ABSTRACT. Group-living animals are often faced with choosing between one or more alternative resource sites. A central question is how a collective decision is taken. This experimental and theoretical review demonstrates that choices can emerge through nonlinear interaction dynamics between « equal » individuals without perfect knowledge or leadership. We explore a number of situations differing in the number and quality of the options, in the type of interactions, and in the number of individuals concerned. The interplay between individual responses to site characteristics and to group-members can give rise to a diversity of patterns of decision-making. We will focus on how the environmental characteristics influence the collective responses and their diversity in different situations. We will also discuss how the individual complexity affects the collective response and how the synergy between artificial agents and organisms is the source of new collective efficiency. Using choice experiments and a theoretical approach, we will show how individuals in a group dramatically outperform the problem-solving ability of a single individual. Our research points towards a generic self-organized collective decision-making process shared by many group living-species and its extension to mixed animal-artificial agents.

ABSTRACT. Swarm robotics is about constructing and controlling swarms of autonomous robots that cooperate to perform tasks that go beyond the capabilities of the single robots in the swarm. In this talk, I will give an overview of recent and ongoing research in swarm robotics in my research lab, IRIDIA, at the Université Libre de Bruxelles. In particular, I will present results obtained with homogeneous and heterogeneous swarms of robots that cooperate both physically and logically in search and retrieval tasks.

ABSTRACT. Collective movement occurs in living systems where the simple movements of individual members of a population are combined to generate movement of the collective as a whole, displaying complex dynamics which cannot be found in the component parts themselves. The plasmodium stage of slime mould Physarum polycephalum displays complex amoeboid movement during its foraging and hazard avoidance and its movement can be influenced by the spatial placement of attractant and repellent stimuli. Slime mould is attractive to robotics due to its simple component parts and the distributed nature of its control and locomotion mechanisms. We investigate methods of automated guidance of a multi-agent swarm collective along a pre-defined path to a goal location. We demonstrate a closed-loop feedback mechanism using attractant and repellent stimuli. We find that guidance by repellent stimuli (a light illumination mask) provides faster and more accurate guidance than attractant sources, which exhibit overshooting phenomena at path turns. The method allows traversal of convoluted arenas with challenging obstacles and provides an insight into how unconventional computing substrates may be hybridised with classical computing methods to take advantage of the benefits of both approaches.

ABSTRACT. Animals in a group have been considered to have a field or zone to avoid a collision among individuals in the group. The repulsive zone was formalized as a symmetry zone i.e. a circle in the theories of collective behavior. We challenge the theories since we believe that animals use anticipation to avoid the collision and the partial evidence has been provided. We investigate a discrete model consisting of velocity-based “oval” repulsive and long-range attractive zones. We show that this model can exhibit highly coherent behavior without explicit alignment force due to asymmetric interaction upon the oval potential. The results would be contributed to future research in collective behavior and robotics.

ABSTRACT. Living in a group can help organisms to find optimal solutions in response to their environment. In this study, we consider the searching strategy of fish (Plecoglossus altivelis, ayu or sweetfish) in an unfamiliar environment. We allow the fish to swim freely in a large shallow tank in a controlled laboratory and we track the fish’s trajectories. We find two main results. One is that all trajectories of individuals show Lévy walk behavior, which is the optimal strategy for resolving the trade-off between exploration and exploitation. This type of behavior was observed for both single individuals and individuals in a group. The second is that the trajectories of individuals were markedly different depending on whether they were alone or in a group. Our results suggest that interactions among fish expand the search area without changing the Lévy walk strategy.

ABSTRACT. Robotic fish are nowadays developed for various types of research, such as bio-inspired robotics, biomimetics and animal behavior studies. In the context of our research on the social interactions of the zebrafish Danio Rerio, we developed a miniature robotic fish lure for direct underwater interaction with the living fish. This remotely controlled and waterproof device has a total length of 7.5 cm with the same size ratio as zebrafish and is able to beat its tail with different frequencies and amplitudes, while following the group of living animals using a mobile robot moving outside water that is coupled with the robotic lure using magnets. The robotic lure is also equipped with an easily rechargeable battery and can be used autonomously underwater for experiments of up to one hour. We performed experiments with the robot moving inside an aquarium with living fish in order to analyze its impact on the zebrafish behavior. We found that the beating rate of the tail increased the attractiveness of the lure among the zebrafish shoal. We also demonstrated that the lure could influence a collective decision of the zebrafish shoal, the swimming direction, when moving with a constant linear speed inside a circular corridor. This new robotic fish design and the experimental results are promising for the fieldof fish-robot interaction.

ABSTRACT. We present a model to generate noise-induced insect movements in a large swarm that are similar to those observed in real-world trajectories. Our approach is based on pre-recorded insect trajectories. We present a novel evaluation metric and a statistical validation approach that takes into account various characteristics insect motions and evaluates well-known noise functions. Finally, we combine Curl noise function with a dynamics model to generate realistic swarm simulations and emergent behaviors of flying insects. We demonstrate their performance on simulating large swarms.

ABSTRACT. This paper proposes a novel method for deploying a wireless mesh network (WMN) using a group of swarm robots equipped with wireless transceivers. The proposed method uses the rough relative positions of the robots estimated by their Radio Signal Strength Indicators (RSSIs) to deploy the WMN. The employed algorithm consists of two parts, namely, (1) a fully distributed and dynamic role decision method among the robots, and (2) an adaptive direction control using the time difference of the RSSIs. In our study, we evaluated the performances of the proposed deployment method and a conventional method in a real environment using 12 real robots. The results of the performed experiments showed that the proposed method outperformed the conventional method with regard to the deployment time, power consumption, and the distances travelled by the robots.

ABSTRACT. The objective of this paper is to develop an index which and when flocking agents are disconnected from a swarm. Each of the flocking agents interacts with neighbor agents in a close distance and determines a motion with relative positions and velocities of the neighbor agents. Therefore, neighbor relations are dynamic and it is difficult to predict an exact motion of each of them. For the objective, we consider an effect of a target agent to a swarm. First, we model a flocking with virtual spring and damper system and develop a disconnection index between the target and another agent with the dynamics. Then, we consider how much the target agent can affect to a group of agents with a graph topology. We verify the index with numerical experiments.

ABSTRACT. Scheduled transportation service is a proper system for mass transportation and it is employed by wide range of transportation modes, such as railway, airline, maritime container shipping and bus. The providers of the service are required to organize effective routes and networks. This paper tackles the problem combining two problems. One is generating Public Transit Network (PTN) as one of the scheduled transportation services. The other is network hub location problem to find out the effective position of node as a hub. The method generating PTN is based on a growing network model and the method for the hub location problem is based on a genetic algorithm. This method can find out effective position of the hub node and lines going through some bus stops before reaching to the hub simultaneously.

ABSTRACT. This paper focuses on the passenger overflow cascades occurred in the bus route network in disaster situations. Cascade means that affected by many passengers overflowing at one station, the passengers at the other station also overflow. The simulations have revealed the following implications: (1) the appearance of passengers overflow can be classified liberally under 4 cases; and (2) our proposed method can optimize the bus route network which is harder to occur the passenger overflow than the conventional one.

ABSTRACT. We report the abstract of our previous research, which theoretically studied collective patterns of swarm dynamics by using statistics and Lyapunov analysis, and individual behaviors in swarms by using speed distributions and individual instability, which is characterized by individual instability exponents (IIEs) we proposed in the paper. The collective patterns are classified by the characteristic quantities of Lyapunov analysis: Lyapunov spectrum, Lyapunov dimension, and stability and instability indices. We show that IIEs is an order parameter which indicates transition of the pattern and the activity of the swarm dynamics.

ABSTRACT. Even though collective animals, e.g., schooling fish and flocking bird, show rapidly synchronized behavior, recently it has been reported that internal structure of such the group is not fixed in time. Such inherent noise, which might be expected to be detrimental for collectivity, plays an important role in facilitating interactions with various neighbors and thereby robust collective motion and information transfer. In our previous work, we studied schools of ayu fish and showed that individual movement relative to the center of the mass of the group displays Lévy walk pattern in which step-length follows power-law distribution with Lévy exponent μ. We here show that schooling fish change internal movement strategy owing to their density, by estimating relative exponent by which the strategy is decided. In particular, we show that Lévy exponent is increased in lower density of the school.

ABSTRACT. I have studied heterogeneous boid models consisting of a large number of boid-type agents divided into a small number of groups. First in this paper, I explain the simplest one, the two-component model. This model generates some typical segmented patterns with a clear interface between two agent groups. These segmented patterns are fragile when additional actions are assigned to agent groups. So I improve this model to enhance the interface, and propose the fixed model and the transition model. Next, the fixed model is modified from the two-component model by adding a third agent group to unify all agent types in a single cluster. The agent type and rate of agent groups are fixed. The fixed model generates enhanced segmented patterns, where the third agents group spreads widely around the interface. Further, the transition model is modified from the fixed model by use of the design method I proposed previously. Agents detect precise position information from the interface. They transmute their type and regulate rate of agent groups based on the position information. The transition model generates enhanced segmented patterns, where the third agents group marks the interface precisely. These enhanced segmented patterns will be useful to improve model’s design further, by assigning additional actions to agent groups.

ABSTRACT. Here, we developed a multi-agent model to describe dynamical pattern formations using pheromones produced by each agent. We compered the balances between exploitation (One site is visited by many foragers) and exploration (many sites are visited by agents) of two algorithms. In both algorithms, agents linearly reacted to pheromone concentrations within their detection fields and changed detection fields depending on the differences between left and right pheromone concentrations. Thanks to this event, we succeeded in agents to produce ordered pattern formations at macro levels even though each agent’s reactions to pheromones were set as linear reactions. However, agents could achieve more moderated pattern formations when they occasionally stopped changing detection fields by considering the relationship between global information (pheromone concentrations) and local information (moving directions of others).

ABSTRACT. The soldier crab appears in great numbers and feeds while wandering en masse during daytime low tide. When they see an approaching object, they screw themselves into the sand at great speed to avoid being caught. The mechanism of formation of mass wandering has not been clarified. In this study, to investigate if the soldier crabs use visual images of neighbors as a stimulus for wandering, dummy crabs were presented to crabs and their reactions were observed. In the experiments, one, two, four, or eight dummies were placed in a circle at regular intervals on a moist sand arena. In addition, an experiment without a dummy was also used. Each crab was placed in the center of the arena and we observed whether it burrowed into the sand or wandered. The proportions of wandering individuals in each experimental treatment were compared with the expected value. Significantly more crabs were wanderers when only two and four dummies were present. This result suggests that soldier crabs might pay visual attention to neighbors and chose burrowing or wandering depending on the distribution of the neighbors.

ABSTRACT. 　The application of robots or drones to search for survivors in disaster sites has recently attracted considerable research attention. In this study, we focus on the behavior of insects and propose a simple path generation algorithm for search and rescue robots. The proposed algorithm is based on the ladybird’s strategy for gathering food. In addition, we optimize its parameters using a genetic algorithm. The effectiveness of the proposed algorithm is demonstrated through simulations.

ABSTRACT. Aggregation is a useful building block behaviour that can allow a swarm of robots to interact with each other or a user more easily. Previous work on swarm robot aggregation has assumed that the capabilities of individual robots are quite limited. We test whether incorporating odometry as an additional capability is helpful and make the argument that odometry is both realizable and biologically plausible. We propose an algorithm called ODOCLUST which takes inspiration from the BEECLUST algorithm but uses a continuously active odometry-based homing process to achieve more tightly packed robot aggregates more quickly than BEECLUST. Initial results in simulation suggest that high-fidelity odometry is not required in order to see these gains.

ABSTRACT. In this paper, we propose a distributed control method to transport a disk shaped object to a desired destination by a group of robot agents without any communication among agents. Using sensor information, the entire group can successfully transport the object to the destination. The process of transportation involves three steps, Approach, Surround, and Transportation. If a target path is given to one robot, leader robot, the other robots will follow the leader while encircling the object among the swarm. Simulation and experiment with real robots are conducted as well to verify the effectiveness of the proposed method.

ABSTRACT. Although the technology for fully autonomous swarms of robots is rapidly progressing, the human operator will continue to play an important role during any swarming mission due to safety, monitoring and control constraints. In this paper, we present the set of features that a Ground Control Interface (GCI) must incorporate to allow monitoring, control and safety of outdoor missions with a swarm of Small Drones (drones of less than 1kg). We extend a widely used GCI by incorporating those features and we demonstrate its usage on a swarm of 10 Small Drones flying outdoor.

ABSTRACT. Swarm robotic systems are usually considered to be very robust as there is no single point of failure. Based on local interactions among robots and between robots and the environment, the swarm is required to complete the task it was designed for and to do that reliably, continuously, and on time. When an individual robot makes wrong decisions, fails to complete its individual task, or just breaks down, the robot swarm might be caught in a deadlock or the desired collective behavior might not emerge at all. How could one guarantee the emergence of desired collective behaviors and how should they correctly emerge from imperfect individuals, that is, faulty robots? This work focuses on the issue of dependability in swarm robotic systems, where dependability of a system is defined as the ability to avoid unacceptable frequent and severe service failures. A general framework on fault classification for swarm robotic systems is proposed based on a case study of sorting in a robot swarm. Different fault tolerance methods in both traditional computing systems and swarm robotic systems are discussed.

ABSTRACT. Recently multi-roter UAV which is known as "drone" is used in many field such as military and industry. Some drones do not need manual control by human and it can perform its mission autonomously. Furthermore, there are some studies on how to control a swarm of drone cooperatively. To control the drones in a group, it is usually more effective to make a leader from the group, as observed in some kinds of wild animals. For this purpose, there are two typical operating algorithms. One of them is Implicit Leadership mechanism and the other one is Leader Selection based on Manipulability. These two algorithms are sensitive to the environmental changes and this makes difficulty in controlling the drones in the real world by these algorithms. Then we tried to make a robust system against the environmental changes by combining these two operating algorithms. As a result, we achieved to make more flexible and efficient system than previous method.

ABSTRACT. Walking is one of the most fundamental activities of human, and there has already been many studies on human walking. However, many of such studies focus on the specific gait patterns of subjects with some symptom from disease or injury, and thus there are not so many studies on the gait patterns of healthy subjects. In the previous study, we analyzed the gait patterns of healthy subjects, and as a result we succeeded to evaluate and classify the healthy gait patterns. In this study, we further analyzed the interaction between healthy subjects walking together, as a preliminary for the analysis on human swarm behavior during military march.

ABSTRACT. Recently, the quality of action cameras have been remarkably improved. In spite of their small size, light weight and durability, they enables the record of high resolution videos. However, high resolution video processing requires much computational resource, and thus it gives too much load on a micro control unit (MCU) implemented on a small scale unmanned aerial vehicle (UAV). Many studies tried to overcome this problem by combining MCU and external processing system, but this causes slow response and limitation in the mobile range of UAV. Therefore, in this study, we developed a hardware using a Zynq board and an action camera that enables high resolution video processing on a small scale UAV to realize autonomous control of the vehicle.

ABSTRACT. Soldier crabs wondering in the lagoon at lower tidal level could make a huge swarm. Since soldier crabs move slowly, and stop and begin to move asynchronously, it is easy to see the swarming mechanism of individual’s contacts. We here show that soldier crabs move based on the pursuit-escape mechanism to some extent, and that they could follow the shadow of the antecessors in a swarm.

ABSTRACT. From colonies of bacteria to swarms of bees and flocks of birds, countless organisms exhibit a swarming behavior based on local, individual decision making. In such species, the information is used efficiently at the group level to reach optimal behaviors in tasks such as food foraging, which allow to overcome noisy sensory inputs and local minima. %Identifying the minimal features of biologically-inspired interacting agents leading to the emergence of such behavior is fundamental to understanding collective behavior, both in artificial intelligence and evolutionary biology. In this paper, we extend an abstract agent-based swarming model based on the evolution of neural network controllers, in order to explore further the emergence of swarming. However, we ground our model in a more realistic setting where information about the resource location made partly accessible to the agents, but only through a highly noisy channel. The swarming is shown to critically improve the efficiency of group foraging, by allowing agents to reach resource areas much more easily by correct individual mistakes in group dynamics. As high levels of noise may make the emergence of collective behavior depend on a critical mass of agents, it is crucial to reach in simulation sufficient computing power to allow for the evolution of the whole set of dynamics. Because this type of simulations based on neural controllers and information exchanges between agents is computationally intensive, it is critical to optimize the implementation in order to be able to analyze critical masses of individuals. In this work, we address implementation challenges, by showing how to apply techniques from astrophysics known as treecodes to compute the signal propagation, and efficiently parallelize for multi-core architectures. The results confirm that signal-driven swarming improves foraging performance. The agents overcome their noisy individual channels by forming dynamic swarms. The measured fitness is found to depend on the population size, which suggests that large scale swarms may behave qualitatively differently. The minimalist study presented in this paper together with crucial computational optimizations opens the way to future research on the emergence of signal-based swarming as an efficient collective strategy for uninformed search. Future work will focus on further information analysis of the swarming phenomenon and how swarm sizes can affect foraging efficiency.

ABSTRACT. During learning a complex task our nervous system self-organizes large groups of neurons into coherent dynamic activity patterns. During this, a cell assembly network with multiple, simultaneously active, and computationally powerful assemblies is formed; a process which is so far not understood. Here we show that the combination of synaptic plasticity with the slower process of synaptic scaling achieves formation of such assemblies. This type of self-organization allows executing, for instance, a difficult, six degrees of freedom, manipulation task with a robot where assemblies need to learn computing complex non-linear transformations and - for execution - must cooperate with each other without interference. This mechanism, thus, permits for the first time the guided self-organization of computationally powerful sub-structures in dynamic networks for behavior control.

ABSTRACT. Despite being the most species-rich, largest, and ecologically significant heterotrophic organisms on earth, unicellulars have been somewhat left out of the recent surge of interest in inter-individual behavioral variation. This may reflect uncertainty on how to deal with such a phenomenon in a unicellular organism that lacks a nervous system. We combine experiments and theoretical models in the true slime mold Physarum polycephalum, and show that behavioral differences underpin some aspects of social interactions. First we demonstrate that cells show a surprising degree of variability in their behavioral repertoire. We identified three distinct behavioral phenotypes; “slow-regular-social”, “fast-regular-social” and “fast-irregular-asocial”. Second we discovered that calcium release and calcium sensitivity are responsible for this variability. Finally, using both experiments and a theoretical model, we show that behavioral differences influence cell-cell interactions. These results provide strong evidence that the true slime molds follow alternative social strategies related to individual behavioral differences. This will aid the understanding of many aspects of unicellulars ecology and evolution.

ABSTRACT. Social insects build various structures such as large nests and underground galleries. Construction is achieved by self-organization rule, whereby colony-level structures emerge from local interactions among members that elicit positive and negative feedback responses [1]. In this building system, stigmergy plays an important role, with local interactions among individuals occurring indirectly through environment modification. The structures produced by this building rule show not only interspecific but also intraspecific variation in size and shape widely. Some studies revealed that intraspecific variation in structures is created by exogenous factors such as group size and environment [2]–[5]. However, although some recent studies have suggested that different construction can emerge even in a homogeneous environment and within a species[4], whether colonies display variations in the structure of architecture under identical exogenous conditions remains unclear. The Japanese subterranean termite Reticulitermes speratus is classified as a multiple-piece nester, wherein the nests of a single colony are interconnected by below-ground tunnels and above-ground shelter tubes [6]. Shelter tubes are made of wood, soil and termite excretions, which provide shelter and protection from predators and the external environment. In Reticulitermes termites, groups of individuals separated from an original colony perform collective activities, including shelter-tube construction. Therefore, we can examine each colony’s characteristics of building behavior by dividing a colony into multiple sub-groups of individuals. In this study, we show that termite colonies construct colony specific structures in shelter-tube construction by comparing morphological parameters of constructed structures among colonies in published work [7]. Eight colonies of R. speratus were collected during April–July 2012 from pine–oak forest at Uryuuyama (A and B), Takaragaike (C), Iwakura (D), and Yoshida (E–H), all northern suburbs of Kyoto, Japan. Four hundred were randomly chosen from each colony and placed on a block of mixed sawdust food (36 mm diameter × 15 mm height) at the center of a short plastic container (221 × 141 × 37 mm). The sawdust block was prepared from brown rotten pinewood and cellulose powder (Nacalai Tesque, Kyoto, Japan) mixed at a ratio of 5 to 1 by volume. We made five replicate groups for each of the eight colonies. To record shelter-tube construction, we took vertical photographs of each plastic container every 24 h with a digital camera (D300, Nikon, Tokyo, Japan). The termites constructed shelter tubes and/or covered the bottom of a container with mats consisting of particles of mixed sawdust and termite feces (Fig1 (a)). To examine the colony characteristics in the consequent structure of shelter tubes, we extracted the total length of shelter-tubes and the total area covered with shelter tubes and/or mats using planimetry in Adobe Photoshop software v.11.0.2 (Adobe Systems, Inc., San Jose, CA, USA). We tested colony differences in these parameters by using ANOVA with Tukey's contrasts. All analyses were performed using the “car” package implemented in R v. 3.1.2 (R Development Core Team, Vienna, Austria). Of the 40 groups (5 groups × 8 colonies), 37 groups formed shelter tubes and three did not, merely covering the container surface with mats of particles. Mat covering patterns were shown only in colony D and H. We found significant differences among the 8 original colonies of experimental groups in the total length of shelter tubes (ANOVA: F7,31 = 32.21, P < 0.001, Fig1 (b)) and constructed area (ANOVA: F7,31=20.88, P < 0.0001, Fig1 (b)). The termites showed distinct colony-specificity in their shelter-tube construction, that is, individual groups of workers derived from the same colony showed similar patterns of shelter-tube construction, whereas groups derived from different colonies showed distinctly different construction patterns. It has been shown that groups of social insects can construct various shaped structures even by using the same building rule, where environmental factor or group size modify the form of interactions by affecting individual responses or building dynamics [4], [8], [9]. In this study, however, colony variation in shelter-tube construction is created under the identical exogenous condition. Therefore, groups originated from different colonies is considered to show different form of interactions in collective building, which will be driven by the colony variation in responses of workers to identical conditions. Because the collective building of social insect colonies emerges from the actions of their individual workers, individual variation among workers will be important in determining the variation of behavior among colonies. There are some hypotheses of how individual level differences in the behavior of workers can lead to colony level variation of collective behavior, such as 1. Colonies differ in their average if worker character; 2. The distribution of worker characters varies among colonies; and 3. Both of them are different among colonies [10]. In this termite, individuals cannot distinguish kin from non-kin members within a nest; therefore, they perform normal altruistic behaviors when workers from different colonies are mixed and re-organize a colony [11]. Colony mixture experiments provide a powerful tool for linking colony level characteristics expressed through individual behaviors with colony-specific patterns that are created through self-organizing processes. In general, one primary goal in studies of self-organization is to understand the mechanisms that construct adaptive structures and organize vast numbers of individuals [1]. Therefore, only a few studies have focused on the colony variation of the structure that emerges from collective building. Here we showed that termite colonies displayed remarkable colony-specificity in construction. This variation is caused by non-exogenous factors, such as genetic differences, physiological differences or learning, which has been rarely considered in most of studies of self-organized pattern forming. By exploring these links, we will be able to understand how self-organization rule works in insect society.

ABSTRACT. The distributed formation control uses local relative measurements for the control of agents; in this regard, it is conjectured that the distributed formation control imitates the feature of collective animal motions. After briefly reviewing the collective behaviors of biological systems, distributed formation control laws are analyzed to reveal how they are related to the animal’s sensing mechanisms. Additionally, as a new viewpoint of this paper, the robustness issues of [13] will be studied through numerical simulations.

ABSTRACT. The coordinated activity of locust swarms (Figure 1) has always been a challenge to laymen and scientists alike. Despite considerable progress in understanding the mechanisms underlying the emergence and synchronization among moving crowds of animals and humans alike, several aspects of locust swarming are still far from fully resolved. These include biological issues related to the interactions among locusts, and between the individuals, the swarm and the environment, and to the functional and evolutionary advantages of locust swarming, but also theoretical questions, related to local and macroscopic dynamics, order and disorder in locust swarms, phase-transitions, and more. As part of our ongoing effort to decipher the rules governing locust coordinated behavior, we use video-recording and cutting-edge movement tracking methods to monitor the marching behavior of five-instar desert locust nymphs in the laboratory. A group of crowd-reared nymphs in our experimental system spontaneously generates robust and consistent coordinated marching (Figure 2). Using high temporal and spatial resolution analysis, we investigate the early stages and maintenance of this behavior, from the movement patterns of individuals and the effective social interaction between conspecifics, to global properties of the swarm. As we have recently reported in Ariel et al. [1], intermittent switching between standing and walking is a major feature of the behavior of individuals within the crowd. We have suggested that this behavior is triggered by tactile and/or visual stimuli and that the intermittent motion pattern constitutes a sequence of individual decisions (Figure 3) in which animals repeatedly reassess their situation and their environment, and decide whether or not to swarm. Accordingly, we introduced a new agent-based modeling approach in which pause-and-go motion is pivotal [1]. As also detailed by Ariel and Ayali [2], this interpretation explains several macroscopic observables of the entire swarm, for example, the measured correlation between the order parameter (which quantifies the level of synchronization) and the fraction of walking animals. Theoretically, it implies the existence of generic characteristics in the emergence of collective order in swarms. The suggested instrumental role of visual stimuli was further augmented by neurophysiological experiments showing that the Descending Contralateral Movement Detectors (DCMD, a pair of interneurons instrumental in directing visual sensory inputs to motor centres) are sensitive to stimuli related to approaching and receding walking insects [1]. Furthermore, the DCMD of crowd-reared, gregarious animals showed a more adapted response to the visual trigger observed (compared to solitary-reared animals). We are currently testing the role of the DCMD in directly inducing walking initiation in response to natural-like stimuli. We also further explore the effects of a changing environment on the behavior of the individual locusts and the swarm. In a recent related project we are looking at the marching locusts as inspiration for swarming robots. As a first step we are looking to incorporate an autonomous agent in a locust swarm to be followed by the creation of a locust-inspired swarm of robots.