuals) the group the pedestrian belongs to. For the The scale of discretization is determined according to the applications presented in this paper, the agents have a single principle of achieving cells in which at most one pedestrian goal in the experimental scenario, but in more complex ones can be present; traditionally the side of a cell is fixed at 40 the environment could be endowed with multiple floor fields or 50 cm, respectively determining a maximum density of 4 and the agent could be also characterized by a schedule, in and 6.5 pedestrian per square meter. The choice of the scale terms of a sequence of floor fields and therefore intermediate of discretization also influences the length of the simulation destinations to be reached. turn: the average speed of a pedestrian can be set at about The behavior of a pedestrian is represented as a flow made 1.5 meters per second (see, e.g., [23]) therefore, assuming up of three stages: sleep, movement evaluation, movement. that a pedestrian can perform a single movement between When a new iteration starts each pedestrian is in a sleeping a cell and an adjacent one (according to the Von Neumann state. The system wakes up each pedestrian once per iteration neighbourhood), the duration of a simulation turn is about and, then, the pedestrian passes to a new state of movement 0.33 seconds in case of a 50 cm discretization and 0.27 in evaluation. In this stage, the pedestrian collects all the infor- case of a finer 40 cm discretization. mation necessary to obtain spatial awareness. In particular, Each cell can be either vacant, occupied by an obstacle or by every pedestrian has the capability to observe the environment a specific pedestrian. In order to support pedestrian navigation around him, looking for other pedestrians (that could be part of in the environment, each cell is also provided with specific his/her group), walls and other obstacles, according to the Von floor fields [9]. In particular, each relevant final or intermediate Neumann neighbourhood. The choice of the actual movement target for a pedestrian is associated to a floor field, representing destination between the set of potential movements (i.e. non a sort of gradient indicating the most direct way towards empty cells are not considered) is based on the elaboration of the associated point of interest (e.g., see Fig.1 in which an utility value, called likability, representing the desirability a simple scenario and the relative floor field representation of moving into that position given the state of the pedestrian. are shown). The GA-Ped model only comprises static floor Formally, given a pedestrian belonging to a group g and fields, specifying the shortest path to destinations and targets. reaching a goal t, the likability of a cell cx,y is defined as: Interactions between pedestrians, that in other models are described by the use of dynamic floor fields [24], in our model are managed through the agent perception model. li(cx,y , g, t) = wt · goal(t, (x, y)) + wg · group(g, (x, y)) B. Pedestrians wo · obs(x, y) ws · others(g, (x, y)) + ✏. (1) Pedestrians in the GA-PED model have a limited form where the functions obst counts the number of obstacles of autonomy, meaning that they can choose were to move in the Von Neumann neighbourhood of a given cell, goal Indiv. Couples Triples Groups of 5 returns the value of the floor field associated to the target Den. Sp. Fl. Sp. Fl. Sp. Fl. Sp. Fl. t in a give cell, group and other respectively count the 0,4 1,54 0,62 1,55 0,62 1,47 0,59 - - number of members and non-members of the group g, ✏ 0,8 1,33 1,06 1,41 1,12 1,32 1,05 1,14 0,91 represents a random value. Group cohesion and floor field 1,2 1,14 1,37 1,19 1,43 1,12 1,35 0,98 1,18 are positive components because the pedestrians wish to reach 1,6 0,95 1,52 0,99 1,59 0,93 1,49 0,83 1,32 their destinations quickly, while staying close to other group 2,0 0,73 1,46 0,78 1,56 0,74 1,47 0,66 1,32 members. On the contrary, the presence of obstacles and other 2,4 0,41 0,98 0,41 0,99 0,44 1,06 0,42 1,00 2,8 0,22 0,60 0,23 0,64 0,25 0,70 0,24 0,66 pedestrians have a negative impact as a pedestrian usually 3,2 0,13 0,42 0,14 0,46 0,16 0,50 0,14 0,46 tends to avoid them. A random factor is also added to the overall evaluation of the desirability of every cell. TABLE I In the usual floor field models, after a deterministic elabora- S IMULATION R ESULTS : VALUES ON AVERAGE SPEED ( METERS PER SECOND ) AND FLOW ( PERSONS / M · S ), CONSIDERING DIFFERENT tion of the utility of each cell, not comprising thus any random DENSITIES ( PERSONS PER SQUARE METER ) OF PEDESTRIANS AND factor, the utilities are translated into the probabilities that the DIFFERENT CONFIGURATIONS OF GROUPS . related cell is selected as movement destination. This means that for a pedestrian generally there is a higher probability of moving towards his/her destination and according to proxemic considerations, but there is also the probability, for instance, to move away from his/her goal or to move far from his/her group. In this work, we decided to include a small random factor to the utility of each cell and to choose directly the movement that maximizes the agent utility. A more thorough comparison of the implications of this choice compared to the basic floor field approach is out of the scope of this paper and it is object of future works. IV. E XPERIMENTAL S CENARIO The GA-Ped model was adopted to realize a set of simula- tions in different starting conditions (mainly changing density of pedestrians in the environment, but also different configura- Fig. 2. Fundamental diagram for different configurations of pedestrian based tions of groups present in the simulated pedestrian population) on simulation results in Table I. in a situation in which experiments focused at evaluating the impact of the presence of groups of different size was being investigated. to move and how to perform the experiment very quickly, since the first experiment took them about 18 seconds while A. Experiments the average completion time over 12 experiments is about 15 The environment in which the experiments took place is seconds. represented in Fig. 1: a 2.5 by 10 m corridor, with exits on The number of performed experiments is probably too the short ends. The experiments were characterized by the low to draw some definitive conclusions, but the total travel presence of two sets of 25 pedestrians, respectively starting at times of configurations including individuals and pairs were the two ends of the corridor (in 2 by 2.5 m areas), moving consistently lower than those not including groups. Qualitative towards the other end. Various cameras were positioned on analysis of the videos showed that pairs can easily form a line, the side of the corridor and the time required for the two sets and this reduces the friction with the facing group. Similar of pedestrians to complete their movement was also measured considerations can be done for large groups; on the other (manually from the video footage). end, groups of three pedestrians sometimes had difficulties in Several experiments were conducted, some of which also forming a lane, retaining a triangular shape similar to the ‘V’ considered the presence of groups of pedestrians, that were shaped observed and modeled in [15], and this caused a total instructed on the fact that they had to behave as friends or travel times that were higher than average in two of the three relatives while moving during the experiment. In particular, experiments involving this type of group. the following scenarios have been investigated: (i) single pedestrians (3 experiments); (ii) 3 couples of pedestrians for B. Simulation Results each direction (2 experiments); (iii) 2 triples of pedestrians for We applied the model described in Sect. III to the previous each direction (3 experiments); (iv) a group of six pedestrians scenario by means of an agent-based platform based on GA- for each direction (4 experiments). Ped approach. A description of the platform can be found One of the observed phenomena was that the first ex- in [26]. We employed the gathered data and additional data periment actually required more time for the pedestrians to available in the literature to perform a calibration of the complete the movement; the pedestrians actually learned how parameters, essentially determining the relative importance of (a) the goal oriented, (b) general proxemics and (c) group V. R EAL W ORLD S CENARIO proxemic components of the movement choice. In particular, we first identified a set of plausible values for the wt and A. Environment and observations wo parameters employing experimental data regarding a one- directional flow. Then we employed data from bidirectional The model was also adopted to elaborate different what- flow situations to further tune these parameters as well as the if scenarios in a real world case study. In particular, the value of the wg parameter: the latter was set in order to achieve simulated scenario is characterized by the presence of a station a balance between effectiveness in preserving group cohesion of the Mashaer line, a newly constructed rail line in the area and preserving aggregated measures on the overall pedestrian of Makkah. The goal of this infrastructure is to reduce the flow (an excessive group cohesion value reduces the overall congestion caused by the presence of other collective means pedestrian flow and produces unrealistic behavior). of pilgrim transportation (i.e. buses) during the Hajj: the yearly pilgrimage to Mecca that involves over 2 millions of people We investigated the capability of our model to fit the coming from over 150 countries and some of its phase often fundamental diagram proposed in the literature for charac- result in congestions of massive proportions. In this work, we terizing pedestrian simulations [27] and other traffic related are focusing on a specific point of one of the newly constructed phenomena. This kind of diagram shows how the average stations, Arafat I. One of the most demanding situations that velocity of pedestrians varies according to the density of the the infrastructure of the Mashaer Rail line must be able to simulated environment. Moreover, we wanted to distinguish sustain is the one that takes place after the sunset of the the different performance of different agent types, and essen- second day of the pilgrimage, which involves the transport of tially individuals, members of pairs, groups of three and five pilgrims from Arafat to Muzdalifah. The pilgrims that employ pedestrians over a relatively wide spectrum of densities. To do the train to proceed to the next phase of the process must be so, we performed continuous simulations of the bidirectional able to move from the tents or other accommodation to the pedestrian flows in the corridor with a changing number of station in an organized flow that should be consistent with pedestrians, to alter their density. For each density value the movement of trains from Arafat to Muzdalifah stations. displayed in the graph shown in Figure 2 is related to at least Since pilgrims must leave the Arafat area before midnight, the 1 hour of simulated time. trains must continuously load pilgrims at Arafat, carry them to Muzdalifah, and come back empty to transport other pilgrims. The achieved fundamental diagram represents in qualita- The size of the platforms was determined to allow hosting tively correct way the nature of pedestrian dynamics: the flow in a safe and comfortable way a number of pilgrims also of pedestrians increases with the growing of the density of exceeding the potential number of passengers of a whole train. the corridor unit a critical value is reached. If the system Each train is made up of 12 wagons, each able to carry 250 density is increased beyond that value, the flow begins to passengers for a total of approximately 3000 persons. In order decrease significantly as the friction between pedestrians make to achieve an organized and manageable flow of people from movements more difficult. outside the station area to the platforms, the departure process An overview on the results of the simulations are shown in was structured around the idea of waiting–boxes: pilgrims Table I in which values on average speed and flow, considering are subdivided into groups of about 250 persons that are different densities of pedestrians and different configurations led by specific leaders (generally carrying a pole with signs of groups are presented. supporting group identification). The groups start from the tents area and flow into these fenced queuing areas located The simulation results are in tune with the experimental in immediately outside the station, between the access ramps. data coming from observations: in particular, the flow of pairs Groups of pilgrims wait in these areas for an authorization by of pedestrians is consistently above the curve of individuals. the station agents to move towards the ramps or elevators. In This means that the average speed of members of pairs is this way, it is possible to stop the flow of pilgrims whenever actually higher than the average speed of individuals. This is the number of persons on the platforms (or on their way to due to the fact that they easily tend to form a line, in which reach it using the ramps or elevators) is equal to the train the first pedestrian has the same probability to be stuck as an capacity, supporting thus a smooth boarding operation. individual, but the follower has a generally higher probability Three photos and a schematic representation of the real to move forward, following the path “opened” by the first world scenario and the related phenomena are shown in member of the pair. The same does not happen for larger Figure 3: the bottom right photo shows a situation in which groups, since for them it is more difficult to form a line: the the waiting-box principle, preventing the possibility of two curves related to groups of three and five members are below flows simultaneously converging to a ramp, was not respected, the curve of individuals for most of the spectrum of densities, causing a higher than average congestion around the ramp. precisely until very high density values are reached. In this This anomaly was plausibly due to the fact that it was the case, the advantage of followers overcomes the disadvantage of first time the station was actually used, therefore also the offering a larger profile to the counter flow and the combined management personnel was not experienced in the crowd average velocity is higher than that of individuals. management procedures. Fig. 3. Photos and a schematic representation of the real world scenario and the related phonomena. B. Simulation Results and the platform (that however increased with the growth of the number of pilgrims in the simulated scenario), but rather Three different scenarios were realized adopting the pre- different metrics of space utilization. This kind of metric viously defined model and using the parameters that were is tightly related to the so called level of service [28], a employed in the previous case study: (i) the flow of a group measure of the effectiveness of elements of a transportation of pilgrims from one waiting box to the ramp; (ii) the simulta- infrastructure; it is also naturally related to proxemics, since neous flow of two groups from two different waiting boxes to a low level of service is related to a unpleasant perceived the same ramp; (iii) the simultaneous flow of three groups of situation due to the invasion of the personal (or even intimate) pilgrims, two as in the previous situation, one coming directly space. from the tents area. Every group included 250 pilgrims. The goal of the analysis was to understand if the model is able The diagrams shown in Figure 4 report three metrics de- to qualitatively reflect the increase in the waiting times and scribing three different phenomena: (i) a situation in which the space utilization when the waiting box principle was not an agent in a cell of the environment was willing to move respected. but it was unable to perform the action due to the excessive The environment was discretized adopting 50cm sided cells space occupation; (ii) a situation in which an agent actually and the cell space was endowed with a floor field leading moved from a cell of the environment; (iii) the “set sum” of towards the platform, by means of the ramp. The different the previous situations, in other words, the situations in which speed of pedestrians in the ramp was not considered: this sce- a cell was occupied by agent, that either moved out of the cell nario should be therefore considered as a best case situation, or remained stuck in there. More precisely, diagrams show the since pilgrims actually flow through the ramp more slowly relative frequency of the above events on the whole simulation than in our simulation. Consequently, we will not discuss here time. The three metrics are depicted graphically following the the changing of the travel time between the waiting boxes same approach: the background color of the environment is Fig. 4. Space utilization diagrams related to the three alternative simulated scenarios. black and obstacles are red; each point associated to a walkable the scenario (also shown in the legend). For instance, in all area (i.e. a cell of the model) is painted in a shade of gray diagrams in the third row the points of space close to the ramp according to the value of the metric in that specific point. The entrance are white or light gray, while the space of the waiting black color is therefore associated to point if the environment area from which the second group starts is black in the first in which the related metric is 0; the white color is associated column, since the group is not present in the related situation to the point in which the metric assumes the highest value in and therefore that portion of space is not actually utilized. The difference between the first and second scenario is [2] R. Challenger, C. W. Clegg, and M. A. Robinson, “Understanding not apparent in terms of different values for the maximum crowd behaviours: Supporting evidence,” http://www.cabinetoffice.gov. uk/news/understanding-crowd-behaviours, 2009. space utilization metrics (they are actually slightly lower in [3] M. Schreckenberg and S. D. Sharma, Eds., Pedestrian and Evacuation the second scenario), but the area characterized by a medium- Dynamics. Springer–Verlag, 2001. high space utilization is actually wider in the second case. The [4] D. Helbing and P. Molnár, “Social force model for pedestrian dynamics,” Phys. Rev. E, vol. 51, no. 5, pp. 4282–4286, May 1995. third scenario is instead characterized by a noticeably worse [5] E. T. Hall, The Hidden Dimension. Anchor Books, 1966. performance not only from the perspective of the size of the [6] J. Was, “Crowd dynamics modeling in the light of proxemic theories,” area characterized by a medium-high space utilization, but also in ICAISC (2), 2010, pp. 683–688. from the perspective of the highest value of space utilization. [7] L. Manenti, S. Manzoni, G. Vizzari, K. Ohtsuka, and K. Shimura, “Towards an agent-based proxemic model for pedestrian and group In particular, in the most utilized cell of the third scenario, an dynamic,” in WOA, ser. CEUR Workshop Proceedings, A. Omicini and agent was stuck about 66% of the simulated time, compared M. Viroli, Eds., vol. 621. CEUR-WS.org, 2010. to the 46% and 44% of the first and second scenarios. [8] V. J. Blue and J. L. Adler, “Cellular automata microsimulation of bi-directional pedestrian flows,” Transportation Research Record, vol. This analysis therefore confirms that increasing the number 1678, pp. 135–141, 2000. of pilgrims that are simultaneously allowed to move towards [9] A. Schadschneider, A. Kirchner, and K. Nishinari, “Ca approach to the ramp highly increases the number of cases in which collective phenomena in pedestrian dynamics.” in Cellular Automata, 5th International Conference on Cellular Automata for Research and their movement is blocked because of overcrowding. Also the Industry, ACRI 2002, ser. Lecture Notes in Computer Science, S. Ban- utilization of space increases significantly and, in the third dini, B. Chopard, and M. Tomassini, Eds., vol. 2493. Springer, 2002, situation, the whole side of the ramp becomes essentially a pp. 239–248. queue of pilgrims waiting to move towards the ramp. Another [10] S. Bandini, S. Manzoni, and G. Vizzari, “Situated cellular agents: a model to simulate crowding dynamics,” IEICE Transactions on Infor- phenomenon that was not highlighted by the above diagrams mation and Systems: Special Issues on Cellular Automata, vol. E87-D, is the fact that groups face a high pressure to mix when no. 3, pp. 669–676, 2004. reaching the entrance of the ramp, which is a negative factor [11] C. M. Henein and T. White, “Agent-based modelling of forces in crowds.” in Multi-Agent and Multi-Agent-Based Simulation, Joint Work- since crowd management procedures adopted in the scenario shop MABS 2004, New York, NY, USA, July 19, 2004, Revised Selected are based on the principle of preserving group cohesion and Papers, ser. Lecture Notes in Computer Science, P. Davidsson, B. Logan, keeping different groups separated. According to these results, and K. Takadama, Eds., vol. 3415. Springer–Verlag, 2005, pp. 173–184. [12] S. R. Musse and D. Thalmann, “Hierarchical model for real time the management of the movement of group of pilgrims from simulation of virtual human crowds,” IEEE Trans. Vis. Comput. Graph., the tents area to the ramps should try to avoid exceptions to vol. 7, no. 2, pp. 152–164, 2001. the waiting box principle as much as possible. [13] W. Shao and D. Terzopoulos, “Autonomous pedestrians,” Graphical Models, vol. 69, no. 5-6, pp. 246–274, 2007. VI. C ONCLUSIONS [14] S. Paris and S. Donikian, “Activity-driven populace: A cognitive ap- proach to crowd simulation,” IEEE Computer Graphics and Applica- The paper has discussed a research effort aimed at investi- tions, vol. 29, no. 4, pp. 34–43, 2009. gating the implication of the presence of groups in pedestrian [15] M. Moussaı̈d, N. Perozo, S. Garnier, D. Helbing, and G. Theraulaz, and crowd dynamics. In particular, the paper has shown a “The walking behaviour of pedestrian social groups and its impact on crowd dynamics,” PLoS ONE, vol. 5, no. 4, p. e10047, 04 2010. sample situation in which data coming from experimental [Online]. Available: http://dx.doi.org/10.1371%2Fjournal.pone.0010047 observations were used to calibrate and validate a simulation [16] S. Xu and H. B.-L. Duh, “A simulation of bonding effects and their model that correctly captures some aspects of the impact of impacts on pedestrian dynamics,” IEEE Transactions on Intelligent Transportation Systems, vol. 11, no. 1, pp. 153–161, 2010. groups in the overall system dynamics. The validation was per- [17] S. Sarmady, F. Haron, and A. Z. H. Talib, “Modeling groups of formed considering both travel times and other data gathered pedestrians in least effort crowd movements using cellular automata,” in actual experiments and also by comparing the achieved fun- in Asia International Conference on Modelling and Simulation, D. Al- Dabass, R. Triweko, S. Susanto, and A. Abraham, Eds. IEEE Computer damental diagram with existing results from the literature. In Society, 2009, pp. 520–525. addition, a real-world case study was also described: this work [18] F. Qiu and X. Hu, “Modeling group structures in pedestrian crowd considered a train station in which different policies for crowd simulation,” Simulation Modelling Practice and Theory, vol. 18, no. 2, pp. 190 – 205, 2010. management were compare adopting space utilization metrics. [19] R. A. Rodrigues, A. de Lima Bicho, M. Paravisi, C. R. Jung, L. P. Ma- The achieved results are in tune with observations carried out galhães, and S. R. Musse, “An interactive model for steering behaviors on the field and the model is able to reproduce phenomena of groups of characters,” Applied Artificial Intelligence, vol. 24, no. 6, related to group behaviours in pedestrian simulation. pp. 594–616, 2010. [20] J. Tsai, N. Fridman, E. Bowring, M. Brown, S. Epstein, G. A. Kaminka, Future works are aimed at modeling and simulating more S. Marsella, A. Ogden, I. Rika, A. Sheel, M. E. Taylor, X. Wang, complex group structures, such as hierarchical group structures A. Zilka, and M. Tambe, “Escapes - evacuation simulation with children, (e.g. families, friends, elderly with accompanying persons authorities, parents, emotions, and social comparison,” in Proc. of 10th Int. Conf. on Autonomous Agents and Multiagent Systems – Innovative inside larger groups) and their implications on overall system Applications Track (AAMAS 2011), Tumer, Yolum, Sonenberg, and dynamics, validating results both quantitatively and qualita- Stone, Eds., 2011, pp. 457–464. tively with specific reference to the morphology assumed by [21] S. Manzoni, G. Vizzari, K. Ohtsuka, and K. Shimura, “Towards an agent- based proxemic model for pedestrian and group dynamics: Motivations the group in medium and high density situations. and first experiments,” in Proc. of 10th Int. Conf. on Autonomous Agents and Multiagent Systems – Innovative Applications Track (AAMAS 2011), R EFERENCES Tumer, Yolum, Sonenberg, and Stone, Eds., 2011, pp. 1223–1224. [1] M. Batty, “Agent based pedestrian modeling (editorial),” Environment [22] S. Bandini, F. Rubagotti, G. Vizzari, and K. Shimura, “An agent model and Planning B: Planning and Design, vol. 28, pp. 321–326, 2001. of pedestrian and group dynamics: Experiments on group cohesion,” in AI*IA, ser. Lecture Notes in Computer Science, R. Pirrone and [26] A. Bonomi, L. Manenti, S. Manzoni, and G. Vizzari, “Makksim: Deal- F. Sorbello, Eds., vol. 6934. Springer, 2011, pp. 104–116. ing with pedestrian groups in mas-based crowd simulation,” in WOA, [23] A. Willis, N. Gjersoe, C. Havard, J. Kerridge, and R. Kukla, “Human ser. CEUR Workshop Proceedings, G. Fortino, A. Garro, L. Palopoli, movement behaviour in urban spaces: Implications for the design W. Russo, and G. Spezzano, Eds., vol. 741. CEUR-WS.org, 2011, pp. and modelling of effective pedestrian environments,” Environment and 166–170. Planning B, vol. 31, no. 6, pp. 805–828, 2004. [27] A. Schadschneider, W. Klingsch, H. Klüpfel, T. Kretz, C. Rogsch, and A. Seyfried, “Evacuation dynamics: Empirical results, modeling and [24] K. Nishinari, A. Kirchner, A. Namazi, and A. Schadschneider, “Extended applications,” in Encyclopedia of Complexity and Systems Science, R. A. floor field ca model for evacuation dynamics,” Special Issue on Cellular Meyers, Ed. Springer, 2009, pp. 3142–3176. Automata of IEICE Transactions on Information and Systems, vol. E84- [28] J. J. Fruin, Pedestrian planning and design. Metropolitan Association D, 2003. of Urban Designers and Environmental Planners, 1971. [25] H. Klüpfel, “A cellular automaton model for crowd movement and egress simulation,” PhD Thesis, University Duisburg-Essen, 2003.