key: cord-211735-qqm4fbor
authors: Gulec, Fatih; Atakan, Baris
title: Mobile Human Ad Hoc Networks: A Communication Engineering Viewpoint on Interhuman Airborne Pathogen Transmission
date: 2020-11-02
journal: nan
DOI: nan
sha: 
doc_id: 211735
cord_uid: qqm4fbor

Pathogens such as viruses and bacteria play a vital role in human life, since they cause infectious diseases which can lead to epidemics. Recent coronavirus disease 2019 epidemic has shown that taking effective prevention measures such as wearing masks are important to reduce the human deaths and side effects of the epidemic. It is therefore requisite to accurately model the spread of infectious diseases whose one of the most crucial routes of transmission is airborne transmission. The transmission models in the literature are proposed independently from each other, at different scales and by the researchers from various disciplines. Thus, there is a need to merge all these research attempts. To this end, we propose a communication engineering approach that melts different disciplines such as epidemiology, biology, medicine, and fluid dynamics in the same pot to model airborne pathogen transmission among humans. In this approach, we introduce the concept of mobile human ad hoc networks (MoHANETs). This concept exploits the similarity of airborne transmission-driven human groups with mobile ad hoc networks and uses molecular communication as the enabling paradigm. The aim of this article is to present a unified framework using communication engineering, and to highlight future research directions for modeling the spread of infectious diseases among humans through airborne pathogen transmission. In this article, we first review the airborne pathogen transmission mechanisms. Then, the MoHANET is given with a layered structure. In these layers, the infectious human emitting pathogen-laden droplets through air and the exposed human to these droplets are considered as the transmitter and receiver, respectively. Moreover, the experimental methods for the proposed approach are reviewed and discussed.

Throughout the history, epidemics caused by infectious diseases have been a major threat to human life. Epidemic diseases such as black plague, smallpox, Spanish flu and recent coronavirus disease 2019 (COVID-19) gave rise to millions of human deaths. In addition, epidemics can induce mental disorders in humans and recessions in the world economy due to prevention and control measures such as lockdown. Owing to these facts, it is essential to understand and accurately model the spread of infectious diseases among humans.

The interhuman spread of infectious diseases occur via direct contact and airborne transmission 1 where pathogens are transferred from an infectious human to a susceptible one. The authors are with the Department of Electrical and Electronics Engineering, Izmir Institute of Technology, 35430, Urla, Izmir, Turkey. 1 Here, transmission is employed synonymously with contagion rather than its usage in communication engineering.

In airborne transmission, these pathogens (viruses, bacteria, fungi, and so on) are carried by large droplets and aerosols (droplet nuclei) which are emitted via breathing, speaking, coughing and sneezing [1] . Throughout this article, we use the term droplet to refer to both large droplets and aerosols together.

As for the airborne pathogen transmission, it is not fully unraveled how its mechanisms operate between two humans, for example, it is still a matter of debate whether large droplets or aerosols are more infectious. In addition, the mobility and interplay of people during their daily life makes the problem of modeling infectious disease spread in an epidemic more chaotic. As people displace, there exist dynamic human groups exchanging pathogens among each other. Due to their mobility, humans form different groups in an ad hoc fashion as their smart phones do in a mobile telecommunication network. Actually, a human emitting expiratory droplets is an information source [2] . When these emitted information carrying droplets are received by another human through sensory organs, we can consider there exists a communication path between them. Hence, an analogy between human groups and mobile telecommunication networks can be established, since they both possess an intermittent connectivity which is detailed later. By utilizing this analogy, we propose an approach to modeling interhuman airborne pathogen transmission with communication engineering perspective where mobile humans forming a group are considered as a mobile human ad hoc network (MoHANET). In a MoHANET, the infectious human is the transmitter (TX), the susceptible human is the receiver (RX) and pathogen-laden droplets are information carriers propagating in the communication channel, that is, air. Here, molecular communication (MC) employing chemical signals instead of electrical signals emerges as an enabler paradigm for the communication among humans due to its biocompatibility with the human body and multiscale applicability.

On the other hand, researchers from many disciplines work separately in different scales to reveal the mechanisms of airborne pathogen transmission and model the behavior of epidemics. In fluid dynamics literature, researchers focus on the propagation of pathogen-laden droplets and their interactions with air [1] . Biologists deal with the survival of airborne pathogens in macroscale [3] and their interactions with the human cells in microscale [4] . Furthermore, the medical literature conducts researches in cellular level to discover new drugs which cure the infectious diseases. In a larger scale, epidemiology literature focuses on the epidemic data to develop mathematical models for the spread of epidemics in time and space [5] . However, there is a need to merge all of these research efforts in a unified framework. Communication engineering approach can provide this framework by combining micro-and macroscale modeling issues. In this way, researchers will be able to utilize theoretical tools of communication theory in order to model the complicated nature of airborne pathogen transmission.

In the remainder of this article, we first review the airborne pathogen transmission mechanisms and the motivation to use MC as the enabler communication paradigm. Then, the communication engineering approach which merges different disciplines. In this approach, the layered architecture of MoHANET is presented in detail and open research issues are discussed. Finally, we give the existing and possible experimental techniques and conclude the article.

This section provides a brief overview for the main issues of the airborne pathogen transmission mechanisms. Then, the roles of molecular signals in the transfer of pathogens among humans are discussed.

A. Overview of Main Issues on Airborne Pathogen Transmission 1) Respiratory Activity and Droplet Size: Pathogen-laden droplets are emitted to the air from an infected human via respiratory activities such as coughing, sneezing, speaking and breathing. These activities have different initial droplet velocities allowing different propagation distances. For instance, the initial velocities for coughing and breathing are about 10 m/s [1] and 2.4 m/s [6] , respectively. Therefore, a cough can infect people at a greater distance than breathing in still air. Furthermore, the expiratory droplets are defined according to their diameters where aerosols and large droplets are assumed to have smaller and larger diameters than the size range 5-10 µm, respectively [1] . While speaking, sneezing, and coughing release more large droplets into the air, breathing mostly contains aerosols.

2) Air Distribution: In addition to the initial velocity, emitted droplets are influenced by the airflows, similar to a MC channel with drift. In outdoor environments, winds carry the droplets and dilute the concentration of pathogens via dispersion. Therefore, it is less probable to get infected in outdoor environments. However, in indoor environments such as hospitals, offices or residential buildings, airflows generated by ventilation systems are critical for the spread of pathogens due to the circulation of air in bounded conditions. Furthermore, personalized ventilation and exhaust systems are proposed as advanced ventilation systems to diminish the infection risk [1] . These air distributions are required to be considered for realistic indoor airborne transmission models.

3) Posture, Relative Orientation, Distance and Movement of the Human: For short distances, the posture, that is, standing, sitting or lying position, and the relative orientation of the infected and susceptible persons are important for the infection risk as shown in Fig. 1 . For instance, a doctor can reduce the exposure from an infected lying patient in a hospital ward via a standing posture and sideways orientation instead of face-to-face orientation [1] . Furthermore, a walking person can increase the infection risk in a closed and ventilated room by increasing the dispersion of the droplets [7] . Another important factor that influences the infection risk is the relative distance of the humans which is also referred as the social distance. Surely, the infection risk decreases, as the relative distance between two people increases.

The temperature difference between the human body surface and the surrounding air generates a thermal plume which is a buoyancy-driven upward flow of the surrounding air. As illustrated in Fig. 1 , this thermal plume leads to a convective boundary layer (CBL) around the human body, which should be taken into account for the movement of the droplets in the breathing zone [8] . This upward flow can change the channel impulse response via generating an upward drift for the pathogens during the reception into the human body. 5) Survival of Pathogens: Subsequent to a respiratory activity, all of the emitted pathogens may not survive. In [3] , it is shown that more than 80 percent of the influenza viruses cannot survive within one minute. However, these survival rates are severely influenced by environmental factors such as temperature and relative humidity (RH). While increasing temperature decreases survival rates of the pathogens due to its effect at molecular levels, increasing RH results in decreasing evaporation of droplets [9] . The decreasing number of pathogens results in a time-varying channel due to the dependence on the previous number of pathogens.

Via the aforementioned respiratory activities, a human can transfer pathogen-laden droplets to another human. This type of transfer (or communication) among humans is investigated in the medical literature where pheromone-based molecular signals are studied for the interaction of humans. In [10] , it is proposed that pheromones secreted from the axillary apocrine glands of women living in close proximity provides a synchronization in their menstrual cycle. Hence, molecular signals may give rise to some biological responses in human organism. As given in the next section, the transfer of the pathogen-laden droplets which cause infection can be considered in the context of MC.

In this section, we present a framework with communication engineering perspective to model the spread of infectious diseases through airborne pathogen transmission. Furthermore, open research issues are given.

As shown in Fig. 2 , the proposed framework merges all of the multiscale research efforts in various disciplines such as fluid dynamics, biology, medicine, and epidemiology under the umbrella of communication engineering. MC emerges as the key paradigm that connects the studies among different disciplines in macro-and microscales. First, the MoHANET is introduced through a layered architecture as depicted in Fig.  2 . Layers are associated with different disciplines from µm to km scale in this architecture where each layer sends its output to a upper layer. The first layer is defined as the physical layer where the infectious human (TX) emits pathogen-laden droplets through the communication channel (air) as illustrated in Fig. 1 . The next layer is the reception layer which takes place at the susceptible human (RX) and includes two sublayers, that is, outer and inner reception sub-layers. The outer reception sub-layer comprises the interactions of the facial sensory organs with the droplets and inner reception sublayer provides the details about the interactions of pathogens with the biological cells in the human body. The networking layer where infectious diseases spread among different people is given at the top of the MoHANET architecture. Here, methods from mobile telecommunication networks literature are exploited and the outputs of the lower layers are employed rather. The details of this layered architecture are introduced as follows.

A. Physical Layer 1) Transmitter: In a MoHANET, an infected person is considered as a TX and her/his respiratory activities determine the TX parameters such as initial droplet velocities and droplet size distribution [2] . The respiratory activities which are mentioned earlier can be classified as impulsive (sneezing and coughing) and continuous (breathing and speaking) emission signals. For continuous emissions, the frequency of the human exhalation is an influential factor for the transmission models. However, it is crucial to characterize speaking, since it is not always periodic and has more complex patterns than breathing. In addition, the respiratory organs such as nose or mouth affect the direction of the emitted signals. For example, the infection risk increases, when the TX uses the mouth instead of nose [1] . Furthermore, the convective boundary layer (CBL) of the human body, posture and relative orientation should be taken into account for accurate TX models. As mentioned earlier, the upward flow stemming from the CBL can affect the direction of the emitted pathogens in a TX model.

2) Channel: From the viewpoint of communication engineering, the channel is the physical medium between the TX and RX including the boundary conditions. As shown in Fig. 2 , channel modeling in the physical layer requires knowledge from fluid dynamics and biology due to the airdroplet interaction and survival of pathogens, respectively. The propagation dynamics of droplets can be examined under two subheadings depending on whether there is an external airflow or not. a) Still Air: In indoor environments such as residential buildings, it is generally assumed that there is no airflow, if there is not any ventilation system. After the emission of pathogen-laden droplets with an initial velocity, they are subject to Newtonian mechanics during their interaction with the air. Emitted droplets can be modeled as a cloud consisting of droplets and air particles. The movement of this cloud can be defined as a two-phase flow where these phases represent the gaseous state of air and liquid state of droplets [11] . Due to gravity, large droplets may fall earlier to the ground with respect to aerosols and evaporation can shrink the size of the droplets. As mentioned earlier, the temperature of the air and evaporation influence the survival rates of the pathogens. For continuous emissions, this fact can affect the channel memory, which is crucial for channel modeling. Furthermore, initial velocities of droplets determined by respiratory activities can Figure 3 . Two-layered Receiver.

give rise to short-term laminar and turbulent flows. These flows fade out as the distance between the TX and RX increases. b) Windy Air: For windy outdoor environments and indoor environments with airflows such as ventilation or wind arising from the open doors and windows, airflows dominate the propagation of droplets rather than other factors given for still air environments. The airflow which carries the pathogenladen droplets can be examined by advection and diffusion mechanisms. Briefly, advection results from the airflow velocity and diffusion depends on the turbulent eddies during the mass transfer [12] . It should be noted that molecular diffusion related with the thermal energy of molecules is negligible in macroscale. In order to calculate the concentration of droplets in time and space, deterministic and stochastic approaches which are based on differential Navier-Stokes and continuity equations are employed. For certain initial and boundary conditions, the solutions of these equations for deterministic concentration are known as Gaussian Plume for steady-state and Gaussian Puff Model for transient analysis [12] . Actually, the concentration and velocity of droplets are random processes whose mean values are represented by these deterministic solutions. Thus, stochastic differential equations are obtained which are non-trivial to solve as a closed form expression. Therefore, these equations are mostly solved by numerical methods using Eulerian and Langrangian approaches [12] . In addition, indoor ventilation types such as under floor air distribution, mixing, displacement, and downward ventilation should be incorporated into these airflow models. For example, downward ventilation can reduce the infection risk by diluting the dispersion of droplets [1] .

A human gets infected, when the transmitted pathogens are received into the body. As shown in Fig. 2 , the reception layer covers the issues related to biology and medicine in microscale where MC is utilized for the interactions of pathogens with the human body. The reception of these pathogens by the exposed human (RX) have not been well investigated, although there are myriads of theoretical, experimental and clinical studies for the propagation of pathogens. To this end, we propose a two-layered RX as shown in Fig. 3 and detailed below. 

The reception of pathogenladen droplets occur in the eyes [13] , mouth and nose for many pathogens such as influenza virus [1] . Hence, we define the first step of reception as the outer layer sensing for the reception via facial sensory organs as illustrated in Fig. 3 . The whole surface of the human face is also important for the reception, since an infection may occur by touching the face contaminated with pathogens and these organs consecutively.

Pathogen-laden droplets emitted via a respiratory activity propagate as a mixture of droplets and air particles, which can be represented as a cloud [11] . This cloud is affected by the momentum due to the initial velocity of droplets, gravity and buoyancy stemming from the temperature difference of the mouth and ambient air. According to this model, the trajectory of the cloud is given in Fig. 4 for a scenario that the TX emits pathogens by coughing. Here, there is no airflow in the channel and, the TX and RX are standing toward each other at the same height as illustrated in Fig. 1. Fig. 5 gives the change of the number of droplets in the cloud by taking settling and reception of droplets into account. The cross-section of the RX is assumed to cover a circular area including eyes, mouth and nose at the outer layer as illustrated in Fig. 3 . At this point, an analogy with the communication systems can be established by considering the infected state of the RX as symbol 1 and no infection as symbol 0. This reception is accomplished by a detection according to a threshold value (γ = 80) indicating the number of droplets required to become infected, as given in Fig. 5 . γ is a critical parameter in the airborne tranmission model, since it depends on the strength of human's immune system. To this end, biomedical data of humans such as body mass index, glucose level and whether or not having chronic diseases can be employed to estimate γ.

In addition to these issues in the outer layer, the posture, relative orientation and CBL of the RX should be taken into account for an accurate receiver model as considered for the TX. Furthermore, the reception of pathogen-laden droplets at the outer layer with different types of masks is an open issue to be investigated.

2) Inner Reception Layer: As shown in Fig. 3 , pathogens actually enter human body at the cellular level and increase their population. For example, viruses replicate themselves by inserting their genetic material (DNA or RNA) into human cells in two ways: They can bind their fusion (or spike) protein on specific receptor sites on the human cell or they can enter by using endosomes like a Trojan horse [4] . Their binding sites can have different concentrations in different parts of the body. For instance, severe acute respiratory syndrome coronavirus-2, which causes COVID-19, binds to angiotensin converting enzyme-2 receptors which are mostly found at upper respiratory tract [14] . While large droplets are effective in upper respiratory tract, aerosols can reach down to alveoli in lower respiratory tract. Hence, the droplet size can be effective to determine the infection risk according to the type of the disease. Moreover, the viruses diffuse among human cells, bind to receptors and copy their genetic material in a random way. All of these issues at the inter-and intracellular level need to be modeled for an accurate transmission model for the spread of infectious diseases in MoHANETs. These modeling efforts can also contribute to drug and vaccine developments.

What we examine up to here in lower layers of the Mo-HANET architecture is about the transmission of infectious diseases between two humans. However, these transmissions occur many times in an epidemic, which requires a perspective to handle the population as a connected group, that is, a network. In the networking layer, the details of the MoHANET architecture are presented in order to model the spread of infectious diseases in a large scale (km) within the communication engineering framework as shown in Fig. 2. 1) Mobile Human Ad Hoc Networks: In epidemiology literature, each human, that is, a node, can be represented as susceptible (S), exposed (E), infectious (I) or recovered (R) according to the SEIR-based models in the infectious disease modeling approaches [5] . According to the disease type, different combinations of these node types can be employed for the models such as SIR or SIRS. For example, COVID-19 is suitable to use all the node types due to a non-infectious incubation period. In the literature, the number of these node types are modeled by ordinary differential equations where the number of the nodes can be deterministic or a stochastic process. The transition among different types of nodes (S,E,I,R) are defined with certain rates which are obtained by fitting statistical epidemic data. In experimental studies, these data are obtained by oral surveys or exploiting wireless sensor network technology [5] . It is noteworthy that very few studies model the spatial change of the epidemic rather than its temporal change.

By utilizing the widespread SIR model, a MoHANET is given in Fig. 6 which gives both the spatial and temporal changes. As the time elapses, the number of nodes may alter and the nodes can make transitions between states such as S, I or R. For example, a susceptible node can become infected, if it is in the transmission range of an infectious node or an infectious node can recover after a certain period.

2) Transmission Types in MoHANETs: As illustrated in Fig.  6 , three transmission types are defined for the propagation of pathogen-laden droplets from the infectious nodes to the susceptible nodes as follows:

• Point-to-Point Transmission includes the communication between two nodes where the infectious and susceptible nodes are the TX and RX, respectively. • Multicast Transmission is the scheme that one infectious node spreads the disease to more than one node within its communication range. • Multiple-Access Transmission comprises the scenario where a susceptible node is exposed to pathogen-laden droplets from multiple infectious nodes.

Humans are susceptible to infectious diseases in indoor places such as public transportation vehicles, shopping malls or offices. However, this is not the case that is encountered continuously. Instead, the risk to get infected is intermittent due to the mobility of humans. As people displace, their smart phones can communicate opportunistically with each other as they are in the communication range. The same type of networking is also used in many applications such as wireless sensor, vehicular, and flying ad hoc networks. These dynamically changing structures defined as mobile ad hoc networks (MANETs) enable communication using the infrastructure at their location without a dedicated router. Therefore, a MoHANET can be resembled as a specific type of MANET, that is, a delay tolerant network (DTN) in which an end-to-end link among the nodes may not always exist. The nodes in a DTN store their data and wait until they find a suitable connection. By considering this waiting delay, the routing algorithms in DTNs provide the path to the desired user. Similarly, an infected human can store its pathogens until finding a susceptible human to infect via airborne transmission. Hence, we propose that opportunistic routing protocols such as epidemic or spray and wait can be adopted to model the spread of the infectious diseases. Interestingly, epidemic routing protocol which is a reference method for routing in MANETs was already inspired by the mechanism of infectious disease spread during an epidemic [15] .

In order to observe and model the airborne transmission mechanisms among humans, experimental setups and computer simulations can be employed. In this section, we present and discuss how the performance of the proposed methods in different layers of the MoHANET architecture can be evaluated.

In physical and reception layers, the emulation of breathing, coughing and sneezing in experimental setups are realized by respiratory machines or thermal manikins which can be heated to change their temperature. These devices emit tracer gases including droplets. The concentration of droplets is measured by air samplers or via imaging techniques such as particle image velocimetry which gives the velocity and directions of droplets [1] . Moreover, sprayer-based MC systems can also be used instead of respiratory machines, manikins and air samplers.

Albeit reliable results can be obtained by the physical experiments regarding the consideration of droplet-air interaction and airflows, the collected data have a low-resolution in space and time and experimental devices are expensive. Therefore, computational fluid dynamics (CFD) simulations are employed to evaluate the airborne transmission mechanisms with a high spatiotemporal resolution and less cost [1] . However, the simulation software programs are based on Navier-Stokes equations which lack the capability to model all of the effects during the transmission realistically.

These experimental techniques and CFD simulations can be employed to model the airborne pathogen transmission with communication engineering perspective for various scenarios between two humans. In a larger scale, for example, in a crowded city, it is essential to model the spread of infectious diseases with an approach that takes into account the interaction of people and their mobility in both time and space. The movement patterns of humans can be simulated by synthetic models such as random waypoint model or tracebased models which rely on real mobility data of mobile nodes as applied in MANETs. The adapted routing protocols for MoHANETs can also be evaluated in time and space by employing these mobility models according to the scenario via network simulation software.

This article presents a framework to model airborne pathogen transmission with a communication engineering perspective. First, airborne pathogen transmission mechanisms are reviewed and MC is utilized to model the propagation and reception of this transmission. The concept of MoHANET is proposed to handle the infectious disease spread modeling problem by using a layered structure in macro-and microscales. Furthermore, simulation techniques and experimental methods to model airborne pathogen transmission are reviewed and discussed. Throughout the article, open research issues possessing the potential for development opportunities are given. The efforts to model the infectious disease spread via airborne pathogen transmission with a novel approach given in this article has the potential for a holistic viewpoint. This communication engineering viewpoint can bring different disciplines such as fluid dynamics, medicine, biology and epidemiology together for accurate predictions about the spread of infectious diseases. Hence, the most proper intervention method (lockdown, social distancing, wearing masks, and so on) can be chosen and how it will be applied can be determined to stop the epidemics in an effective way.

Airborne spread of expiratory droplet nuclei between the occupants of indoor environments: A review

Communication through breath: Aerosol transmission

Survival of airborne influenza virus: effects of propagating host, relative humidity, and composition of spray fluids

How viruses invade cells

Dynamics of infectious diseases

Close contact behavior in indoor environment and transmission of respiratory infection

Performance of "ductless" personalized ventilation in conjunction with displacement ventilation: Impact of disturbances due to walking person (s)

Human convective boundary layer and its interaction with room ventilation flow

Mechanistic insights into the effect of humidity on airborne influenza virus survival, transmission and incidence

Regulation of ovulation by human pheromones

A molecular communication perspective on airborne pathogen transmission and reception via droplets generated by coughing and sneezing

Air dispersion modeling: foundations and applications

The severe acute respiratory syndrome

A pneumonia outbreak associated with a new coronavirus of probable bat origin

Epidemic routing for partially connected ad hoc networks

Turkey as a research/teaching assistant under the supervision of Assoc. Prof. Dr. Barış Atakan. His research interests include micro and macroscale molecular communications and molecular networks

He is currently an Associate Professor with the Department of