Simulating the movement of wild meerkats (Suricata suricatta)

(1)

Simulating the movement of wild meerkats

(Suricata suricatta)

The social force model applied to foraging meerkats

Noa Visser Karlijn Limpens Kiki de Waart

10531610 10615342 10653481

Abstract

In this research the social force model is modulated to be applied to a group of foraging wild meerkats. The social structure within a group of meerkats is needed for this modulation. The different forces in the model are adjusted in order to implement this social structure within the model. This modified model are simulated in NetLogo. Two important features are modelled: the division of the group between leaders and subordinates and the differences in dominance between individuals. In this modelelate realistic movement of meerkats, group cohesion and a key role for the leaders in direction and speed of the group should be detectable. Thereby, the speed of the individuals should be around their average speed in real life and no faster than their maximum possible speed. During simulation the meerkats made unrealistic big steps. It can be concluded that the model does not show realistic movement. Reasons for these results can be found in the biological assumption made and the implementation of the model.

(2)

Introduction

With the current knowledge of group movement and artificial intelligence, group movements can be described in models and represented in simulation. Movements in models are described by simple rules and laws. With these models it is possible to simulate real life movements. The models give an insight in factors that influence individual movement behaviour (Li & Jiang, 2014) and make it possible to detect certain patterns in movement. The movements of large flocks of starlings is an example of modelled animal movement (Hemelrijk & Hildenbrandt, 2015).

Moving behaviour of humans has also been captured in models. One of these models, the social force model, describes the movement of pedestrians. Helbing (1995) used a model which simulates the movement of gas molecules and applied this to the movement of pedestrians. The model is frequently used in simulations of pedestrians in different situations to predict their movement. For example, this can be useful in constructing the movement of people in crowded areas (Saberi, Aghabayk & Sobhani, 2015).

Li and Jiang (2014) succeeded in creating a working model for the foraging behaviour of sheep, based on Helbings social force model (Li & Jiang, 2014). Besides this model there has been no research in which the social force model is applied to animals. In this research the social force model (Helbing, 1995) will be applied to the movement of foraging wild meerkats (Suricata suricatta). An important difference between meerkats and sheep is the social structure of the group. Every group consists of a dominant pair, which lead the group, and subordinates (Madden, Drewe, Pearce & Clutton-Brock, 2011). Also, there are differences in dominance between different individuals. The differences in these social structures could result in a different movement of the individuals within a group, due to the role of a leader and differences in dominance (Madden, Drewe, Pearce & Clutton-Brock, 2011). Because sheep and crowds of people do not have leaders within the group, meerkats are an interesting species to modulate. Applying the model to a group with a leader has not been done before. The role of the leader strongly influences the movement of the group. If the model works with the implementation of a leader, this could mean that the social force model is applicable to different types of group structures. The crucial position of leaders in the movement of a group of meerkats makes it an interesting species to test whether the social force model can be modulated to simulate the movement of a group with a leader. Also, creating a model that describes the movements of meerkats, creates an opportunity to research the different influential factors on the movement within the group by modifying the model.

An interdisciplinary approach is required in the application of the social structures of meerkats to the social force model and formulation of a new model. To create this model, understanding of both the social force model and the foraging behaviour of meerkats is needed. To accurately describe this behavior, the social structures of meerkats need to be thoroughly understood. This requires biological insights. The formulas used in the Social

Force Model are based on the physical description of gasses. Physics is needed to

understand the formulas. For the simulation of the model the movements of meerkats and their social structures need to be described in the formulas. This requires an interdisciplinary approach in which biology and physics cooperate. The formulas of the social force model will be used as a basis for the new formulas. The general idea of the social force model will be explained in the theoretical background. The modified formulas are used to create a simulation of meerkat movement while foraging in Netlogo (Wilensky, 1999). To apply the new formulas in a useable model, artificial intelligence is needed. The behaviour and foraging circumstances of meerkats need to be simplified to create an efficient model. This is done by presuming several assumptions which are partly obtained from correspondence with Joah Madden, who has done a lot of research on meerkats and their social structures.

This research will find out whether it is possible to make a modulation of the foraging behaviour of meerkats based on the social force model. To be able to describe this, it is necessary that the movements of individual meerkats in the simulation reflect realistic

(4)

movements. These realistic movements should be comparable to the movements of wild meerkats. To compare these movements, an interdisciplinary cooperation between artificial intelligence and biology is required. The exact criteria confirming realistic movement will be specified in the “Method” section. If the model does not represent realistic movements, it will be adjusted in order to detect the causes of the unrealistic movements. These adjustments create several simulations under different circumstances. Although the model may not be a perfect reflection of reality, it could be useful for different purposes within the research of meerkats.

Theoretical background

To modify the Social Force Model such that it is applicable on meerkats, an understanding is required of both the Social Force Model as of the social structures and movements of meerkats. Therefore it is necessary to approach the modification interdisciplinary, using biology and physics. The social structures and movements of meerkats have to be implemented in the formulas of the social force model. In this section the knowledge need for these modifications is elaborated.

The social force model (Helbing, 1995) is based on the physics which describes the movement of gases. The formulas which describe the movement of gases are modulated to be applicable on humans by introducing new parameters. These parameters depend on distances between pedestrians and their environment. For every pedestrian there are also constants, such as the maximal physical velocity of a pedestrian, that influence the ‘social’ forces on the different pedestrians. The ‘social’ forces are as impulses that change the walking direction and velocity of the pedestrians. These social forces are divided in repulsive forces and attractive forces. An example of attractive forces are buskers. Pedestrians could want to slow down and listen to them. An example of repulsive forces are aggressive appearances of other pedestrians. One would move away from them, rather than move closer to them. The essence of the model is that the sum of all the different ‘social’ forces, which are working on a single pedestrian, determine the direction and velocity of the pedestrian.

The social force model has been used to simulate all kinds of different situations of humans. For example, Saberi, Aghabayk and Sobhani (2015) used the social force model to simulate the spatial fluctuations of pedestrian velocities in bidirectional streams and explored the effects of self-organization. They describe the specific situation in which there are two streams of pedestrians walking through a passageway past each other in opposite direction. They found that the density of pedestrians is influenced by the number of walkways in a crowd. With the use of the social force model they found that the walking behaviour of pedestrians is self-organised. The increment in walkways due to a higher density found in the model, is also found in real life observations of the walking behaviour of pedestrians.

The social force model has often been applied on pedestrians. In contrast, there has been almost no research simulating the movement of animals using the social force model, even though the social force model offers the possibility to be adapted to animals since it is based on the physics of gas. As mentioned before, the social force model has been modified to simulate the movement of sheep within a flock while grazing (Li & Jiang, 2014). In this research, the attractive force depends on the attractiveness of patches of grass. The repulsive forces mainly depend on the maximal and minimal comfortable distance between individuals. Li and Jiang (2014) did not include influences of dominance in their model.

An important difference between sheep and meerkats is the social structure within the group. This social structure within the group makes it interesting to adjust the social force model on meerkats instead on a sheep flock. According to Madden et al (2011), individual meerkats interact with each other within different social networks. They live in groups with two dominant individuals, of which one male and one female, and 2 to 50 subordinates. The dominant pair are non-related and are also the only ones within the group that reproduce (Young, Levesque, Harrison & Clutton-Brock, 2011). The subordinates fulfil different tasks, such as grooming the offspring of the dominant pair and guarding the nest while the rest

(5)

forages (Sharp & Cutton-Brock, 2011). The group leaves the burrow to forage for 5 to 8 hours a day. When they forage, they are more vulnerable because they have to dig 20 centimeter in the sand with their heads down while eating. To decrease vulnerability, they frequently stand up and look around to detect danger, an activity called ‘guarding away’. Meanwhile they produce close calls to communicate with each other. Half of the time one of the subordinates climbs to a higher position, such as a tree, to guard. When this is the case, the others are less vigilant because the guarding subordinate will produce alarm calls when it sees a danger approaching (Clutton-Brock, O’Rain, Brotherton Gaynor, Kansku, Griffin & Manser, 1999).

The different roles that the individuals can fulfil within the group influences the social interactions between the meerkats, such as dominance (Madden et al, 2011). This will influence the movement of the meerkats and should be accounted for while forming the model.

Methods

This section explains the method that has been used for this research. First, a detailed description of the realistic movement that should be seen in the simulation in order to fulfil the hypotheses is given. Next, the biological assumptions that were needed to form the model will be examined. These assumption were made to simplify the model. Then follows an explanation of the formulas that express the social forces that are used in this model. The formulas are explained from both a physical and biological point of view. Also, an explanation is given about how the the formulas will be used in the modeling of the three different situations.

Realistic movement

In this research, the model was tested on realistic movement of the meerkats. This means that the individuals within the simulation should move according to expectations based on their natural movement while foraging. The simulation should fulfill three qualifications in order to present this natural movement:

1. The meerkats will move with their natural speed. This means that they will not move any faster than 7.06 m/s which is their natural maximum speed. Secondly, their average speed should be around 3.31 m/s (Bousquet, Sumpter & Manser, 2011). 2. Meerkats live and forage in a group (Clutton-Brock et al, 1999). From video material

of groups foraging in nature, it can be stated that they stay close together within a certain distance. This group movement should be reflected by the simulation.

3. During foraging the group will follow leading individuals that decide when and where to go (Bousquet and Manser, 2011). In our model, it should be detectable that the group follows these leading individuals.

In order to recognize realistic movement, the simulation should reflect these three criteria. In the sections Formulas there is described how the different forces in the model are adjusted so the model contains these qualifications.

Assumptions of the model

Several assumptions were made so that the model did not become too complex. The more parameters that were considered within the model, the more complex the model would become. Because of the limited amount of time and knowledge, assumptions were made to be able to generate a conclusion. By doing so, the model was not a perfect reflection of reality. The model could still be a good approach and can be used as a starting point for further research.

Based on correspondence with Madden, some general assumptions were made. The first assumption was that the food sources which the meerkats look for were evenly and randomly distributed in the environment (comm. pers., 23 November 2015). The model

(6)

became less complex because all the meerkats had an even chance on every place to find food. Thereby, the distribution of different food sources is not known.

Thereby, there was assumed that handling time of the food was not taken into account and individuals avoid coming too close to each other. When a meerkat finds food, he will start digging. For other meerkats there is no reason to go to this place, unless they are looking for competition. Although the digging actions were not modelled so no arrestment for food sources will be seen in the simulation, there was assumed that individuals are avoiding coming too close to each other unless they are competitive. This is why they spread through the environment instead of moving to the same place (comm. pers., 23 November 2015). As Madden stated, not all food sources need the same treatment. Some can be eaten straight away, others have to be prepared and some can be shared (comm. pers. 23 November 2015). To model this, a lot of extra variables are needed. This would be too detailed and time consuming for this research.

Next assumption was that danger is sporadic. Meerkats are endangered by several predators such as snakes, raptors and some mammals. They have distinct reactions on them, influenced by the type of predator and the presence of other group members (Graw & Manser, 2007). According to Madden it is possible to form a model without predators involved because they do not occur frequently (comm. pers., 23 November 2015). Thereby, meerkats that are foraging are still aware of potential danger. This means that “guarding away” actions should be modelled as well. To prevent this, it is assumed that a raised guard is present so that the rest of the group is less vigilant and these actions can be neglected.

At last, there was assumed that the simulations takes place in the winter season, which is the nonbreeding season. All group members are foraging for themselves so they do not have to find food to bring to the puppies.

The modelled group consisted of 18 individuals divided in three subgroups. This number is chosen because it is close to the average of 2 and 50 individuals within a group and it can easily be divided in three groups. Thereby, the simulation was not too chaotic so the movements of individuals were clear. The first group consisted of the dominant pair and the other two groups of eight subordinates in each one, of which the two subordinates groups have a difference in dominance. How this differences are modelled can be found in the section “Formulas”. Bousquet and Manser (2011) found that leadership during foraging, consisting of initiating the action and determining the direction, can be done by every individual in the group. Leadership depends on an individual's foraging success of previous days and a female's reproductive state. This means that the dominant pair are not per se the leaders during foraging. Increase or decrease in speed or change of direction of the leaders during foraging will result in change of speed or direction for the whole group (Bousquet & Manser, 2011). In this model, the dominant pair were chosen to be the leaders of the foraging group. The features of each group were modelled differently. What type of features will be discussed in the next section.

The modelled group moved over a grid of 64 x 64, which represented a 320 x 320 meters area in real life. According to Madden the distance between two holes was about one kilometer. In the simulation, the group moved in the area between two holes. It was assumed that in an area of 320 x 320 meters no holes occurred.

Formulas

Base formulas

In this section all the formulas that were used in our model for meerkats will be explained. In the new model there were several forces taken in account to form a realistic model of movement while foraging. The sum of these forces (formula 2) led to the acceleration of the meerkats. The new position of each meerka was calculated (formula 1) from the current position and the acceleration caused by the total force (Taylor, 2005). Every 1/30 second the new position for all the meerkats was calculated.

(7)

Every formula for the forces will first be explained from a social force model perspective. The variables used to create formulas for ‘social’ forces can vary from distance between individuals and individual speed to dominance. Each force has a biological meaning for the behaviour of meerkats. In the implementation of the formulas, methods from the Netlogo libraries “Gas in a Box” (Wilensky, 1997), “Flocking” (Wilensky, 1998) and “Shepherds” (Wilensky, 1998) were used. It was chosen to implement the formulas as literal as possible.

The first formula (formula 2) was a sum of all the social forces. The first term was an attractive force towards the leaders. The second term was a force giving initial speed to the leaders. The last term was a repulsive force between individual meerkats. The forces in the first and third term were summations of all the attractive and repulsive forces caused by each individual upon each individual. To simplify the simulation, each meerkat took the closest individual of each dominance group in account for the calculation of the force for that individual. For example, for one of the leaders the forces are caused by the other leader, the nearest individual of the more dominant group of subordinates and the nearest individual of the less dominant group of subordinates was summed. When the distance between two individuals was 0 the model made this distance 0.5. Otherwise the force calculated with that distance of 0 was not defined unambiguously.

The second formula (formula 3) was an attractive force. The force depended on the distance between a meerkat and its nearest leader. The formula implemented that the force became larger when a meerkat was too far from the nearest leader. The force had a direction pulling towards the nearest leader. The reaction speed of each of the meerkats was set to 1. This meant that all the meerkats reacted with the same speed to change in the distance between themselves and the nearest leader.

Attractive Force

During simulation, the meerkats were moving in a group. That meant that they could spread within a limited distance from each other and would only split up in case of danger, which is not included in the model. To make sure the group would not spread too far, there

(8)

was a maximal desired distance between each individual and its nearest leader. This maximal distance was 15 meter between subordinates and their leader. The maximal desired distance was 8 meters between the two leaders . These distances were based on observation. There was no explicit data available about the spreading of a group of meerkats,but observation showed that there was no strong cohesion in a group of foraging meerkats. Thereby NetLogo displayed a surface of 320 x 320 meters, so the group movement should be detectable within these maximal distances.

Base Speed

The leading meerkats determined the direction and base speed of the entire group of meerkats. This was implemented in a force (formula 4) which only applied on the leaders. This force depended on a desired speed to which the current speed was corrected. The desired speed also contained the direction in which the leaders were moving. The base speed of the leaders was determined by this force. Since the meerkats wanted to stay close to the leaders, the subordinates had to constantly adapt their speed and direction to be able to follow their leaders. The reaction speed of the leaders is set to 1, the same value of the reaction speed in formula 3.

Group cohesion is partly retained by the sounds meerkats use to communicate. Besides the vocal alarm calls that they use in case of danger, they also produce close calls. These calls are produced while foraging and are probably used to remain relatively to each other. Moving calls can change the speed of the group. This will only happen when at least two other individuals also start producing these calls. As a consequence, the group will start moving faster. Their average speed increases from 3.31 m/s to the maximum of 7.06 m/s. The direction of the movement does not change by these calls. Without any moving calls, but only close calls, a group will not move faster than 2.69 m/s (Bousquet, Sumpter & Manser, 2011). The variation in speed will be taken into account in this formula. To make sure the group moves from their starting point the leaders will be given an initial speed of 2.69 m/s and a desired direction of one, since there is no desired direction in the beginning.

(9)

Repulsive Force

Third, there is a repulsive force that depends on the distance between 2 individual meerkats. This force can be interpreted as the personal space of the meerkats. When two meerkats are too close to each other this repulsive force increases, so that the meerkats are pushed away from each other. The formula (formula 5) for this force is influenced by the dominance of the meerkats. This manifests itself in the reaction speed Τα, which varies between 1 and 5. When there is a hostile relation between two meerkats, the reaction speed increases, resulting in a small Τα.

It is important to include competition between individuals in the model. If an individual has higher dominance than another individual, there is a greater chance that the weaker individual will avoid the stronger one. Dominance could give a good indication of the influence that the character of each meerkat has on other meerkats’ movement. Elaborate research of Madden et al. (2011) provides an overview of the social interactions within groups of meerkats, including dominance. They found more dominance interactions between males and males than between males and females. This also occurred between females in one of the research groups. Thereby, they found that the dominant pair gave higher amounts of dominance to other meerkats, and did this to more meerkats (Madden et al., 2011). The consequence for our model is that there has to be a difference between the individuals of the group and the amount of competition between them. This is why the group of subordinates will be divided in two; dominant individuals and subordinate individuals. The leading pair will have the highest dominance. Half of the group of subordinates will have an intermediate dominance, and the other half a low dominance. When an individual with a high dominance approaches another one with less dominance, this individual will try to avoid contact which could result in intimidation. This will only be the case when they are closer than 0.5 meter away of each other; this measure is chosen because Madden et al. (2011) used this in their research as an indication of food competition, so this could be an indication of the personal space that meerkats like to retain.

(10)

Figure 1: Dominance interactions between the different groups.

Implementation NetLogo

The formulas were implemented in NetLogo in the most literal way possible. The formulas had to be adjusted to NetLogo, since NetLogo is a round world. This means that if an individual leaves the display at a certain place, the individual will re-enter the display opposite of the place where the individual left the screen. For example, in figure 2 the green meerkats were situated in the lower left corner of the display. In figure 3, one meerkat (green) moved to the left center of the display. Following the green line, one can conclude that the meerkat moved outside of the display, from the place where he left the screen, reentering the display on the opposite side of the screen, as if the display was a globe instead of a flat area. This round world is not possible in real life,making it necessary to bound the area. In the simulation, whenever a force moved a meerkat off screen, the meerkat was moved to the borders of the screen, making the x or y coordinates 0 or 64. For example if a force moved a meerkat to coordinate (3, 65) the coordinates were readjusted to (3,64).

(11)

Figure 2: Unbound world, the green individuals were situated in the lower left corner of the display

Figure 3: Unbound world, the green individuals have moved to the middle of the display, via the upper left corner of the display.

Results

A group of eighteen individuals was simulated, of which two leaders, eight more dominant subordinates and eight less dominant subordinates. The meerkats were represented by arrows in the simulation. The orange arrows represented the leading individuals and the green arrows and turquoise arrows represented the two different groups of subordinate meerkat. To analyse whether the movement of the meerkats was realistic, the movements were tracked of one individual of each group. Whenever one of the tracked meerkats moved, a line was drawn on the display. The line was drawn in the same color as the meerkat was displayed in. Two different simulations were made. The location in which the meerkats were situated at the beginning of the situations were different in these situations.

At the beginning of the first simulation, the meerkats were situated at the left lower corner of the display. This was chosen as starting point, since it made all coordinates positive in Netlogo. Also, since the meerkats move in a group, it is likely that they would enter an area together.

(12)

Figure 4: Bound world, startingpoint left bottom corner, position at t = 0.

(13)

figure 6: Bound world, startingpoint left bottom corner, position at t = 2∆t.

figure 7: Bound world, startingpoint left bottom corner, position at t = 3∆t.

During the simulation, all meerkats moved from the left lower corner to one of the four corners of the display. They kept moving from one corner to another. The first steps of this movement is shown in figures 5 to 7. Figure 5 represents the first step of the meerkats after the begin situation. The meerkats moved close to each other. One leader was moving ahead of the group, along the lower border of the display. Figure 6 represents the second step of the meerkats. One leader moved to the right lower corner of the display. Two blue subordinates followed. One blue subordinate moved up, drawing a line. In figure 7 is seen that this blue meerkat moved down, drawing a line towards the lower left corner. All meerkats were situated in one of the four corners of the screen.

(14)

To test whether this movement towards the corners was caused by the start position of the meerkats the meerkats were also simulated in a bound world, with random starting points. The first movements of this simulation is shown in figures 8 to 12.

figure 8: Bound world, random starting point, position at t = 0.

(15)

figure 10: Bound world, random starting point, position at t = 2∆t.

(16)

figure 12: Bound world, random starting point, position at t = 4∆t.

Figure 8 shows where all meerkats were situated at the start of the simulation. In figure 9, all meerkats moved to either the lower left corner of the upper right corner. In figures 10 to 12, the meerkats moved step by step to one of the four corner of the display. Despite the disabling of moving off screen and the random start points, the meerkats were still able to move unproportional steps. For example, some meerkats were situated in the lower left corner of the display in figure 10. In figure 11, three blue subordinate meerkats moved almost one third of the screen up. The meerkats had moved from their positions in figure 10 to their positions in figure 11 in one step. The three blue subordinate meerkats had moved almost 100 meters in one step in real life. This is not possible for meerkats.

Conclusion

In this research a model is proposed that was supposed to simulate the movements of meerkats while foraging. To be verified, the model should have shown realistic movements of the individuals. The model should show meerkats moving with an average speed of 3.31 m/s with a maximal speed of 7.06 m/s (Bousquet, Sumpter & Manser, 2011), group movement and individuals that follow their leaders. This has been tested in two different simulations. In the first simulation, the meerkats begin in the lower left corner of the display. The meerkats move towards the corners of the display during simulation. In the second simulation, the meerkat start with random begin situations. The meerkats also move towards the corners during simulation. The movement towards the corners seems to be caused by the size of the forces. The meerkats are simulated on a 64 x 64 grid. To bound the world, the coordinates are readjusted if the forces move them outside the grid. Since the meerkats are able to move unrealistically large steps, the forces need to be very large as well. If the forces are too large, the meerkats keep moving along the grid. It can be concluded that as long the forces are this big, it is impossible to simulate a realistic movement of the meerkats.

(17)

Discussion

There are several possible reasons why the hypothesis was not validated. Firstly, this can be found in the biological assumptions made on the distribution of food, the occurrence of danger and use of subgroups representing the dominance of several meerkats. Secondly, the use of Netlogo restricts the simulation of the model. Each of these reasons are suggested to be taken into account for further research.

Biological assumptions

Because of the biological assumptions that are made, the model became less realistic. These assumptions were important for the feasibility of this research, and although the model is not working as expected, it is important to take their impact on the formed model in account for future research.

First of all, the assumption made about evenly and randomly distributed food sources is not realistic. This assumption causes dispersal of the individuals because they are not attracted to one point in space. In reality, this is quite different. Food is not evenly distributed, and according to Madden there are also different types of food. Some of these need to be handled by two meerkats or more, and others can be shared or have to be brought to the sleeping holes (comm. pers. 23 November 2015). In the model is assumed that they are not shared and do not need any extra handling time or actions. This will change the simulation, because competition will be different and collaboration between individuals can be seen.

Thereby, assumed is that danger is sporadic and the meerkats will not be continuously aware of danger because there is a raised guard. Although they do become less vigilant, they are still aware and ‘guarding away’ from time to time. This is not taken in account, but if it would have been modelled it could be interesting to make differences in vigilance of different individuals. The leading individual for example could be ‘guarding away’ more than the rest, because it decides where the group will move. Reactions of the group on danger would also be an interesting simulation for the future.

Another simplistic feature is that the group is split in three groups in which the individuals have the same characters. In reality, this is of course much more complicated. Further research could give each actor its own character, which would result in whole different dynamics. Other characteristics then dominance, such as grooming or amicable actions, could be modelled for each individual. Thereby, the characteristics are static now, but in reality roles of the individuals can change, such as the dependency of leadership on the success in feeding in the past and. More extended versions of this model could take this in account as well.

Simulation

During simulation, the model shows several problems. The meerkats keep moving towards the corners of the display. This might be caused by restrictions in the use of NetLogo. The individuals are only able to calculate the forces of several individuals, instead of all of them. Also, the meerkats have to be restricted from moving off screen. Lastly, the calculated forces are translated to an x and a y coordinate to determine the step an individual has to make. The translation of the forces may cause the unproportional steps the meerkats are able to make.

The first restriction on the implementation of the formulas during simulation, caused by the usage of Netlogo, is the calculation of distances between individuals. In Netlogo, the individual distance between one individual and all other members of the group can not be calculated easily. Netlogo does not contain a function in which this can be done and also can not easily work with lists of individuals to calculate the different distance values. In the simulated situations each individual only calculates the forces of the nearest member of

(18)

each group. If more individuals are used to calculate the forces, the resulting force could be larger and more accurate.

Secondly, to restrict the meerkats of moving off screen, the x and y coordinates were readjusted if they were situated outside the display. In real life, meerkats would be able to move on and off a field. Since NetLogo represents a round world and meerkats live in a flat area, it is necessary to restrict the meerkats. This restriction may cause that meerkats kept moving towards the corners of display.

Lastly, the calculated total force is translate into an x and y coordinate. This is done by formula 1. The total force is the acceleration and will be multiplied by t2= 1900. This makes the total addition of the acceleration smaller. To implement this formula into NetLogo the initial speeds gets translate into the step a meerkat would make in t in its current velocity. The meerkats still make very big steps, which suggests that the initial step of the meerkats is very large and causes the unproportioned movements. In this research, the probable influence of the initial step on the movements of the meerkats cannot be proved. In further research this could be explored.

This research does not verify or contradict whether the social force model can be used to simulate the movement of meerkats. In the social force model individuals simulated have a determined direction in which they move. Meerkats do not have a determined direction during foraging, but their direction is determined by each individual of the group and adjusted during movement. They do not have one leader to solely follow, as sheep have, neither do they have an individual destination to walk to, as humans have. It may be that the complex social structures of meerkats make them unfit to be applied on the social force model. This can be verified in further research, if the steps made by the meerkats will not be so big and food sources are put in the field of the simulation.

(19)

References

Bousquet, C. A., & Manser, M. B. (2011). Resolution of experimentally induced symmetrical conflicts of interest in meerkats. Animal Behaviour, 81(6), 1101-1107.

Bousquet, C. A., Sumpter, D. J., & Manser, M. B. (2011). Moving calls: a vocal mechanism underlying quorum decisions in cohesive groups. Proceedings of the Royal Society of London B: Biological Sciences, 278(1711), 1482-1488.

Clutton-Brock, T. H., O'Riain, M. J., Brotherton, P. N. M., Gaynor, D., Kansky, R., Griffin, A. S., & Manser, M. (1999). Selfish sentinels in cooperative mammals. Science, 284(5420), 1640-1644.

Graw, B., & Manser, M. B. (2007). The function of mobbing in cooperative meerkats. Animal Behaviour, 74(3), 507-517.

Helbing, D., & Molnar, P. (1995). Social force model for pedestrian dynamics.Physical review E, 51(5), 4282.

Hemelrijk, C. K., & Hildenbrandt, H. (2015). Diffusion and Topological Neighbours in Flocks of Starlings: Relating a Model to Empirical Data.

Li, Z., & Jiang, Y. (2014). Friction based social force model for social foraging of sheep flock. Ecological Modelling, 273, 55-62.

Mares, R., Young, A. J., Levesque, D. L., Harrison, N., & Clutton-Brock, T. H. (2011). Responses to intruder scents in the cooperatively breeding meerkat: sex and social status differences and temporal variation. Behavioral Ecology, arr021.

Madden, J. R., Drewe, J. A., Pearce, G. P., & Clutton-Brock, T. H. (2011). The social network structure of a wild meerkat population: 3. Position of individuals within networks. Behavioral Ecology and Sociobiology, 65(10), 1857-1871.

Sharp, S. P., & Clutton-Brock, T. H. (2011). Reluctant challengers: why do subordinate female meerkats rarely displace their dominant mothers?. Behavioral Ecology, arr138.

Saberi, M., Aghabayk, K., & Sobhani, A. (2015). Spatial fluctuations of pedestrian velocities in bidirectional streams: Exploring the effects of self-organization. Physica A: Statistical Mechanics and its Applications, 434, 120-128.

Taylor, J. R. (2005). Classical mechanics. University Science Books.

Wilensky, U. (1997). NetLogo GasLab Gas in a Box model.

http://ccl.northwestern.edu/netlogo/models/GasLabGasinaBox. Center for Connected

Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.

Wilensky, U. (1998). NetLogo Flocking model.

http://ccl.northwestern.edu/netlogo/models/Flocking. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.

Wilensky, U. (1998). NetLogo Shepherds model.

http://ccl.northwestern.edu/netlogo/models/Shepherds. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.

(20)

Wilensky, U. (1999). NetLogo. http://ccl.northwestern.edu/netlogo/. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.

Simulating the movement of wild meerkats (Suricata suricatta)