ABSTRACT. Consider the proverb “It would not do for us all to be the same”. This proverb is a declaration that human society only works because of differences among its　members. It implies that the uniqueness of individuals is vital to the common　good. In this presentation, I will consider a case in which differences among ants in a single task, new nest site evaluation, may benefit their colony hugely. Indeed, the uniqueness of individuals may be the basis of the　imagination of the system. The principle that “It would not do for us all to be the same” is also manifest in science in　general and interdisciplinary science especially. Moreover, if a novel bridge is to be built between biology and engineering &emdash; the effectiveness of that new enterprise will depend on the bridge sustaining two-way traffic. Perhaps controversially, I will argue that we will need something far more useful than mutual “inspiration”: we will need the different approaches of biologists and engineers to develop deeper understanding rather than superficial inspiration. Virtuous cycles of biological and engineering experimentation, where each informs the next revolution, should favour new and very exciting technology and science.

ABSTRACT. The implementation of coordination and control algorithms for multi-robot systems often depends explicitly on the sensing modalities available to the robots. Common coordination goals, such as formation control, can have vastly different control algorithms depending on what information is available to each robot. In this talk, we explore how various sensing modalities, such as distance measurements or bearing measurements, are used to achieve coordination tasks like formation control and localization. Despite the differences of the sensing modalities, we show that they share a common underlying conceptual and theoretical framework based on rigidity theory. The talk will focus on recent results in bearing-only formation control and highlight open challenges in this arena.

ABSTRACT. There are many types of foot sensors (e.g., force and pressure sensors) available for legged robots. Most of them are expensive or have hysteresis effects. In this paper we demonstrate the design of a novel but simple and cheap foot sensor. The sensor uses a photo resistor and a light emitting diode enclosed in a 3D printed cylindrical assembly separated by a spring. The sensor’s response can be adjusted by changing the spring stiffness. The sensor is based on a simple mechanism and uses a basic voltage divider circuit to obtain the output signal. Though the design of the sensor can be transfered to other systems, our initial design is targeted to the robotic toolkit, LocoKit, for monitoring walking behavior and detecting different types of surfaces.

ABSTRACT. This paper investigates the constrained formation tracking of multiple Lagrangian systems with input dead zones and external disturbances. A family of artificial potential functions are derived for Lagrangian systems. Besides, a smooth inverse of input dead zone is employed, and the uncertain parameters of dead zone nonlinearity can be estimated by adaptive laws. The unknown external disturbances are suppressed by sliding mode control together with dynamic gaining factors. The proposed coordinated control framework is applied into the constrained attitude formation control of multiple rigid bodies, and the according simulation results are provided to show the effectiveness of the proposed scheme.

ABSTRACT. In the last years, various hardware-based Spiking Neural Networks (SNN) systems were developed and the description of those pioneer platforms have gained remarkable attention. On account of their parallel and distributed structures, spiking neuronal networks can simulate neuronal activities, potentially realizing an extremely large-scale network comparable to that of the human brain in future. Neuromorphic engineering aims to design SNNs which will be used in brain-like computers. A second approach in the neuromorphic community concerns biomimetic systems, which mimic the activity of biological cells and could replace the living part. A neuromorphic system facilitates the building of a hybrid network incorporating both silicon and biological neurons. In recent times, the term neuromorphic has been used to describe analog, digital, mixed-mode analog/digital VLSIs, and software systems that implement biologically realistic neural network models, from the electrophysiology of a single neuron to the network plasticity rules. In this work, we describe a biomimetic SNN for pattern recognition used for biological data processing. Comparisons with different neuron model are done. The success rate for the pattern recognition algorithm is better using DSSN neuron model. As this model can realize Class I and Class II properties, the interesting way to explore is to use SNN with complex neuron model for increasing the success rate of usual bioinspired techniques. Next step is to implement this architecture on FPGA for real-time applications. This work is expected to be used for pattern recognition directly from the biological neuron recording from MEA (MicroElectrode Array) for biomedical applications

ABSTRACT. Modern deep neural networks usually consist of many layers with unidirectional feed-forward connectivity. By contrast, the architecture of the cortex of vertebrates contains fewer hierarchical levels but many recurrent and feedback connections. Here we show that a small, few-layer artificial neural network that employs feedback can be used for image recognition when controlling a grasping operation with a robot. This network reaches top level performance on a standard benchmark recognition task, otherwise only obtained by large feed-forward structures.

ABSTRACT. Group-living organisms must make decisions to optimise life-history trade-offs in the same way as individuals. However, and inherent advantage of groups is the capacity to pool information among individuals. This can allow groups to make better decisions, an idea expressed colloquially as the ‘wisdom of the crowds’. However, crowds are equally famous for making poor decisions, when initial errors are magnified by the same positive feedback mechanisms that underlie the group advantage. The effectiveness of decision making in groups therefore relies on maximising the advantages and minimise the costs of the collective. Myrmecina nipponica forms small colonies of 10-70 ants and is an ideal model for exploring the mechanisms of decision making in natural societies. Colonies of this species uses a combination of pheromone trails and a quorum-based decision-making process to select among candidate nest sites when deprived of their original home. Ants ‘vote’ for a preferred site by laying a chemical trails between it and the original colony location. These trails recruit additional scouts to the site, and those which deem the site acceptable reinforce the trail with pheromones of their own. No final decision is reached however, until the number of scouts at one site exceeds a critical number. Once this quorum-threshold is achieved, ants begin transporting brood to the new site, thus concluding the emigration. Many ants use pheromone trails as a means of information sharing, and this can be a very effective way of recruiting a large number of nest-mates to sites of interest. Pheromone trails can also suffer from ‘lock-in’ problems, however, because the cycle of positive feedback is difficult to break once one option becomes dominant. Combining a quorum threshold with chemical recruitment helps insure against runaway positive feedback. This group-level mechanism is supported by individual-level vetting of social information (pheromone trails) with private information (direct assessment), providing two mechanisms via which negative ‘information cascades’ are avoided. These mechanisms allow colonies to reap the benefits of sharing information while being insured against the costs. Indeed, there is evidence that colonies do benefit from a ‘wisdom of the crowds’ effect from collating information among individuals, as larger colonies of M. nipponica can make better decisions. These decisions are achieved in the same amount of time as in small colonies but involve more information collectors, because quorum thresholds scale with colony size. Not to be outdone, individuals in small colonies increase per-capita information by working harder during decision making, and this may help ameliorate the shortage of information collectors.

ABSTRACT. Complex systems are characterized by behaviour at the system level that is difficult to explain by examination of interactions between system components. Interactions between components have a synergistic effect, such that group level products exceed the sum of the parts, allowing complex collective tasks or collective behaviour to be completed by teams of the simplest creatures. Understanding complex systems is a topic of current interest in a range of fields, from improving our understanding of biological systems to the application of these concepts to computer science, the developed swarm robots, and efficient transport systems. The very complexity of complex systems makes them difficult to study, and research has thus focused either at the level of the group (system), or individual components (individuals). While valuable, these approaches miss a critical component in the interaction between these levels. Complex systems resemble a ‘black box’ as while the group state may be clear, it is not obvious how this is reached through the interactions between system components. Network analysis explores relationships between nodes (individuals) connected by links (interactions), allowing us to integrate analysis of both group level and individual level components. This approach can elucidate how the complex group level behaviour is derived from interactions between, and actions of, individual components (nodes and links). In this study, we characterize network structure of individual interaction networks in a small-colony ant, Myrmecina nipponica using an AR tracking system. In this study, we present preliminary findings from the application of a network analysis approach to shed light on collective behaviour in an ant model system. We used the ant Myrmecina nipponica, which is an established model of collective behaviour in ants. Individual ants were tagged with 0.8 mm AR barcodes printed on high density paper using offset printing and treated with polyacrylate resin. Tags were fixed to each ant’s thorax using chemical reaction type adhesive and remained robust for the duration of experiments. BEEtag software was used to extract x and y coordinates of tags from digital photographs taken at 1 second intervals with a high-resolution mirrorless camera (Panasonic, DMC-GH4H). Experimental chambers were illuminated using LED light sources (Style+, PAR38-12*3W) covered with red filters to limit light exposure of the ants. Each colony was photographed for 4 hours. An example video is available at “https://youtu.be/9fD_WR0A2k0”. Interactions between individuals were calculated based on a individual heading and distance. Edge lists (interactions) and an adjacency matrix data were generated from these interaction data for network analysis. Network structural indicators (e.g., Small world-ness, average path length) were calculated using each colony’s edge lists and adjacency matrix in the R packages (igraph, sna, SocialNetworks, qgraph). These networks were compared to 1000 randomly generated networks which had the same number of nodes and links (generated using the Erdös-Rény method). In this presentation we discuss i) how network structure varied in colonies of different size and ii) how real networks compared to random networks.

ABSTRACT. Many of the collective activities performed by group living animals result in the formation of complex spatio-temporal patterns. In social insects, collective building often produce complex structures involving self-organization, where colony-level patterns emerge from local interactions among members. Although some studies have revealed the mechanism of building sophisticated or complex structures in several species, little is understood about its evolutionary perspectives. In this study, I purposely focus on the building behavior of primitive termite species (Zootermopsis nevadensis) which is known not to build outstanding structures but repair their nests by plugging openings of their nests with barricades. I found that even this species have building algorithm similar to more derivative species that build largest mounds. Moreover this species show highly skewed task allocation in their building behavior. By comparing the collective building of termite species, I will discuss how the self-organization works in the insect society and the evolutionary perspectives of collective behavior of group living animals.

ABSTRACT. Bacterial swarming is a collective mode of cell motion in which rod-shaped, self-propelled bacteria rapidly migrate over surfaces. Swarming is associated with several biological manifestations such as cell elongation, secretion of wetting agents and increased antibiotic resistance. During swarming, densely-packed groups of bacteria move in coherent swirling patterns of whirls and jets that can persist for several seconds.

Experiments analyzing the dynamical swirling patterns of this group-phenomenon suggest a range of physical mechanisms such as steric (excluded volume) and hydrodynamic interactions to explain this dynamics. For example, it was shown that dense suspensions of self-propelled rod-shaped particles are subject to orientational order instabilities which may be driving the vortex-like and irregular dynamic patterns. In other words, the collective swirling dynamics is a physical consequence of the mechanical characteristics bacteria exhibit during swarming.

Recent experiments (Ariel et. al., Nature Communications 2015) analyzing the behavior of individuals within such dense swarms have revealed a significantly more complex dynamics – that trajectories of swarming cells are super-diffusive. Moreover, they are consistent with a particular stochastic process known as a Lévy walk. Such movement patterns, often attributed to the execution of an advantageous searching strategy, have been found in a wide variety of organisms, from cells to humans. Although the mechanisms underlying Lévy walks vary, in the past, they have been mostly related to sparsely distributed independent individuals. In contrast, Lévy walks in swarming bacteria are attributed to the collective rather than individual movements, namely that the group is a keystone of this phenomenon.

The talk will describe and explain these observations. Furthermore, we present a simplified model (Physical Review Letters, accepted), explaining the observed Lévy walks in swarming bacteria. We show that due to the interaction with the collective swarm, the dynamics of individual cells becomes chaotic, which in turn results in super-diffusion and Lévy walking. Our findings suggest that bacteria can manipulate their large numbers in order to change the geometrical properties of their trajectories.

Our results suggest new possible advantages of swarming and collective motion in nature, in particular under adverse conditions, by demonstrating that the dynamics can be acted upon by selection pressures for advantageous searching.

REFERENCES [1] G. Ariel, A. Rabani, S. Benisty, J.D. Partridge, R.M. Harshey and A. Be'er, Swarming Bacteria Migrate by Levy Walk, Nature Commun. 6, 8396 (2015). [2] G. Ariel, A. Be'er and A. Reynolds, A chaotic model for Levy walks in swarming bacteria, Phys. Rev. Lett., accepted.

ABSTRACT. The authors proposed a novel underactuated locomotion robot that generates a crawling-like gait on a downhill by utilizing the effects of sliding and wobbling, and numerically clarified that the central angle of the sinusoidal vibration of the wobbling mass is the most important parameter for control of the moving speed. In this paper, we describe the development of an experimental machine and report the experimental results. First, we outline our previous mathematical investigations and the prototype experimental machine developed based on the simulation model. Second, we report the experimental results obtained using the improved experimental machine, and discuss the gaps between those and simulation results.

ABSTRACT. The authors have investigated the method of collisionless walking for point-footed walkers. The generated motion consists of the single-limb support and double-limb support phases, and the gait inefficiency is mainly caused by the period of the latter phase. In this paper, we propose a method for achieving an efficient collisionless walking without including the period of double-limb support. First, we consider a model of an underactuated rimless wheel with an upper body, and develop the equation of motion. Second, we design an output-following controller for the linearized equation of motion, and mathematically derive the condition for achieving collisionless walking without including the period of double-limb support motion

ABSTRACT. The walking motion of combined rimless wheel with wobbling mass demonstrated a dependence of its entrainment property on the frequency of the wobbling mass. However, complex nonlinear phenomena have not yet been thoroughly investigated. To resolve the remaining issues, this study presents nonlinear analysis of the walking dynamics. Arnold tongues are computed for the combined rimless wheel entrained by different waveforms. Furthermore, chaotic gaits are investigated under different conditions. Sine wave input shows a better synchronizability than impulse wave input. Chaotic gaits are more often induced by light wobbling mass with weak wobbling amplitude.

ABSTRACT. The proposed method can generate an optimal feedforward control input and the corresponding optimal walking trajectory minimizing the L2 norm of the control input by iteration of laboratory experiments. Since a general walking motion involves discontinuous velocity transitions caused by the collision with the ground, the proposed method consists of the combination of a trajectory learning part and an estimation part of the discontinuous state transition mapping by using the stored experimental data. We apply the proposed method to a kneed biped robot with a torso, where we also provide a technique to generate an optimal gait not only energy-efficient but also avoiding the foot-scuffing problem.

ABSTRACT. This paper investigates the stability of underactuated bipedal walking incorporating telescopic-leg actuation. In human walking, knee joints of swing and support legs are bent and stretched. The telescopic legs mimic the motion of the center of mass of human legs via their telescopic motion during the stance phase. First, underactuated telescopic-legged biped robot models are introduced. Second, an output-following control law is applied to the linearized equation of motion of the robot, and the controlled robot's equation is then specified as a linear time-varying system. The error transition equation is developed to evaluate the stability during the stance phase. The numerical calculations are performed to show the influences of leg telescopic motion on the stability.

ABSTRACT. This paper discusses a method for achieving high-speed collisionless walking of a point-footed walker. We introduce an underactuated rimless wheel model for analysis, and develop two controllers for the single-limb support and double-limb support phases so that they minimize the corresponding control period. The former is determined under the unilateral constraint condition, whereas the latter is determined under the ZMP constraint condition, respectively. The validity of the proposed method is investigated through numerical simulations.

ABSTRACT. Our research group has developed robot for a rice field ("Aigamo robot"). The problem of Aigamo robot are motion is uncertainty by disturbance and robot damages to rice plants. Sweep path for Aigamo robot generates by separating sweeping field into square cells. The sweep path planning at traditional type that sweep all cells in sequence is difficult and taking too much time for Aigamo robot because robot motion is uncertainty. Aigamo robot cannot enter into the target cell, and robot sweep many times at same point when robot introduces it. Aigamo robot actually sweep every day. Aigamo robot can sweep upward another day when robot cannot sweep completely. Therefore, it is assumed that sweep rate of 80[%] is enough. We propse to reducing the number of visiting cells. The proposed method is realized that sweep rate is 80[%] or more and the damage to rice plants is smaller.

ABSTRACT. This paper presents a cooperation of swarm robots based on casualty. The robots are initially placed in a field with a ditch which single robot cannot get across. All robots aim for a common destination. An algorithm for each robot is so simple that some of them are pushed out towards the ditch. The sacrificed robots are left as a part of a bridge. Completed bridge enables the rest to go across the ditch. This phenomenon is considered as an unintentional cooperation. We show fundamental characteristics of this system investigated by computer simulation and small scale experiments by real robot system.

ABSTRACT. This paper proposes to use evolutionary computations to design probabilistic finite state machine for the controller on an aggregation problem of swarm robotics. This problem, formulated as an optimization problem was solved by the PSOs. Several computer simulations were conducted to investigate the validity of the proposed method. The results obtained in this paper show to us that the proposed method is useful for the aggregation problem and the best optimized controllers are interpretable. This would be transferable to real swarm robots problems.

ABSTRACT. Optimization problems have some characteristics: Dependency among decision variables such as separable or non-separable problems, and landscape modality such as unimodal or multimodal problems. Particle swarm optimization (PSO) has been shown powerful search performance especially in separable and unimodal problems. However, the performance is deteriorated in non-separable problems such as rotated problems. Although velocity updating rules using random rotation matrices have been proposed to solve non-separable problems, the computational cost of generating the random rotation matrices is very high. In this study, a new velocity updating rule according to an oblique coordinate system, instead of an orthogonal coordinate system, is proposed to solve non-separable problems. Also, two mutation operations for the worst particle and the best particle are proposed to improve the diversity and the convergence of particles, respectively. The advantage of the proposed method is shown by solving various problems including unimodal problems, multimodal problems, and rotated problems and by comparing the results of the proposed method with those of standard PSO.

ABSTRACT. This paper proposes a distributed model predictive control method for a consensus of robotic swarm systems. The proposed method is free from any precomputed trajectories, contrast to previous studies. This is huge advantages in the cases systems do not work as previously planed, communication topologies change frequently, and the communication is unreliable. We introduce a concept of local subsystem regulator first, and then the distributed control method and its design concept are described. We also analyze stability of the proposed method and collision avoidance constraints are integrated into our method. Computer simulations are demonstrated to show effectivity of our method.

ABSTRACT. In this paper, we investigate realization of trial and error behavior at the swarm level in order to obtain new emergence. We attempt the realization by the change of agreement of the members from hour to hour. In this case, it is required that the agreement can be achieved promptly even if the number of individuals of swarm becomes large and various agreements can be utilized. In this paper, computer simulations make it clear that trial and error behavior can be realized by designing individuality well.

ABSTRACT. Snake robot locomotion in cluttered environments where the snake robot utilises a sensory-perceptual system to perceive the surrounding operational environment for means of propulsion can be defined as perception-driven obstacle-aided locomotion (POAL). The development of POAL is challenging. Moreover, testing new control methods for POAL in a real setup environment is very difficult because potential collisions may damage both the robot and the surrounding environment. In this perspective, a realistic simulator framework may enable researchers to develop control algorithms for POAL more safely, rapidly and efficiently. This paper introduces SnakeSIM, a virtual rapid-prototyping framework that allows researchers for the design and simulation of control algorithms for POAL. To demonstrate the potential of SnakeSIM, a possible control approach for POAL is also considered as a case study.

ABSTRACT. This paper studies a method to climb ladder with a snake robot. We propose a method of ladder climbing for snake robot whose surface shape is smooth by using the gait design method which configurate target form of snake robot by connecting simple shapes. Climbing motion is executed by shift control and the motion to catch a next step of a ladder. We demonstrate the effectiveness of proposed motion using a physical simulation.

ABSTRACT. The Underwater Swimming Manipulator (USM) is a snake-like, multi-articulated, underwater robot equipped with thrusters. One of the main purposes of the USM is to act like an underwater floating base manipulator. As such, it is essential to achieve good station-keeping and trajectory tracking performance for the USM using the thrusters, while using the joints to attain a desired position and orientation of the head and tail of the USM. In this paper, we propose a sliding mode control (SMC) law, in particular the super-twisting algorithm with adaptive gains, for trajectory tracking of the USMs center of mass. A higher-order sliding mode observer is proposed for state estimation. Furthermore, we perform a simulation study to verify the applicability of the proposed control law and show that it has better tracking performance than a linear PD-controller.

ABSTRACT. A snake, which has a simple shape without limbs, can move on a rough terrain, climb a wall and a tree, swim, and glide in the air. It is difficult to understand the gliding flight of a snake in aerodynamics assuming a static model, because a snake undulates the body trunk and turns in the air. This paper presents an experimental system to investigate relationship between the shape of gliding snake model and glide characteristics. As a result of the gliding experiment, it was confirmed that the reproducibility of the experimental result is sufficiently high.

ABSTRACT. Snake robots have a possibility of utilizing in many fields using its property of hyper redundant, although there are control complexity problems when it is applied to complicated environment. For example, helical rolling motion have been used to climb a pipe. By using the helical rolling motion, a snake robot can move along the inside or the outside of a pipe. However, this motion has limitation when it moves in a pipe with a high gap or a branch point. In this study, we propose a motion of a snake robot, which wraps around the outside of a pipe, to overcome a branch point on a pipe. The new motion uses a hyperbolic function to make a helical wave curve. The helical wave curve is propagated by shifting the shape of hyperbolic function along snake robot's body. The joint angles of a snake robot is derived by calculating curvature and torsion of the curve based on the formula of continuous curve model. Finally, the result of simulational experiment using ROS and Gazebo are shown to validate the effectiveness of the new motion.

ABSTRACT. This paper studies the interaction between a snake robot and an object to be pushed. The influence of the snake robot's configuration and the friction with the ground is modeled and studied. It is theorized that the snake robot's configuration must have a significant influence on the task. The model presented here can be applied to any (planar) snake robot and is presented under the framework of differential geometry, making it possible to be extended in the future to a full spatial snake robot. The model is verified with simulations on a three-jointed snake-robot, while varying the friction it has with the ground (from no friction to ideal non-holonomic constraints). The model and results can be used as a basis for future tasks combining locomotion and dexterous manipulation. It is shown that the configuration of the snake robot has a big impact on the resulting acceleration of the object. However, in addition to the object's motion, it is also shown the effect on the slippage of the robot itself.

ABSTRACT. This paper proposes a new self-localization method based on Extended Kalman Consensus Filter (EKCF) using only angular informations observed by many standing robots each other that are not controlled and that are dispersed in an environment as a localization method using only local information around robots. In a localization method by Extended Kalman Filter (EKF), to localize more dozen robots is difficult because matrix calculation are difficult for a regular computer. Therefore, to solve this problem a consensus is introduced to the EKF. Each EKF localize some robots in each EKF group and take consensus with each other to localize as a whole. To verify this method the numerical experiments were carried out and the effectiveness is confirmed.

ABSTRACT. Animals like as the birds and fish in the natural world form swarm in various forms. However, it is not clear how individuals affect each other and form and keep a swarm. In this paper, we show how the moving swarm keep the formation so as not to get separated as a group in the computer simulation. In the simulated environment, 10 agents equipped two wheels and eight IR sensors are controlled by neural network and evolved by genetic algorithm. Our result shows that the individuals keep the swarm behaviors by moving fluidly and changing their positions.

ABSTRACT. This paper proposes a novel specialization method by response threshold model using local interactions for swarm robotics. The response threshold describes a sensitivity for pheromones of ants. Actually, there are high and low pheromone sensitivity ants. And it is known that the pheromone sensitivity of ants is related with specialization. In conventional response threshold model, it is assumed that individual ant can gain the number of all workers in an ant group. However, it is difficult for an ant to gain and deal with the number of all workers because the ant functions are very limited. Thus, the proposed response threshold model uses the number of encountered foraging ants instead of the number of all workers. This paper investigates the robustness of the proposed method in the environment where the number of ants in a group fluctuates through an ant foraging simulation.

ABSTRACT. Redundancy is a fundamental feature of robotic swarms to confer robustness, flexibility, and scalability. However, robots tend to interfere with each other in a case where multiple robots gather in a limited environment. The aim of this paper is to understand how a robotic swarm develops an efficient strategy to manage the congestion. The behavior of the robots is obtained by an evolutionary robotics approach. The strategy of utilizing congestion is observed in the process of generating a path of robots visiting two landmarks alternatively. We will demonstrate that the robotic swarm emerges an adaptive division of labor to utilize congestion.

ABSTRACT. To apply Artificial Bee Colony (ABC) algorithm to the multimodal problems which locations of the Optimal solution change dynamically. This paper improves ABC algorithm based on Local Information Sharing (ABC-lis) for various landscape problem. Concretely, the improved ABC-lis called ABC-lis AD searches the optimal solution and several local solutions by changing sharing range adaptively in various landscape problem. To investigate the effectiveness of the global search ability of ABC-lis AD, we compare ABC-lis AD with ABC-lis in two types of landscape problems. The experimental result revealed the following implications: (1) ABC-lis cannot sometimes maintains the searching capability, while ABC-lis AD can exert a high performance in environment where the intervals of local minimum solutions are not uniform with each other; and (2) ABC-lis AD does not need to change parameters according to the landscape in question, and the behavior of ABC-lis AD does not change depending on the problem.

ABSTRACT. Massive earthquakes has been striking Japan in the recent years and posing large scale disasters such as the great Hanshin-Awaji earthquake in 1997, the Niigata Prefecture Chuetsu earthquake in 2004, the great east Japan earthquake in 2011, and the Kumamoto earthquake in 2016. In the all disasters above, logistics system for relief supply collapsed and this situation emerged as social issue in Japan. The reason repeating the situation that the supply goods does not delivered to the evacuation refugees properly stems from the lack of knowledge to estimate the performance of the logistics system in the disaster prevention plan. Recently many publications relevant to the logistics under disaster situation were presented. However the heart of the studies is optimization techniques for logistics under disaster circumstances. Before optimization, it is effective to investigate the transportation capability of the logistics system in the disaster prevention plan already exist. In this paper, using transportation simulator for disaster situation based on the Multi-Agent System, analytical solutions to grasp roughly the performance of logistics system in disaster prevention plan are developed. The analytical solutions give us the information on transportation throughput of the logistics system and also find out the bottlenecks in the system. Applying the analytical solutions to a city in Tokyo metropolitan, the solution quality were verified by comparing with that of the simulation analysis.

ABSTRACT. Cockroach antennae are multimodal sensory appendages engaging in active olfactory and tactile sensing. They are involved in unisensory behaviours such as chemotaxis, thigmotaxis, obstacle negotiation and tactile orientation. Here we report on an integrated experimental and computational approach to investigate how sensory information affects antennal movements. We present a modelling approach to characterise the relationship between antennal searching movement of the American cockroach \emph{Periplaneta americana} and the spatial properties of encountered odourant concentrations. Video recordings reveal that the antennae exhibit systematic movements in the presence of behaviourally relevant odourants. We hypothesise a dynamic coupling between the left and right antenna modulated by odour stimulation. To test this we designed experiments in which cockroaches were exposed to odourants arriving with different temporal patterns. We measured the spatial structure of the odourant plume encountered by the antennae at any given time and generated a spatial odourant concentration map. We then employed an adaptive frequency, coupled Hopf oscillator to model antennal movements in response to odourant concentration, using this map for sensory drive signals for the model. We present simulation results of antennal movements in response to odourant concentrations.