Akihiro Fujihara

Disaster Evacuation Guidance Using Opportunistic Communication: The Potential for Opportunity-Based Service

In recent years, as a practical use of Delay Tolerant Network and Mobile Opportunistic Network, disaster evacuation guidance effective against situations of large-scale urban disasters have been studied. We have proposed a disaster evacuation guidance using opportunistic communication where evacuees gather location information of impassable and congested roads by disaster into their smartphones by themselves, and also share the information with each other by short-range wireless communication between nearby smartphones. Our guidance is designed not only to navigate evacuating crowds to refuges, but also to rapidly aggregate the disaster information. On the other hand, the Great East Japan Earthquake in 2011 taught us a lesson: the only helpful services in disaster situations are services that are daily used by everyone. Inversely services prepared only for disaster situations have never been used in a disaster situation because of lack of maintenance or unawareness by ignorance. To effectively utilise our evacuation guidance, therefore, some service using opportunistic communication should be firstly widespread across the world as an infrastructure and everyone naturally receives much benefit from the service in daily life. In this chapter we consider a possibility of the service, which we call Opportunity-based Service (OBS). We discuss some practical usages not only for disaster situations, but also for daily life, for example, an autonomous human navigation avoiding congestion by crowds. Through reviewing our past works, we try to foresee a possible next-generation information communication technology regarding Big Data, IoT, and pervasive computing on smart environments.

Effect of Traffic Volume in Real-Time Disaster Evacuation Guidance Using Opportunistic Communications

In case a large-scale disaster hits a big urban area, studies have been undertaken on effective disaster evacuation guidance using mobile phones of evacuees. In our previous work, we proposed a real-time disaster evacuation guidance using opportunistic communications as an application of Delay Tolerant Networking to disaster management. In the guidance, evacuees gather disaster information, such as impassable roads and dangerous areas, on their mobile phones by themselves and share it with each other using opportunistic communications between nearby phones on their evacuation routes to refuge areas. We numerically evaluated the performance of the guidance and showed that the guidance considerably reduces evacuation time. In this evaluation, however, we neglected the effect of traffic volume of evacuees, which might cause serious traffic congestion, inhibiting a smooth evacuation. In this paper, we consider the effect of traffic volume on evacuation time and numerically evaluate the performance again by introducing capacity-constrained nodes on road graphs. We show that the guidance also effectively reduces the evacuation time regardless of the traffic volume.

Efficiency Analysis on an Information Sharing Process with Randomly Moving Mobile Sensors

For the recent remarkable developments of wireless communication technology, information communication networks between mobile sensors become possible and have been studied recently. In the networks, however, the theoretical understanding of efficient content delivery schemes for collecting information is quite primitive at present. To promote the understanding, we investigate analytically and numerically an information sharing process with mobile sensors where the motion of sensor is a random walk and each sensor exchanges possessing information with the other sensors if they are at the same position on the d-dimensional unbounded square lattice. We find that probability distribution functions for information collection times obey power laws at the tails. We introduce three classes for the characterization of information collection performance and classify the system by recursiveness of random walk and the power-law exponents. Consequently, information collection is successful in one and two dimensions, while unsuccessful in larger dimensions. It also become efficient in one and two dimensions as increasing the number of sensors sufficiently. However, the system in two dimension needs much more sensors for the efficient collection compared to that in one dimension.

Homesick Lévy Walk: A Mobility Model Having Ichi-Go Ichi-e and Scale-Free Properties of Human Encounters

In recent years, mobility models have been reconsidered based on findings by analyzing some big datasets collected by GPS sensors, cell phone call records, and Geotagging. To understand the fundamental statistical properties of the frequency of serendipitous human encounters, we conducted experiments to collect long-term data on human contact using short-range wireless communication devices which many people frequently carry in daily life. By analyzing the data we showed that the majority of human encounters occur once-in-an-experimental-period: they are Ichi-go Ichi-e. We also found that the remaining more frequent encounters obey a power-law distribution: they are scale-free. To theoretically find the origin of these properties, we introduced as a minimal human mobility model, Homesick Lévy walk, where the walker stochastically selects moving long distances as well as Lévy walk or returning back home. Using numerical simulations and a simple mean-field theory, we offer a theoretical explanation for the properties to validate the mobility model. The proposed model is helpful for evaluating long-term performance of routing protocols in delay tolerant networks and mobile opportunistic networks better since some utility-based protocols select nodes with frequent encounters for message transfer.

Homesick Lévy Walk and Optimal Forwarding Criterion of Utility-Based Routing under Sequential Encounters

The Internet of Things (IoT) is going to develop integrated and organised networks of all things and beings in the world enabling autonomous computing and information communication for the creation of new values in the future. For such networks by IoT that accept a certain level of communication delay, but that must realise highly-reliable message forwarding, Delay Tolerant Network (DTN) gives a possible solution. Recently, DTN has attracted attention as a future network under challenged network environments where communication delay, disruption, and disconnect frequently occurs. In this chapter, we review some routing protocols for efficient message forwarding in DTN. We also review some mobility models often used for simulating motions of mobile nodes to evaluate the performance of DTN. In this review, we propose our mobility model called Homesick Levy Walk that mimics human mobility patterns of an universal scale-free property of the frequency of human contacts. After this, we also propose our utility-based routing protocol which maximises the expected number of selected relay nodes being likely to encounter a destination node under sequential encounters with nodes. We evaluate the performance of our routing protocol by comparing with some performance measures of some existing routing protocols under the condition that the Homesick Levy Walk is adopted as mobility model. We show that our protocol is comparable to others in arrival rate of messages under a smaller number of message forwarding.We also find that the performance of our protocol is stable up to a few hundred mobile nodes and tends to be scalable with the number of nodes.

On the Use of Congestion Information for Rerouting in the Disaster Evacuation Guidance Using Opportunistic Communication

In recent years studies on disaster evacuation guidance effective against situations of large-scale urban disasters have been undertaken. Recently we have proposed the disaster evacuation guidance using opportunistic communication where evacuees collect disaster information of the locations of impassable and congested roads in their smartphones by themselves and share it with each other opportunistically by short-range wireless communication between nearby smartphones in order to not only navigate crowds of evacuees to refuges, but also rapidly aggregate the disaster information in real time. The proposed guidance was valid in evacuation scenarios under the condition that the mobile communication is disabled by disaster. We numerically showed by simulating a simple mathematical model that the guidance effectively shortened evacuation time of the evacuees. Since the crowds of evacuees simultaneously rush to refuge areas in the evacuation, on the other hand, the emergence of congestion by evacuees is inevitable and the evacuation time could be seriously delayed by the congestion. In this paper we investigate effects of the use of congestion information on the evacuation time by switching a detour evacuation route around congested roads. We introduce three simple algorithms to divert the route and numerically compare their performances using a new stochastic congestion model. In consequence we find that the use of congestion information has less impact on the evacuation time than that of the information of impassable roads, and also that slower route switching decision coming around congested roads is worse in the performance because evacuees tend to be sandwiched between some congested areas and are forced to go back and forth in vain by repeated evacuation route changes.

Universal power laws in the threshold network model: A theoretical analysis based on extreme value theory

We theoretically and numerically investigated the threshold network model with a generic weight function where there were a large number of nodes and a high threshold. Our analysis was based on extreme value theory, which gave us a theoretical understanding of the distribution of independent and identically distributed random variables within a sufficiently high range. Specifically, the distribution could be generally expressed by a generalized Pareto distribution, which enabled us to formulate the generic weight distribution function. By using the theorem, we obtained the exact expressions of degree distribution and clustering coefficient which behaved as universal power laws within certain ranges of degrees. We also compared the theoretical predictions with numerical results and found that they were extremely consistent.

Necessary Condition for Self-Organized Follow-Me Evacuation Guidance Using Opportunistic Networking

There is a serious problem in disaster evacuation guidance in urban areas: When a large-scale disaster strikes in the areas, how can many crowds of evacuees are smoothly evacuated to refuges without any major confusion by lack of disaster information? In recent years, we have proposed a method for disaster evacuation guidance using opportunistic networking where the evacuees gather and share disaster information using wireless communication by smartphones on the way in their evacuation, and we have shown numerically that our method drastically reduces evacuation time and efficient disaster information gathering. In our numerical simulations, however, there was an assumption that all evacuees totally follow our evacuation guidance. In this paper, we loosen the assumption for investigating the scenario that only a certain percentage of evacuees can follow our guidance, but the rest of unguided evacuees tends to follow movement of the majority of nearby evacuees. By this synchronous movement, a small group of evacuees including a core leader who instantly leads the other nearby evacuees by our guidance at the initial stage of evacuation will gradually aggregate nearby evacuees to form a self-organizing large group of evacuees, which creates a evacuation flow of evacuees along evacuation routes. We call these processes self-organized follow-me evacuation guidance. To evaluate the performance of the newly proposed guidance, we propose a model of Majority Voting Mobility (MVM) for simulating the formation of the self-organizing groups of evacuees. We use evacuation time and the cumulative number of evacuees arrived at a refuge for evaluation measures. Furthermore, we also consider a simple mean-field theory of MVM to discuss the necessary condition for the self-organized follow-me evacuation guidance. Consequently, we can numerically obtain reasonable numerical results which are consistent with an earlier study by Sugiman and Misumi, and we also show a threshold of the ratio of the core leaders to total evacuees that they can successfully form the self-organizing group.

Analyzing scale-free property on human serendipitous encounters using mobile phone data

The recent expanded use of mobile phone dataset for research purpose enables studies not only on human mobility patterns, but also on human contact patterns including proximal contact, face-to-face meeting, and serendipitous encounter. In our previous work, we analyze our dataset of long-term experiment with periodic scanning of proximal devices emitting close-range radio waves for wireless communication to investigate the frequency of human serendipitous encounters. As a result, we have found that the frequency of human serendipitous encounter has a scale-free property where the number of encounters is highly biased and its average means nothing. Our dataset is not Big data (the number of participants in our experiment is a few dozen), but it is Long data, but the property is universally observed. It is important to check from a multilateral perspective whether mobile phone data, which is one of well-used Big datasets, also support this fundamental property of human serendipitous encounter. In this paper, we investigate the scale-free property of human contact frequency using mobile phone data of D4D Challenge Senegal which contains hundreds thousands of human mobility traces. Because the data is not provided with high resolution enough to detect exact human encounter at proximity, we roughly assumed that humans who are considered to be placed around the same cell tower encounter with each other. Despite this rough assumption, we successfully reproduce the scale-free property.

ZebraNet and its theoretical analysis on distribution functions of data gathering times

We theoretically investigated a general property of data gathering times in a wireless communication system with randomly moving sensors which share data only with nearby ones. We proposed a stochastic model of the system to analyse distribution functions of data gathering times. We found that the time distribution asymptotically obeys a power-law decay in infinite space, while it becomes exponential in finite space. Mean and variance of the time distributions become finite as the number of sensors is sufficiently large, meaning efficient data gathering can be accomplished by deploying a large number of sensors when sensors are spreading data epidemically. In the finite space, moreover, both power-law and exponential distributions coexist in general. We proposed a truncated power-law distribution for a least-square fitting of the time distribution on the whole range to estimate their accurate mean and variance.