Emergency Messaging System for Urban Vehicular Networks Inspired by Social Insects’ Stigmergic Communication

Abstract

For occupant safety in vehicular networks, emergency messages derived from vehicular incidents should be exchanged only during their validity period and in zones containing involved entities. Problems arise for mobile entities in vehicular networks that change their location over time, where data may be further communicated in out-of-context space and time. Current solutions extend from the naive assumption of notifying every entity in the network about emergencies with data flooding and clusters and by means of specific communication only in the affected zones—geo-routing—of incidents’ relative data. However, delivering useless data to uninvolved entities results in wasted resources and more overheads in the former cases and the work of obtaining knowledge and secondary site services from neighbors in the latter. In this paper, we propose that the common task of disseminating emergency messages for occupant safety among entities should only be communicated only where and when useful, namely, if spatio-temporal constraints apply regarding those incidents. Our solution is inspired by the communication of working social insects that exchange data through pheromones regardless of closeness or knowledge among colony members for food retrieval. The results show that communication based on space–time constraints makes better use of resources than other solutions.

1. Introduction

Emergency message systems (EMSs) are mandatory in order to communicate and prevent further consequences wherever and whenever a traffic incident occurs in a vehicular network [1,2]. Especially in those applications that ensure the physical well-being of vehicle occupants at all times, also known as safety applications, vital information messages should arrive before any other [1,3]. However, on the one hand, the mobility factor in these networks works against the opportune reception of emergency messages by affected entities, given the intermittent connectivity with the current communication paradigms [4,5]. On the other hand, due to retransmission and/or non-controlled broadcast storms such as flooding approaches, non-affected entities might receive useless data, resulting in a waste of resources, even if the vehicles are grouped in clusters [6,7]. Furthermore, other approaches with spatial constraints, such as active or passive geo-routing, assume prior knowledge or the on-demand search for current neighbors, respectively, to the zone where data are communicated: distant or outdated data are received and then discarded by most entities with inherent channel bandwidth, buffer, and processing misuse [8]. In addition, traditional communication paradigms are still in place: static network architectures, the requirement for prior knowledge of communicating entities, assumed universal mobile global third-party communication and/or location services, human or manual involvement, and the largely time-only design of communication protocols.

Contrary to current approaches, we propose the use of emerging vehicular networks with dedicated infrastructures in dynamic urban environments and the delivery of only useful data under the premise that spatio-temporal constraints define both the useful data and the receivers that participate in the common task of the disseminating of emergency messages. Moreover, in nature, social insects base their communication and coordination on indirect pheromone messages, called stigmergy, such as in ants [9,10]. Therefore, any individual not having knowledge about other colony members is not an obstacle for successful communication; in this case, regarding food retrieval, these individuals are secondary, while achieving the common benefit is primary. In our proposed solutions, the communicating entities and the EMS are prioritized in the same way, as described in the literature as task-oriented communication [11,12]. This biological behavior inspired our EMS solution: failed or stranded mobile entities report to fixed entities located at crossroads acting as intermediaries. In turn, mobile entities are aware of the problem because they consume incident-related data—similar to a pheromone—from fixed entities. Data in the form of pheromones held by fixed entities describe close and recent incidents that only entities defined by space–time constraints are able to consume. Eventually, in a passive manner, a pheromone expires if it is not reaffirmed by the entity that provoked the incident, thereby marking the end of the incident.

Note that this behavior is actually a spatio-temporal behavior, since representative data in the pheromone are related to the location of food sources or incidents and are valid only during certain periods of time; in the literature, this is known as degradation [13,14]. Namely, data are only useful if they are communicated in the space–time they are created, while communicated data are only useful if they describe the close and recent space–time context. As seen, our proposed EMS communicates data under space–time constraints and is intuitively more effective than other solutions described previously that try to reach as many entities as possible and neighbor discovery overhead communication. We name our solution the Pheromone Emergency Message System For Vehicular Networks, abbreviated as PhEMS4VN. Highlights of its operation in dynamic environments include the addition of a communication protocol over the vehicular network infrastructure for the dissemination of emergency messages (EMs, not to be confused with EMS), the communication of EMs only, and not additional control messages, and the nonmandatory knowledge of communicating entities or the concurrent run of entities’ location services. The results of the experimental environments, with varying vehicular density, indicate that the total number of bytes communicated is lower than in other solutions while producing similar or identical results, achieving the more effective use of the spectrum and storage.

The remainder of this paper is organized as follows: in Section 2, we explore the theoretical background and review work related to our research; a proposed reclassification of approaches that address the problem of message dissemination among vehicles in those networks is shown in Section 3; in Section 4, we provide a detailed description of PhEMS4VN, define the system model for its operation, propose the experimental setup and performance metrics, and the observe results; finally, Section 5 includes the conclusions and future directions of this research field.

2. Related Work

We have divided this section into three parts concerning vehicular networks, current communication approaches in vehicular networks, and space–time-constrained communication.

2.1. Vehicular Networks

Vehicular networks, or Vehicular Adhoc NETworks (VANETs), are a particular kind of Mobile Adhoc NETworks, or MANETs [1]. The main difference is that MANETs are mostly composed of mobile nodes connected by wireless links that move on random trajectories ([15] in [16]), while in VANETs, fixed entities or roadside units (RSUs) might be present in addition to mobile entities or onboard units (OBUs) [1] that move along predefined roads [3]. A recent study also described a remote controller called a Trusted Authority (TA) [17] with functions such as registering nodes and network security. However, we aim to communicate with nodes faced with the task of communicating EMs only under space–time constraints; thus, security and related matters are outside of the scope of our research, and therefore this term will be intentionally omitted from the rest of this paper.

A major objective of VANETs is to provide physical welfare to the vehicle occupants through the opportune communication of data; therefore, several safety applications have been proposed and studied [1]. Other non-safety applications include route efficiency and fuel economy (see

Table 1. VANET applications classification and time requirements (taken from [1]).

Type	Delay Tolerance	Purpose
Safety	Most strict, latency should be less than 100 ms	Main purpose of these applications is to avoid collisions and accidents.
Non-Safety: Efficiency	On-time information needed, latency can be a few seconds	These are not as important but they can save time and money if they are in place.
Non-Safety: Comfort	No latency constraints as long as the information is received in a time bound	These applications provide information about hotels, restaurants, etc.
Non-Safety: Entertainment	Generally, these do not have any real-time constraints	These applications are related to multi-media sharing, the Internet, etc.

); for these applications, communication among entities in VANETs is mandatory.

In general, authors in the literature agree that there are three kinds of communication depending on the entities involved [1,3]: infrastructure-to-infrastructure (I2I), i.e., RSU to RSU communication; vehicle-to-infrastructure (V2I), i.e., OBU to RSU communication; and vehicle-to-vehicle (V2V), i.e., OBU to OBU communication. See Figure 1. Mobility issues, however, represent problems in the desired communication, namely packet routing and packet delivery [18].

2.2. Communication in Vehicular Networks

The simplest way to disseminate messages in a VANET is by means of simple broadcast and relay by flooding to entities in range [7]. However, two main problems arise: uncontrollable broadcast storms and non-receiving nodes not located inside the broadcast region. For this, Ghazi et al., 2020 [2], classified message dissemination approaches as follows:

• Dissemination by Intelligent Traffic Lights (ITL). Mobile entities that provoke or detect an emergency create messages communicated to ITLs, and then forward them to other ITLs and vehicles in advance before they reach the emergency zone. Another goal is to provide better routes by adjusting speeds or clearing lanes for emergency vehicles such as ambulances, for example.
• Dissemination using Internet of Things (IoT). Oriented to the detection and priority flow of emergency vehicles by means of Radio Frequency Identification (RFID) tags and Worldwide Interoperability for Microwave Access (WiMAX) wireless technologies with centralized management for the clearance of lanes and streets. Nevertheless, the proposed solutions are more focused on transportation than on communication issues.
• Dissemination using Priority Messaging. A broadcast is made before other messages and storms are controlled by an algorithm. Communication is location-sensitive by using the real-time updated location of where incidents happened; despite this, discrimination of messages for storm control generates overheads in message processing in the network.
• Dissemination using a Clustering Approach. Messages received, either through routing tables or opportunistic receipt, by a member are delivered in a unicast way to heads within groups of neighbor vehicles (clusters) that share a heading and speed at determined times in the environment. Those heads are responsible for broadcasting messages to all vehicles inside clusters. This approach retain a controlled and geographically limited scheme, but flooding still occurs, and it tries to reach every entity currently in the environment.
• Dissemination using Software-Defined Networks. Central or in-the-cloud entities control data flows by means of on-demand requirement analysis and, in this case, emergency messages are sent dynamically with higher priority among vehicles; however, connectivity with mobility is the main issue, not to mention the fact that data are communicated over previously deployed and dedicated expensive infrastructure or resources.
• Dissemination using Fog-Based Approaches. Fog computing basically delegates tasks from cloud to edge devices that will allow the dissemination of messages dynamically. These approaches nonetheless inherit the issues and requirements related to resources deployment as those seen under Software-Defined Networking (SDNs).
• Dissemination using 5G Technology. These approaches use the benefits given by 5G technology over mobile devices, such as connectivity and a broader bandwidth. Some drawbacks are that the operation is offered by third-party carriers, and the relatively new and expensive cellular technology requires time to be deployed to rural zones or zones through which highways are routed.

The classification of approaches for message dissemination provided by [2] does not match or exceed the performance of other feasible approaches, while the means of data transport are also classified as mechanisms. Other works include the following studies, including a proposed classification for message dissemination approaches in vehicular networks.

In [19], the authors propose a synchronous system for clustered vehicles traveling in the same direction for highway scenarios with only a V2V scheme. Their objective is to deliver the warning messages with higher priority, given that these kind of messages are time-sensitive. However, the application of this solution to urban scenarios where vehicles display more random movements is not considered, not to mention the fact that preserving synchronized clocks requires more control packets (overheads) than asynchronous ones.

An infrastructure-based solution involving clustered vehicles is proposed in [20]. RSUs are used as gateways in order to maintain the connection throughout the network, while they compute more distant movements and topology changes with the participation of an additional transport control center (TCC). The extra deployment of infrastructure might increase the costs of this infrastructure and the centralized entity.

The authors in [21] propose the communication of head clusters with close fixed nodes called sinks, as an additional infrastructure with fog computing, in order to avoid congestion. The purpose is to maintain connectivity throughout the network. Nevertheless, congestion is only avoided at sinks, since beacon packets that maintain the clusters are not avoided.

Two mechanisms for the dissemination of emergency messages are proposed in [22]. The first is a known code that reduces the packet size to be communicated among vehicles. The second is an inter-cluster communication protocol based on the A* algorithm [23] for message delivery. The solution is, however, cluster-based, with centralized (heads) dissemination and beacon messages that are larger than the emergency ones, meaning more overheads.

Reduced relay nodes in a geo-routing protocol are proposed in [24] for an overall reduced hop number and end-to-end propagation delay time. Frontier vehicles are selected as relays in consecutive areas that might not be permanently connected, i.e., the store–carry–forward approach is also applied. Since the scheme is only V2V and under ideal conditions, the authors do not mention how they deal with line-of sight and density problems.

A predictive vehicular location is proposed in [25] by means of a Kalman filter [26]. The main motivation is to avoid the excessive use of beacon messages and predict more distant positions. The vehicle with minimum error acts as the relay for the data geo-route to a destination zone; however, it is unclear if the vehicles execute the algorithm with limited resources or offline and how they obtain the actual position if the Global Positioning System, or GPS, is partially reliable.

The authors in [27] present a bio-inspired approach based on Particle Swarm Optimization (PSO) and aim to achieve a high data delivery rate, while the critical response time is reduced. Hello packets determine communication ranges and a Next Hop Vehicle (NHV), computed by the mentioned heuristic, relays data to the next region. Full knowledge of neighbor vehicles is assumed.

Ant Colony Optimization (ACO) is used in a bio-inspired geo-routing protocol [28]. Besides RSUs, a central entity sends agents (ants) with the purpose of finding less congested routes in the physical environment to provide shorter paths and travel times. High pheromone intensity means a high vehicular density to be avoided if a dynamic threshold is exceeded. Note that this solution requires reliable links and extra entities and its scope is to avoid traffic jams, instead of dealing with intermittent communication due to vehicular mobility.

A bio-inspired protocol based on spider behavior is presented by [29]. Their aim is to avoid V2I communication by dispatching agents from point to point in a spiderweb-like network and returning with the shortest path to the destination node. However, the reported overheads are higher than those of the compared works and reliable geographic services and connectivity is assumed.

Data dissemination under unstable links solution is provided by [30]. Both sparse and dense vehicular scenarios are considered, while alternative routes are computed if the main route presents failures. Despite the recovery mechanism, a greedy algorithm is used to calculate routes, where frontier nodes are used as relays if moving in the same direction.

In [31], the authors deal with the problem of broadcast storms using a barrier time mechanism, where a super-node delays data delivery in a cluster-based solution in order to control flooding instead of immediate retransmission. Nevertheless, their aim is to extend the messages network-wide to the majority of clusters.

The distributed data dissemination protocol DV-cast is presented in [32]. Fully connected, fully disconnected, and sparsely connected networks are observed in addition to the changes in topology at the same time. Broadcast suppression and store–carry–forward approaches are applied if connected or disconnected networks are detected, respectively. The geographical position, heading, and local or distant neighbors are identified through hello packets. An epidemic routing is used in sparse scenarios.

The authors in [33] propose location services for mobile nodes in order to locate and deliver data with reduced overheads. Described as semi-flooding, the nodes periodically broadcast their current position with a time stamp and maintain neighbor tables to be updated with received data. Two-dimensional and uni-dimensional scenarios as urban and highway scenarios are considered with uniform distribution densities. Problems such as topology changes or disconnected networks are not treated.

Head of clusters are computed according to the prey location behavior of whales in the work of [34]. The heuristic determines the optimal cluster formation, since the problem is NP-Hard. The revealed benefits of clustered-based solutions are reduced delay, scalability, no hidden node problems, and topology stability, among others. However, the vehicles must match the optimal speed and acceleration, while the iterative evolutive-like algorithm for cluster conformation must be run on every agent.

Open problems such as security in heterogeneous networks, non-standardization in car makers and local regulations, scalability, priority, and compatibility in future communication frameworks are explored in [35]. The authors also classify routing protocols in VANETs as follows:

• Topology-based protocols: Routing tables are created from the topology in the network. The same routes are maintained proactively or reactively, i.e., routes are calculated before being required or on-demand when required, respectively, or are locally proactive and externally reactive in hybrid scenarios.
• Position-based protocols: Information regarding position and geographic location is required for every node. Data are sent at once from the source to the destination in the non-DTN mode; meanwhile, in DTNs, data are communicated when connectivity happens. In hybrid modes, data run through both non-delay-tolerant networks (non-DTNs) and delay-tolerant networks (DTNs).
• Broadcast-based protocols: Packets are forwarded in multiple hops and delivered intentionally to all nodes in the network.
• Cluster-based protocols: Groups of vehicles are formed (clusters) where receivers forward data to designated heads responsible for delivering data to every node in the cluster. This approach is used to deal with blind flooding; however, the objective is to forward packets to all nodes in the network, similar to a broadcast, and thus it is questionable whether this can be described as routing.
• Geo-cast protocols: Destinations are computed by means of geographical locations.

Ross et al. [36] define routing as follows:

… ‘the network-wide process that determines the end-to-end paths that packets take from source to destination’…

Note that it is unclear if broadcast and cluster-based message dissemination approaches might be considered as routing given their controlled or uncontrolled forwarding to every feasible node in the network, as opposed to the other approaches that determine paths according to their path discovery algorithms. Furthermore, forwarded data are valid only where and when they represent close and recent events, i.e., distant and outdated events might be redundant, currently invalid, and physically far enough away that they have no effect on the receivers; however, network overheads, bandwidth abuse, and the local waste of storage and decision-making can happen if the acquired data have no use.

2.3. Spatio-Temporal Studies

Events in physically dynamic environments, specifically with spatial constraints, are related to each other according to the first law of geography by Tobler [37]:

… ‘everything is related to everything else, but near things are more related than distant things’…

Note that the same law is also applicable to more recent events in time. Thus, vehicular networks have an inherent interaction with their physical environment in which entities move; consequently, in addition to recent events that happen in time, the near spatial closeness of event also impacts the network behavior and current topology. Furthermore, the exchanged data describe these events. Therefore, the term spatio-temporal constraints in communication refers to the relationship between the space and time in which the events happen, and this needs to be communicated for the system to be useful.

In the literature, three basic behaviors and the combination of spatio-temporal phenomena in nature are presented: affectation, degradation, and propagation [13,14]. These behaviors are detailed in

Table 2. Comparison and description of natural spatio-temporal behaviors.

Behavior	Characteristics	Bio-Inspired Examples
Affectation	Maintains a sustained intensity, action or influence into a focalized location during a period of time	Mammals’ immune system
Degradation	Similar to affectation, however, the intensity, action or influence varies in space and/or time, or else, it depends on the beholder’s perspective	Stigmergy through pheromone
Propagation	Dissemination of the intensity, action, or influence through a physical environment and with continuous change and during the time the effect lasts	Epidemic dissemination
Combined behaviors	Spatio-temporal combinations of the behaviors or the observance of more than one at the same space and/or time	Pheromone trails and marks

, as well as analogue or equivalent bio-inspired behaviors examples:

Additionally to the spatio-temporal constraints, we state that spatio-temporal coupled entities are those that enable these constraints, i.e., they exist as message transmitters and receivers and share a region in space during a period of time, while feasible communication among them is achieved. If one condition, either space or time, is not satisfied, even if the other is satisfied, then the coupling is not achieved.

Finally, unlike routing and cluster approaches, previous knowledge among entities or discovery or location services is not mandatory, and the message receivers are determined by the space–time constraints, as in stigmergic communication, which is a type of indirect communication in which individuals take advantage of the environment by changing it [38]. Social insects such as ants use stigmergic communication through pheromones by means of traces to food sources in a spatial way, and the pheromone will lasts as long as it is reaffirmed depending directly on the time validity of the food source [9,10]. Therefore, stigmergic communication through pheromones is considered a spatio-temporal behavior (degradation), as noted in Table 2. Furthermore, communicating entities are defined through these spatio-temporal constraints; however, these entities become secondary as the due communication is established in order to achieve the common task of food retrieval or dissemination of emergency messages, in the cases of described social insects and vehicular networks, respectively [11,12].

3. Reclassification of Emergency Messages Dissemination Approaches

We propose a new message dissemination classification from global and local message dissemination perspectives, as shown in Figure 2, based on the previous classification from works observed in Section 2. A first superclass to be discerned is whether the approaches have the objective of making messages reachable for the majority of the population in the network, or in a reduced area in which data have space–time constraints and are considered relevant.

Scope and Limitations of Current Solutions and Why the Presented Work Is Necessary

It is clear that emergency message dissemination, a special and high-priority case of the task of communication in vehicular networks, remains an open problem mainly due to mobility of the entities involved. Therefore, several approaches and solutions have been proposed in the literature, as observed in Section 2. However, the main objective of the aforementioned approaches is to reach as many entities as possible with single broadcasts. Model systems include previous knowledge of the neighbors and/or data recipients, as in the routing cases, as well as location services for geo-routing, or both in the case of clustering approaches. We include a qualitative comparison of the dissemination approaches in

Table 3. Qualitative characteristics comparison of message dissemination approaches for vehicular networks.

Approach	Highlights and Advantages	Assumptions for Models Systems	Disadvantages and Drawbacks
Flooding	Simplest in dissemination	Previous knowledge of neighbors is not mandatory	If not controlled, broadcast storms hoard communications and messages overflow incoming buffers; the highest cost of message dissemination to as many vehicles as possible
Clustering	No broadcast storms are present; only heads disseminate EMs among clusters	Full knowledge of neighbors due to control messages for cluster formation and maintenance; an objective is that EMs reach most vehicles, whether they are useful or not	Control messages might be larger than EMs; cluster members must have the same speed and heading, which results in multiple executions of the cluster formation phase for urban scenarios and is impractical for sparse vehicular networks
Topology-based routing	Feasible delivery of EMs in unicast ways	Clusters of nodes are mandatory and the maintenance of partial (passive) or full (active) routes to the destination through the discovery of neighbors (control messages)	EMs might not arrive on time because of the on-demand route discovery for passive cases; the overheads of discovering routes actively represent more costs regarding bandwidth and local storage; in such cases, destinations must be known, at least in the last phase of message delivery
Location-based routing	Support of disconnected networks or DTNs	Knowledge of frontier nodes that relay messages to other zones of the environment; location services for mobile nodes are mandatory	Location services introduce more overheads to control messages and local storage for neighbor locations; DTNs might not be suitable for EM dissemination, as the delivery time limits are short
Geocast-based routing	Delivery of messages to entities currently in specific geographic zones; an effort to disseminate EMs only in locations where data are useful	As in the other routing approaches, routes to delivery zones are maintained; if not the recipients, previous knowledge of regions where data are to be delivered	Data might not be created where they are useful to deliver, and therefore a delay while data propagation happens might be present, i.e., time constraints are not fully covered; several hops might be present before final data delivery at the geographic destination, meaning that infrastructure availability is mandatory
Proposed space–time-constrained data delivery through pheromones	Data are only useful where abd when they are created; recipients are determined under space–time constraints and their previous knowledge and current location are not mandatory; no broadcast storms are present	Entities collaborate and communicate in order to achieve a common task; no cluster formation or maintenance is required; no control messages are communicated, and the topology in local storage is not required	Indirect and partially centralized communication through RSUs is mandatory; multiple notifications of the same incident are necessary

.

Note that issues such as data integrity, encryption, authentication, and privacy are security matters that are beyond the scope of the current research work and are not considered as characteristics in Table 3 or in this paper as a whole.

4. Proposed Approach: Pheromone Emergency Message System for Vehicular Networks (PhEMS4VN) and Experiments

This section includes a proposed feasible bio-inspired space–time communication approach to deal with the task of dissemination of emergency messages for mobile entities in vehicular networks and a comparison with flooding mechanisms.

4.1. Model System for PhEMS4VN

Contrary to any approach reviewed in Section 2, to make as many entities aware as possible in the whole environment, or to route packets through contiguous entities from points to other regions in the environment and/or exclusive entities, we propose that the communication enable collaborative behaviors among entities that pursue a common task and are in specific regions spatially close and recent in time in relation to where and when data are created and, thus, describe that part of the environment with more accuracy than other data. A feasible solution, as anticipated in Section 1, PhEMS4VN is a bio-inspired communication approach with a focus on achieving a common task through exchange of messages under space–time constraints based on stigmergy found in social insects colonies, more specifically ants’ pheromones. Therefore, in our mechanism, the messages are consumed only by entities that are coupled in space–time with a pheromone. The pheromone contains space–time data about incidents, provoked by entities with sudden stops that might endanger the occupants of other entities.

The operation of PhEMS4VN in a basis case is summarized as follows:

• A vehicle experiencing an undetermined failure stops and an incident begins.
• The failed vehicle periodically segregates messages and sends them to the closest RSU in order to indirectly inform other vehicles of the incident for as long as the incident lasts.
• The fixed entity or RSU acquires the message and creates a pheromone to be sent, also periodically, to those vehicles coupled in space–time with the incident so that they can be made aware of it.
• Vehicles around the location of the incident while it lasts consume pheromone messages and, if the planned route is obstructed by the incident, they will be instructed to take a detour.
• Note that the pheromones sent are the emergency messages disseminated to the mobile entities coupled in space–time with the incident.
• Entities not coupled in space–time are never aware of the incidents, as they probably are if broader other mechanisms are applied.
• If the incident ends and messages stop being produced, the pheromone will decay and die, thus delimiting its lifetime.

The notation used for further reference in the formalization of the mechanism is shown in

Table 4. List of notation for indicated data structure.

Set, Tuple, Interval or Variable	Elements, Entries or Limits	Description
$ME;M{E}_{inc};M{E}_{st}$	$\{m{e}_1,m{e}_2,...,m{e}_n\}$	Set of every mobile, provoked incidents, and space–time-coupled entities with incidents: representative $m{e}_k$
$FE$	$\{f{e}_i,f{e}_2,...,f{e}_m\}$	Set of fixed entities: representative fixed entity $f{e}_g$
$re\left(f{e}_g\right)$	$\{a,b,c,d\}$	Outbound exits of region in azimuth degrees defined for each fixed entity $f{e}_g$, where $a=0$, north; $b=90$, east,…
$ri\left(f{e}_g\right)$	$\{a^{\prime },b^{\prime },c^{\prime },d^{\prime }\}$	Inbound inlets of region in azimuth degrees defined for each fixed entity $f{e}_g$, where $a^{\prime }=0$, north; $b^{\prime }=90$, east,…
$IN{C}_k$	$\{in{c}_k^1,in{c}_k^2,...,in{c}_k^r\}$	Local set of all $(k,r)$ incidents provoked by mobile entity k, where k denotes the entity identifier and r denotes the number of incidents
$in{c}_k^p$	$(k,p,(x_k,y_k,t_{\alpha }^{k,p}))$	Spatio-temporal coordinates of location and initial time $\alpha$ of incident $(k,p)$ provoked by mobile entity k
${\Delta }_k^p$	$[t_{\alpha }^{k,p},t_{\omega }^{k,p}]$	Interval with final time $\omega$ of incident $(k,p)$, only known when finished
$mes{s}_k\left(in{c}_k^p\right)$	$(k,p,(x_k,y_k,\lambda t_{\alpha }^{k,p}),Stt)$	Message containing spatio-temporal coordinates of incident $(k,p)$ provoked and segregated by mobile entity k 1
${\lambda }_{\beta }^{g,k,p}$		Periodical physical time(s) when $mes{s}_k\left(in{c}_k^p\right)$ is acquired by fixed entity g and the related pheromone’s lifetime begins or is extended by reaffirmation
${\lambda }_{\gamma }^{g,k,p}$	$={\lambda }_{\beta }^{g,k,p}+\delta t$	Periodical physical time(s) when the pheromone relative to incident $(k,p)$ is going to expire after being started or restarted by fixed entity g at time ${\lambda }_{\beta }^{g,k,p}$
$\Delta {\varphi }_g^{k,p}\left(f{e}_g\right)$	$[1t_{\beta }^{g,k,p},\lambda t_{\gamma }^{g,k,p}]$	Physical countdown timer that extends the lifetime of the pheromone for $in{c}_k^p$ on $f{e}_g$
$I\left({\varphi }_g^{k,p}\right)$	$(g,in{c}_k^p,timer\_\varphi (g,k,p),({\iota }_a,{\iota }_b,...))$	Intensity in subregions $a,b,...$ computed with $value$ of $timer\_\varphi (g,k,p)$, normalized in interval $[0,1]$
$mes{s}_g\left({\varphi }_g^{k,p}\right)$	$(g,in{c}_k^p,timer\_\varphi (g,k,p),({\iota }_a,{\iota }_b,...))$	Pheromone message on $f{e}_g$ containing the indicated incident and intensities in subregions

¹ Stt = 0, initial segregation; Stt = 1, subsequent segregation; Stt = 2, incident is external to the mobile entity k.

.

In the context of distributed systems, we explore the events, internal and external, that happen in the involved processes or entities, either mobile or fixed, in 

Table 5. Event definition for mobile processes.

Internal Events’ Mobile Processes $m{p}_k$	Description
$Crt\left(in{c}_k^p\right)$	Creation of incident
$MGen\left(mes{s}_k\left(in{c}_k^p\right)\right)$	Generate incident message to be segregated
$MSt\left(in{c}_k^p\right)$	Countdown timer start or restart for indicated incident
$Mdec\left(mes{s}_h\left({\varphi }_g^{k,p}\right)\right)$	Intensities decapsulate from consumed pheromone
$MEn\left(in{c}_k^p\right)$	End of countdown timer for indicated incident
$MTrm\left(in{c}_k^p\right)$	Termination of indicated incident
External Events’ Mobile Processes $m{p}_k$	Description
$Seg\left(mes{s}_k\left(in{c}_k^p\right)\right)$	Segregate message of incident
$Con\left({\varphi }_g^{k,p}\right)$	Consume pheromone
$Act\left({\varphi }_g^{k,p}\right)$	Action after consuming pheromone

and

Table 6. Event definition for fixed processes.

Internal Events’ Mobile Processes $f{p}_g$	Description
$FDec\left(mes{s}_k\left(in{c}_k^p\right)\right)$	Decapsulate incident from message acquired
$FGen\left({\varphi }_g^{k,p}\right)$	Generate pheromone message to be notified
$FSt\left(timer\right(g,k,p\left)\right)$	Countdown timer start or restart for indicated pheromone
$Upd\left({\varphi }_g^{k,p}\right)$	Update intensity in pheromone with timer
$FEn\left(timer\right(g,k,p\left)\right)$	Countdown timer end for indicated pheromone
$FTrm\left({\varphi }_g^{k,p}\right)$	Termination of indicated pheromone
External Events’ Fixed Processes $f{p}_g$	Description
$Acq\left(mes{s}_k\left(in{c}_k^p\right)\right)$	Acquisition of incident message
$Ntf\left(mes{s}_g\left({\varphi }_g^{k,p}\right)\right)$	Notify pheromone

.

In the same line, we specify the respective algorithms for the mobile entity that provokes the incident, the RSU that acquires and notify pheromones, and the mobile entities that consume the pheromone messages in Algorithms 1–3. Note that in a future expansion of the basic case, i.e., if the EMs are to be segregated indirectly by consuming vehicles and used as mules to disseminate the message, if required, the incident is beyond the space limits of the original pheromone in Algorithm 4.

Algorithm 1: Incident creation and message incident segregation by mobile entity $m{e}_i$ on fixed entity $f{e}_h$

4.2. Simulation Parameters and Setup

In order to illustrate the operation of our protocol, we propose an environment with a central crossroad where a single incident happens and ten routes, with three of them directly obstructed and where/when emergency messages (EMs) will prevent a further traffic jam, as observed in Figure 3 for the space setup. The same environment is used to represent basis cases throughout the proposed protocol and comparison mechanisms of flooding in order to disseminate EMs. Two flooding variants are compared: flooding constrained to one and only one relay of an EM, and flooding that relays or forwards every message received. Both mechanisms are addressed as flooding-c and flooding-r, respectively. Meanwhile, the pheromone follows the system model at the start of Section 4. We also show timelines of these mechanisms in Figure 4.

Algorithm 2: Incident message acquisition on fixed entity $f{e}_h$ from mobile entity $m{e}_i$ and creation and notification of pheromone from fixed entity $f{e}_h$ for consumption on $m{e}_k$

Algorithm 3: Pheromone message consumption on mobile entity $m{e}_k$ from fixed entity $f{e}_h$

Algorithm 4: Non-local incident relay and message incident segregation by mobile entity $m{e}_j$ on fixed entity $f{e}_h^{\prime }$

The physical area where EMs are disseminated is the variant for the pheromone and both flooding mechanisms compared in Figure 5 and Figure 6. The detailed parameters for radio, communication space and time, messages, and vehicles which support our simulations are reported in

Table 7. Simulation parameters: radio.

Parameter	Pheromone with Degradation	Flooding with Control (Single Packet Relay)	Flooding with Relay (Multiple Packet Relay)
Radio standard	802.11 p	802.11 p	802.11 p
Radio frequency band	5.9 GHz	5.9 GHz	5.9 GHz
Radio channel (literature/FCC)	3/176	3/176	3/176
Radio channel central frequency	5.880 GHz	5.880 GHz	5.880 GHz
Radio bandwidth	10 MHz	10 MHz	10 MHz
Radio Tx power	0.50 mW	20 mW	20 mW
Ideal range (approximate)	80.91 m *	511.74 m *	511.74 m *

* Computed from frequency, Tx power, and isotropic antenna under the Free-Space Path Loss model with GatcomSUMO [45] v 1.06.

,

Table 8. Simulation parameters: space and time for communication.

Parameter	Pheromone with Degradation	Flooding with Control (Single Packet Relay)	Flooding with Relay (Multiple Packet Relay)
Global simulation space (area approximate)	47,700 m2	47,700 m2	47,700 m2
Potential communication area (approximate)	20,566.21 m2	47,700 m2	47,700 m2
Global simulation period	300 s	300 s	300 s
Incident period duration	200 s	200 s	200 s
Overall pheromone lifetime	210 s	-	-
Period of EM retransmission	5 s	5 s	5 s
Overall EM transmissions (occurrence)	42	41	41

,

Table 9. Simulation parameters: packets or emergency messages (EMs).

Parameter	Pheromone with Degradation	Flooding with Control (Single Packet Relay)	Flooding with Relay (Multiple Packet Relay)
Communicated types of packets	Emergency only	Emergency only	Emergency only
Relay counts	Only RSU relays packets	1	≥1
Transport protocol used	UDP	UDP	UDP
Packet header length	20 B	20 B	20 B
Packet payload length (average)	93 B	42 B	45 B
Payload contents	Entities’ IDs, space–time data for incident, pheromone levels, lifetime left, consecutive IDs	Entities’ IDs, space–time data for incident	Entities’ IDs, space–time data for incident, consecutive IDs

, and

Table 10. Simulation parameters: vehicles.

Parameter	Pheromone with Degradation	Flooding with Control (Single Packet Relay)	Flooding with Relay (Multiple Packet Relay)
Communication capabilities for EM	Rx only	Tx/Rx	Tx/Rx
Fixed entities able to communicate	Yes	No	No
Number of vehicular routes	10	10	10
Maximum speed in all street segments	40 km/h (11.11 m/s)	40 km/h (11.11 m/s)	40 km/h (11.11 m/s)
	1 veh/min and 50	1 veh/min and 50	1 veh/min and 50
Vehicular creation	2 veh/min and 100	2 veh/min and 100	2 veh/min and 100
rate per route	3 veh/min and 150	3 veh/min and 150	3 veh/min and 150
andtotal vehicles	4 veh/min and 200	4 veh/min and 200	4 veh/min and 200
	5 veh/min and 250	5 veh/min and 250	5 veh/min and 250
	6 veh/min and 300	6 veh/min and 300	6 veh/min and 300

, respectively.

All experiments were made with a machine with the following specifications:

• Hardware: Intel©i7-7700k microprocessor generic machine at Puebla, México, with 64 GB RAM, 4 GB VRAM, and a single 120 GB SSD for storage.
• Software: Linux Ubuntu 20.04.6 LTS x64 operating system
• Simulation environment and extra tools: Simulation of Urban MObility (SUMO) v. 1.16.00 [46]; OMNeT++ Discrete Event Simulation System IDE v. 6.0.1 (https://omnetpp.org (accessed on 25 February 2025)); veins v. 5.2 (https://github.com/sommer/veins, (accessed on 25 February 2025) and INET v. 4.4 (https://github.com/inet-framework, (accessed on 25 February 2025)); and Gatcom SUMO v. 1.06 [45] (code version 1.0.0 available at GitHub© repository (https://github.com/OjilvieAvila/PhEMS4VN, (accessed on 25 February 2025)), as we encourage the community to use these data and contribute to future research. Thank you).

4.3. Metrics and Results Summary

A direct comparison with the existing literature is non-viable, given the differences, for instance, with cluster approaches, given that we do not take advantage of control packets or explicit data routing. Metrics are also intended for a time-based focus only. For this reason, we propose the metrics for an intended fair comparison, as far as possible, among the mechanisms shown.

4.3.1. Packet Reception per Vehicle in the Environment

Metric description: Total emergency messages (EMs) or packets received per entity present in the environment during the whole simulation period.

Formula:

$$avg\_pck\_R{x}_{sim}={\displaystyle \frac{{\sum }_{i=1}^{n}pck\_Rx\left(m{e}_i\right)}{\left|ME\right|\setminus |M{E}_{inc}|}},$$

(1)

where $pck\_Rx\left(m{e}_i\right)$ is any packet received, such as a pheromone message ($mes{s}_g\left({\varphi }_g^{k,p}\right)$ as observed in Table 4), and $\left|ME\right|\setminus |M{E}_{inc}|$ is the total number of enabled vehicles or mobile entities minus entities that were involved in incidents during the indicated simulation.

Desired values: Less is better.

Discussion: The number of pheromone EMs is observed to be a stable constant number (four packets per vehicle) despite the increase in the number of vehicles that enter into the environment. The bandwidth used to communicate pheromone EMs remains proportional to the vehicular density (pheromone). In the last case (6 veh/min x route), there is no substantial difference between flooding-c and flooding-r, meaning that relayed messages are less commonly received with higher densities (see Figure 7).

4.3.2. Packet Reception per Vehicle Space–Time-Coupled with the Incident

Metric description: Total number of emergency packets received per entity present in the potential communication area during the incident period of time.

Formula:

$$st\_veh\_pck\_R{x}_{sim}={\displaystyle \frac{{\sum }_{i=1}^{n}pck\_Rx\left(m{e}_i\right)}{|M{E}_{st}|\setminus |M{E}_{inc}|}},$$

(2)

where $pck\_Rx\left(m{e}_i\right)$ is any packet received, such as a pheromone message ($mes{s}_g\left({\varphi }_g^{k,p}\right)$ as observed in Table 4), and $|M{E}_{st}|\setminus |M{E}_{inc}|$ represents the number of vehicles or mobile entities space–time-coupled with the incidents (i.e., in the potential communication area and during the pheromone’s lifetime or incident period) minus entities that were involved in incidents during the indicated simulation.

Desired values: Less is better.

Discussion: The number of pheromone EMs is observed to a stable constant number (7.5 packets per vehicle) despite the increase in the number of vehicles that are space–time-coupled. The bandwidth used to communicate EMs remains proportional to the vehicular density located in the communication area (pheromone footprint), compared with a linear increase in flooding-c. The last case remains practically unchanged for flooding-r due to previous considerations, as in Section 4.3.1, and the hidden node problem [2,34] (see Figure 8).

4.3.3. Total Payload Bytes Received

Metric description: Total number of emergency packets received per average payload length in bytes.

Desired values: Less is better.

Formula:

$$total\_pld\_bytes\_R{x}_{sim}=avg\_pld\_bytes\_le{n}_{sim}\ast total\_pck\_R{x}_{sim},$$

(3)

where $avg\_pld\_bytes\_le{n}_{sim}$ is the average payload length in bytes for the current simulation (see Table 9) and $total\_pck\_R{x}_{sim}$ is the total number of EM packets received by relevant mobile entities or vehicles during the simulation period. Additionally, a growth formula for the metric is also included:

$$growth\_rat{e}_{sim-(sim-1)}=(({\displaystyle \frac{resul{t}_{sim}}{resul{t}_{sim-1}}})-1)\ast 100,$$

(4)

where $resul{t}_c$ is the current simulation result (for example, 2 veh/min x route in pheromone) and $resul{t}_{c-1}$ is the previous simulation result (1 veh/min x route in pheromone) for each mechanism.

Discussion: Despite an apparent initial overpass of Rx bytes for the pheromone (see Figure 9), the average growth rate (obtained with Formula 4) is 45%, compared to 79% for flooding-c and 64% for flooding-r in Figure 10. This means that for more dense environments, the pheromone is observed to have a better or more efficient use of the bandwidth and throughput available in the wireless channel, as well as the buffer and storage space in receivers than flooding-c and flooding-r. This effect is due to the average payload length for each mechanism (see Table 9).

4.3.4. Detours Made per Vehicle Receiver of EMs

Metric description: How many vehicles had to detour per total EMs received

Desired values: More is better

Formula:

$$avg\_det\_R{x}_{sim}={\displaystyle \frac{|M{E}_{det}|}{{\sum }_{i=1}^{n}pck\_Rx\left(m{e}_i\right)}},$$

(5)

where $|M{E}_{det}|$ is the total number of vehicles that had to detour from their original route due to the received EMs and ${\sum }_{i=1}^{n}pck\_Rx\left(m{e}_i\right)$ is the number of overall received packets in the indicated simulation.

Discussion: The pheromone mechanism outperforms both flooding-c and flooding-r. The number of detours in every environment for the respective vehicular density was actually the same, however, the overall outcome of detours was achieved with a reduced number of packets or EMs received for the pheromone mechanism, i.e., the pheromone mechanism is more efficient with respect to the number of packets received than any of the flooding mechanism compared. It is also observed that no substantial change happens comparing both flooding mechanisms (see Figure 11).

4.4. Remarks About the Results

• We have observed that the pheromone mechanism performs better or more efficiently regarding channel bandwidth in time and storage space than flooding-c and flooding-r as the vehicular density increases.
• The pheromone footprint is considerably smaller than the whole potential communication space of any flooding mechanism.
• A major drawback for the pheromone mechanism is that packets’ payload length is about double that of any flooding mechanism, i.e, more overheads are generated in pheromone EMs. However, the increase in the overall received bytes ratio is lower for the pheromone mechanism than it is for the flooding mechanisms, and the pheromone notification is provided once more than for flooding in the experiments reported.
• The pheromone mechanism achieves the same detour outcome with fewer packets exchanged among entities coupled in space–time.

5. Conclusions and Future Work

We have presented a novel bio-inspired approach with the common task of disseminating emergency messages among mobile entities. The space–time constraints define both EM-receiving entities and useful data. The proposed approach, named PhEMS4VN, generally overcomes the global scopes of clustering and flooding approaches. The usefulness of the data received and space–time constraints demonstrated that this approach is more efficient in terms of the number of packets and the total bytes received than the compared approaches. We have also shown an updated and enriched emergency message dissemination approach classification in vehicular networks in space–time-constrained focused and global scopes, depending on the usefulness of the data received and the space–time constraints themselves. Future work will show that our proposed solution in simulated environments outperforms regular physical daily traffic disruptions, as we simulate more than one incident simultaneously within larger urban scenarios and more than one RSU or extend the same incident to other RSUs. We aim to propose a generalized metric of how space–time-constrained data impact on close and instant decisions and events in order to achieve the common task that enables communication, i.e., what the literature calls effectiveness [11,12].