Intelligent High-Awareness and Channel-Efficient Adaptive Beaconing Based on Density and Distribution for Vehicular Networks

Abstract

In vehicle ad hoc networks (VANETs), a beacon is a periodic message sent to nearby vehicles containing essential details like the sender’s vehicle ID, location, speed, and direction. Maintaining the freshness of this information without causing network congestion requires adaptive beaconing to adjust to changes in mobility and network density. Our research, based on extensive simulation experiments, identifies specific parameter sets optimal for adapting beaconing rates to different scenarios. From this analysis, we introduce a novel scheme called high-awareness and channel-efficient adaptive beaconing (HACEAB), employing fuzzy logic to adapt to various environments and conditions. Initially, the protocol gauges network density using an adaptive threshold function, followed by estimating the node spatial distribution through the quadrat statistic method to discern uniform distribution or clustering. Utilizing these data, the protocol adjusts beaconing rates via appropriate input parameters for the fuzzy logic system. Remarkably, HACEAB represents the first beaconing scheme capable of simultaneously adjusting to changes in network density and spatial distribution. Furthermore, the protocol enhances performance by adapting transmission power to fluctuations in node density and distribution. NS-3 simulations validate the efficacy of these improvements.

1. Introduction

Since the number of vehicles on the road increases every day, traffic accidents and congestion become critical global issues that draw attention all over the world [1]. A vehicular ad hoc network (VANET), an emerging intelligent transportation system (ITS) technology, holds promise to resolve these issues and provide further services. VANETs offer a wide array of applications, encompassing safety, traffic efficiency, control, comfort, and entertainment. Safety-related applications within VANETs involve features like intersection collision alerts, lane change/merge warnings, blind spot warnings, precrash sensing, road condition alerts, emergency vehicle warnings, electronic brake lights, and various other safety-oriented functionalities [2,3].

Various types of messages are exchanged within VANETs, with beaconing messages being particularly significant (alternatively known as cooperative awareness messages or basic safety messages). A beacon, a periodic transmission, is disseminated to all one-hop neighbors and contains fundamental details about the transmitting vehicle, including its ID, location, speed, and timestamp (as illustrated in Figure 1). A VANET operates within the 5.9 GHz frequency band [4]. Safety-centric applications, as highlighted in [5,6,7], typically necessitate a minimum update frequency of 10 Hz.

In a VANET, there are seven channels: one control channel (CCH) and six service channels (SCHs). According to the IEEE 802.11p WAVE standard [8], the CCH is designated for safety-related messages, service announcements, and context-aware information. Meanwhile, the SCHs are allocated for comfort and driving efficiency messages. Beacons are exchanged over the control channel [9,10], which is the most important channel in VANET communications.

The dynamic topology is a challenging characteristic in VANETs, where the number and distribution of vehicles can quickly change. Furthermore, exchanging beacon messages at a fixed rate can cause the communication channel to be congested, which will affect the overall performance of vehicular networks [11,12,13,14,15].

Cooperative awareness is critical for VANETs because many applications depend on it to make the correct decision and launch the appropriate application to avoid dangerous situations. There are trade-offs between information freshness, cooperative awareness, and channel congestion. Sending beacons at high frequency is favorable because it can possibly increase cooperative awareness and improve the freshness of exchanged information. However, this leads to more congested networks and increases resource consumption. On the other hand, sending beacons at low frequency reduces network congestion and saves more bandwidth but leads to outdated information.

Transmission power plays a very important role in terms of cooperative awareness and channel congestion. Increasing transmission power can allow more vehicles to communicate with each other and hence improve cooperative awareness. However, this leads to more congested channels and network resource saturation. On the other side, reducing transmission power can alleviate channel congestion since the number of vehicles that can hear beacons are fewer, but this affects cooperative awareness. Fixed transmission power is inappropriate in a VANET environment. Therefore, adaptive transmission power is highly recommended besides the adaptive transmission rate [16].

Due to those challenges, it is important to come up with an adaptive beaconing protocol that adapts its transmission parameters (transmission rate and transmission power) based on different situations, conditions, and environments. Thus, the main goal is to increase the cooperative awareness and information freshness without causing congestion in the network.

In this paper, we present a new approach named high-awareness and channel-efficient adaptive beaconing (HACEAB), which employs fuzzy logic to dynamically adjust beaconing in diverse networking scenarios.

The primary contribution of HACEAB is to utilize an adaptive threshold function which defines the network density. HACEAB uses the quadrat statistical method to determine the spatial distribution of vehicles in the network. Then, based on the determined network scenario, the protocol applies the appropriate input parameters to a fuzzy logic system, which adapts the beaconing rate. Afterwards, HACEAB further optimizes the performance by adapting the transmission power to variations in node density and distribution.

The remainder of this paper is structured in the following manner: In Section 2, we present a review of related work on beaconing in VANETs. Section 3 offers an overview of the fuzzy logic system and its application in VANETs. Section 4 explores the effects of various parameters on beaconing. Our proposed beaconing protocol is detailed in Section 5. Section 6 delves into the performance evaluation and simulation results. Finally, Section 7 provides the conclusion of the paper.

2. Related Work

Channel congestion is a well-recognized problem in vehicular ad hoc networks. Literature approaches propose to mitigate this problem by regulating the network load according to different parameters such as vehicle density, vehicle speed, and distance between vehicles. In particular, each vehicle monitors and/or calculates the selected parameter(s) around itself and utilize this information to adapt its transmission parameters.

In [17], the authors introduce a protocol called adaptive beaconing rate (ABR), which employs fuzzy logic to adjust the frequency of beacon transmissions. The ABR system utilizes vehicle status and the proportion of vehicles moving in the same direction as inputs to regulate the beacon frequency. Similarly, in [18], a fuzzy logic-based protocol that adjusts the beacon interval using inputs such as vehicle speed, number of single-hop neighbors, and carry time is presented. However, both protocols have only been tested in urban environments and do not incorporate channel status in their beacon rate adaptation mechanism.

A Joint space-division multiple access and rate control (JSRC) scheme is introduced in [19]. JSRC joins an adaptive rate control scheme with a space-division multiple access (SDMA) approach to alleviate interference between vehicles and solve the hidden-node problem. In JSRC, vehicles must be sufficiently separated to share the same time slot for message transmission. JSRC utilizes carrier-sense multiple access with collision avoidance (CSMA/CA), allowing for multiple message transmissions per road segment.

A beacon rate adaption based on fuzzy logic (BRAIN-F) to control the beacon frequency is proposed in [20]. The fuzzy logic system in BRAIN-F operates across two tiers. At Level 1, it takes into account the vehicle’s individual velocity and the average directional speed of neighboring vehicles as inputs, generating a traffic density degree as its output, which subsequently becomes an input for Level 2. At Level 2, the fuzzy logic system factors in vehicle traffic density (output from Level 1), location status, and vehicle status to determine the beacon rate degree as the final output. However, the scheme does not incorporate considerations for the channel status.

The impact of radio signal shadowing dynamics on the efficiency of beaconing protocols is studied in [21]. An expression for the upper limit of the channel utilization is derived, and it is shown that beaconing should not exceed approximately 60% of the control channel interval. Then, a dynamic beaconing approach (DynB) is proposed. DynB employs two control parameters: $b_t$, representing the proportion of occupied time between $t-I$ and t, alongside N, denoting the number of neighbors. These variables are utilized to adjust the beacon interval I towards a target value $I_{des}$, ensuring that the channel utilization remains below a specified threshold $b_{des}$.

In [22], based on noncooperative game theory, a beacon rate and awareness control method is introduced. In order to ensure fairness among vehicles with similar needs, each vehicle is allocated a beacon rate tailored to its specific requirements.

A distributed beacon congestion control scheme is proposed in [23]. This scheme considers link conditions to control beacon activities, which means a higher beacon rate will be assigned to the vehicles that have more neighbors and better conditions of links.

In [24], an adaptive beacon generation rate congestion control scheme for urban environments is proposed. In this method, the beacon generation rate is adjusted to the level of vehicle density in the network.

A speed-adaptive beaconing method is proposed in [25]. This approach considers the vehicle count when computing the forwarding probability.

The study described in [26] introduces a beacon rate control technique based on link conditions. In that approach, nodes with a greater number of neighboring nodes are allocated a higher beacon rate.

In [27], the authors present an adaptive method for determining the beacon interval within a position-based protocol, taking into account vehicular traffic conditions. This strategy relies on three main mobility factors: vehicular density, the average speed difference among neighboring vehicles, and the spatial distribution of these neighbors.

A context-aware hybrid beaconing algorithm is proposed in [28], aiming to enhance safety by disseminating driver status via beacons to neighboring vehicles in IEEE 802.11p-based networks. The comprehensive context is taken into account to derive driver status and initial network status. Real-time network congestion status is considered, and adjustments to transmission range and power parameters at the PHY layer, as well as beacon interval and contention window at the MAC layer, are dynamically made using the enhanced distributed channel access (EDCA) mechanism.

Ensuring the effective adjustment of beaconing transmission parameters requires the consideration of various factors such as channel conditions, traffic situations, and link quality. Many existing schemes in the literature concentrate on just one or two of these aspects, constraining their adaptability. In our study, we introduce an innovative scheme that simultaneously takes into account all three crucial factors. Furthermore, our proposed scheme is crafted to dynamically respond to alterations in network density and environmental conditions, offering a comprehensive and adaptive strategy for beaconing across diverse scenarios.

3. Overview of Fuzzy Logic System

Fuzzy logic is one of the most efficient soft-computing methods for solving uncertainties in high changing and dynamic environments such as VANETs [29,30].

3.1. Fuzzy Logic System Components

In [31], L. Zadeh introduced fuzzy logic that uses a range of input values and produces estimated results as output. As shown in Figure 2, a fuzzy logic system basically includes the following components:

• A fuzzifier;
• An inference engine;
• A defuzzifier.

During the fuzzification stage, precise values (inputs) are transformed into fuzzy values. Subsequently, the inference engine interprets the fuzzy values obtained from the fuzzifier into alternative fuzzy values using IF-THEN rules. In the defuzzification phase, fuzzy values are converted back into precise values (outputs). Three frequently utilized defuzzification approaches include the center of gravity method, height method, and mean of maximum method.

3.2. Fuzzy Logic Application in VANETs

Fuzzy logic has been employed in VANETs to improve decision-making processes and mitigate communication and computation delays. VANET applications leveraging fuzzy logic include:

• Routing algorithms;
• Broadcasting;
• Contention window size adjustment;
• Clustering control;
• Data aggregation;
• Localization;
• Trust management.

In [32], the authors proposed an innovative intelligent geographic routing scheme that uses fuzzy logic system to determine the ideal node to forward the packet. The fuzzy logic system uses direction, distance, achievable throughput, and link quality to generate a weight value that determines the optimal forwarding node. In [33], another geographic routing protocol is proposed which utilizes fuzzy logic to make packet forwarding decisions. It uses direction and distance as input to the fuzzy logic system. A routing method, GRVAD (geographic routing method based on velocity, angle, and density), leveraging the fuzzy logic system is introduced in [34]. This method uses relative velocity, angle between nodes, and neighbor node density as inputs for fuzzy logic. In order to store data in vehicular ad hoc networks, a fuzzy logic-based transferring data protocol is proposed in [35]. Considering throughput, vehicle velocity, and bandwidth efficiency, the fuzzy logic system manages the next data carrier node selection (short-term evaluation), while using a Q-learning algorithm to achieve long-term efficiency.

In [36], a sender-oriented broadcasting scheme based on fuzzy logic is proposed. The fuzzy logic system uses vehicle velocity, antenna height and the number of neighboring vehicles heading in the same direction as inputs. Thus, the fuzzy logic system selects the backbone node to forward data packets. A fuzzy logic-assisted cross-layer receiver-oriented broadcast scheme is proposed in [37]. In that scheme, based on locally calculated mobility and coverage factors, the proposed fuzzy logic system checks the qualification of each receiving vehicle to rebroadcast the message. Also, to avoid packet collision and increase the delivery rate, considering the local density and distribution of vehicles, a fuzzy logic-based system is used to perform contention window adjustment at the MAC layer.

The authors in [38] introduce a solution employing a fuzzy logic system at the MAC and network layers to effectively broadcast safety messages. The proposed rule-based model enhances the contention window (CW) and relay selection process for adaptation to diverse traffic conditions.

In [39], a cluster head selection scheme that utilizes fuzzy logic is introduced. A node is selected as a cluster head based on criteria related to each candidate node: received signal strength, speed of vehicle, vehicle location, spectrum price, reachability, and stability of node.

In [40], a structure-free aggregation scheme for VANETs is presented. This scheme uses fuzzy logic to make aggregation decisions relying on an extensible and resilient set of criteria. Fuzzy logic rules are utilized to base aggregation decisions upon qualitative metrics including induced quality loss caused by aggregation.

Moreover, a fuzzy logic-based scheme for localization in VANETs has been proposed in [41]. This scheme combines a weighted centroid localization (WCL) with fuzzy logic. Two parameters are considered as fuzzy logic inputs, the distance between sender and receiver and the heading information. Beacons’ (periodic) messages are utilized to exchange status information such as ID, location, and velocity. The fuzzy logic module generates the output in the form of a weight value. Using WCL, a weight value is assigned to each neighboring vehicle. Finally, the vehicle location is estimated using the neighboring vehicles’ weighted coordinates.

To facilitate trust management within VANETs, ref. [42] introduces a fuzzy trust model. This model, based on experience and plausibility, conducts a sequence of security assessments to verify the reliability of information received from authorized vehicles in the network. In [43], a fuzzy-based trust authentication scheme employing the Mamdani fuzzy inference system to identify malicious nodes is utilized.

4. Investigating Parameters’ Impact on Beacon Rate Adaptation

Several parameters impact beacon rate adaptation. In this section, our discussion focuses on five influential parameters and their impact on beacon rate adaptation.

4.1. Parameters Used and Their Significance

In this subsection, we discuss the parameters used in our proposed protocol and their significance. In VANETs, the beaconing rate can be influenced by different parameters. Through surveying the literature and conducting extensive simulations, we found that beaconing rate adaptation was highly affected by five parameters as follows: channel busy time, mobility, packet delivery ratio, number of neighbors, and signal-to-interference-noise ratio.

The channel busy time (CBT) is a good measure for channel load [44,45]. A high CBT means the channel is more occupied; in that case, the beaconing rate should be reduced to alleviate the load on the channel. The CBT is defined as the proportion of time the channel is sensed occupied by vehicle x (vehicle x is the vehicle that will adapt its beaconing rate) during a monitoring interval because of vehicle x or its neighbors’ activity. The acquisition of the CBT value occurs when the signal strength received by the measuring node surpasses the clear channel assessment threshold (CCA) over a specified time window. The $ChannelBusyTimeFactor$ is computed as:

$$ChannelBusyTimeFactor=\frac{CB{T}_x-CB{T}_{min}}{CB{T}_{max}-CB{T}_{min}}$$

(1)

where $CB{T}_x$ represents the duration where vehicle x detected the channel as occupied during the monitoring interval. $CB{T}_{min}$ and $CB{T}_{max}$ denote the minimum and maximum channel busy times for vehicle x and its neighbors, respectively.

Mobility is one of the most important parameters in VANETs, since it reflects the vehicles’ movement and traffic status. In high traffic mobility, the beaconing rate should be increased to ensure that more packets reach the fast and distant vehicles. The proposed schemes in the literature use a mobility factor that takes into account either the distance between vehicles or the speed of vehicles [20,36,41,46,47,48,49]. However, in this work, we use a more representative vehicle mobility factor, which we proposed in our preliminary work [50,51,52]. It considers both the distance between vehicles and the speed of vehicles. Hence, we obtain the mobility factor by computing the distance $d_i$ between vehicle x and its neighboring node i. We use the speed to calculate the distance, $d_{i{T}_b}$, traveled by vehicle i during the beacon interval $T_b$:

$$d_{i{T}_b}=s_i\ast T_b$$

(2)

where $d_{i{T}_b}$ is the distance traveled by vehicle i during the beacon interval, $s_i$ is the speed of vehicle i, and $T_b$ is the beacon interval. We then calculate the average K of ($d_i$ + $d_{i{T}_b}$) over all the neighboring nodes ($i=1,\dots ,n$) of vehicle x, as shown in Equation (3).

$$K={\displaystyle \frac{{\displaystyle \sum _{i=1}^n}(d_i+d_{i{T}_b})}{n}}$$

(3)

Determining the final mobility factor value requires dividing K by the transmission range R and subtracting a smoothing factor $\alpha$, as depicted in Equation (4).

$$Mobility={\displaystyle \frac{K}{R}}-\alpha$$

(4)

The packet delivery ratio (PDR) reflects the link quality [53]. The PDR is influenced by the quantity of transmitting vehicles. With an increase in the number of sending vehicles, the PDR tends to decrease. This is attributed to the diminished success probability in accessing the medium, resulting from an escalation in the number of vehicles contending for channel access [54]. The PDR is characterized as the ratio of packets successfully received to the total number of packets transmitted within the current time interval. In this time frame, any packet received from a neighboring vehicle situated within the transmission range is deemed as a successfully received packet.

The number of neighbors (local density) is a very important parameter that needs to be considered in VANETs [18,21,36,55,56,57]. The number of neighbors needs to be estimated before attempting to adapt the beaconing rate. In the presence of a substantial number of neighboring nodes, it is advisable to decrease the beacon frequency to alleviate the load on the channel. The count of neighbors is determined by the number of adjacent vehicles within the transmission range of vehicle x.

The signal-to-interference-noise ratio (SINR) is computed by utilizing beacons obtained from neighboring vehicles. It is utilized to estimate the link quality [53,58]. A large value of the SINR could mean a neighboring vehicle is closer to the sending vehicle, requiring beacon frequency to be reduced. The SINR is computed using the following equation:

$$SINR=\frac{S}{I+N}$$

(5)

In this context, S denotes the received signal power, I represents the cumulative power of interference signals, and N indicates the noise power [59]. Subsequently, the span SINR is calculated to obtain the SINR factor, given by:

$$SINRFactor=\frac{SIN{R}_x-SIN{R}_{min}}{SIN{R}_{max}-SIN{R}_{min}}$$

(6)

where $SIN{R}_x$ is the SINR computed by vehicle x. $SIN{R}_{max}$ = $max\left[SIN{R}_i\right]$ and $SIN{R}_{min}$ = $min\left[SIN{R}_i\right]$, where $i=\left\{1,\dots ,n\right\}$.

4.2. The Impact of the Selected Parameters

We investigated the impact of the five aforementioned influencing factors. The preliminary results of our investigation were reported in [50,51,52]. This led us to establish in this work, and through extensive simulation experiments, that specific sets of parameters were best suited to adapt the beaconing rate for specific environments. Set 1 comprises the SINR, the count of neighboring nodes, and mobility. Set 2 is constituted of mobility, the CBT, and the PDR. Set 3 encompasses the SINR, the number of neighboring nodes, mobility, the CBT, and the PDR.

We ran thousands of simulations and obtained consistent results that confirmed the effect of these parameters in different scenarios. We ended up with the following observations (

Table 1. Summary of experimental findings.

	Scenario 1	Scenario 2	Scenario 1	Scenario 2
Set 1 (SINR, number of neighbors, mobility)	✓
Set 2 (Mobility, CBT, PDR)		✓
Set 3 (SINR, number of neighbors, mobility, CBT, PDR)			✓	✓

). In nondense networks, we had two scenarios. The first scenario was a nondense clustered network environment. In that scenario (scenario 1), we obtained good performance when we adapted the transmission rate based on Set 1’s parameters. The second scenario was a nondense uniformly distributed network environment. In that scenario (scenario 2), we obtained good performance when we adapted the transmission rate based on Set 2’s parameters. In the third scenario, we found that in the case of dense networks, regardless of the type of environment (uniformly distributed or clustered), we obtained good performance when we adapted the transmission rate based on Set 3’s parameters. These findings appeal to intuition. For instance, in nondense networks in urban environments, Set 1’s parameters did not work very well, since a small value of the SINR does not always mean a vehicle is far away; it could be close, but there are some impeding objects such as buildings in between the two vehicles. From the above observations, it is clear that there is no one solution that fits all and every scenario requires different parameters to be considered. This strongly motivates the design of a protocol that adapts to the different scenarios.

5. Proposed Adaptive Beaconing System

We present the design of an intelligent high-awareness and channel-efficient adaptive beaconing protocol (HACEAB). Figure 3 shows beaconing in WAVE, while the HACEAB flow chart is shown in Figure 4.

In our protocol, we assume that each vehicle is equipped with a Global Positioning System (GPS) and a half-duplex transceiver, enabling it to have access to its position information. Furthermore, it is assumed that each vehicle appends its identification, speed, and coordinates to the transmitted beacon packet (Figure 1). The receiving vehicle utilizes these data to compute the input values.

First, HACEAB estimates the network density using an adaptive threshold function that depends on the beacon frequency to determine whether the network is dense or not dense. If the network is dense, it uses the dense conditions’ fuzzy rules (we explain it in detail later on). If the network is not dense, based on the exchanged position information, it uses the quadrat method to determine the type of distribution, whether uniform or clustered. Based on the determined environment, HACEAB uses the appropriate fuzzy rules (i.e., in case of a nondense network, if based on the quadrat method, nodes are clustered (scenario 1), and HACEAB uses the fuzzy rules of Set 1. Otherwise, it uses the fuzzy rules of Set 2. After that, HACEAB calculates the next beacon rate utilizing the rate control function.

Now, we explain how the flow-chart activities work in details. The first step is to receive beacons from other vehicles. Then, the protocol determines the local density (number of direct neighbors). After that, the protocol calculates the density threshold and compares it with the local density as is explained in the following subsection.

5.1. Density Estimation

In this section, we explain the method used by the proposed protocol to determine the network density level (dense or not dense). First, we derive an expression to estimate the density threshold ($DT$). IEEE 802.11p operates in the 5.9 GHz frequency band, with 7 channels of 10 MHz each. One of these channels is a control channel (CCH) and the other six are service channels (SCHs). The CCH is reserved for safety related messages, service announcements, and context-aware information, while SCHs are reserved for comfort and driving efficiency messages. Beacons are exchanged over the control channel (CCH) [9,10]. The channel time in IEEE 802.11p is divided into synchronization periods of 100 ms each, composed of equal-length alternating 50 ms control channel interval (CCHI) and 50 ms service channel interval (SCHI). There is a 2 ms guard interval (GI) at the start of each one of the 50 ms time intervals to account for the timing inaccuracies of the communicating devices and the radio switching delay [9,10,60].

The beacon load ($BL$) is the maximum portion of the CCHI that can be used by beaconing, and the beacon rate ($BR$) is the number of packets sent per second. Consider a frequency of 10 Hz and a CCHI of 50 ms. Also, consider a maximum beacon load of 60% [21] and an approximate beacon transmission time of 1 ms [61]. Over a 1000 ms interval, the density threshold ($DT$) can be estimated as follows:

$$DT=(1000/(CCHI+SCHI\left)\right)\ast (CCHI-GI)\ast BL\ast (1/BR)$$

(7)

Using the following values (discussed above) $CCHI$ = 50 ms, $SCHI$ = 50 ms, $GI$ = 2 ms, $BL$ = 0.6, and $BR$ = 10, in Equation (7) results in a $DT$ of 28 vehicles. Every vehicle determines the number of direct neighbors, $ND$, and compares it to $DT$. We consider the network to be dense if $ND$ > $DT$ since in that case the channel is saturated. Otherwise, the network is not dense. The density is estimated by each vehicle utilizing the received beacons from neighbors within the transmission range.

If the network is determined to be dense, the protocol uses Set 3’s parameters as inputs to its fuzzy logic system. Otherwise, the protocol determines the type of environment to select the suitable set of parameters that it needs to use in its fuzzy logic system as is explained in the next subsection.

5.2. Characterization of Node Distribution

We use the $quadrat$ $statistic$ $method$ to characterize the node spatial distribution [56]. The quadrat method measures how a set of nodes are evenly spaced. We obtain the quadrat value (Q) which is defined as the ratio of the variance to the mean. When Q > 1, it means that vehicles are clustered in 1D and oriented uniformly (scenario 1), while Q near 1 means vehicles are uniformly randomly distributed in 2D (scenario 2).

From Table 1, in nondense conditions, if the quadrat value characterized the environment as scenario 1, Set 1 is used, and if the quadrat value characterized the environment as scenario 2, Set 2 is used. If the network is dense, Set 3 is used. Once the protocol selects the suitable set of parameters, it utilizes the fuzzy logic system to produce a final congestion rank.

5.2.1. Fuzzification

For real-time applications, the triangular membership function stands out as one of the most frequently employed, effective, and recommended functions. To refine our fuzzy logic system’s membership functions, we employed a trial-and-error approach based on the specific application requirements. Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9 show the membership functions for the SINR, CBT, PDR, the number of neighbors, and the mobility.

Each vehicle calculates the CBT degree using Equation (1) and applies the CBT membership function to classify the computed value as either small, medium, or large. Similarly, the mobility factor is computed using Equations (3) and (4), with the mobility membership function then determining the degree of mobility as either low, medium, or high. Furthermore, each vehicle utilizes the PDR membership function to determine the degree of packet delivery ratio (PDR) as either small, medium, or large. Likewise, the number of neighbors’ membership function is employed to determine the degree of the number of neighbors as either small, medium, or big. Subsequently, the SINR factor is computed using Equations (5) and (6), and the SINR membership function is applied to establish the degree to which it belongs, categorized as either small, medium, or large. Prior to utilizing any of the aforementioned metrics, they are normalized using the min-max normalization method.

5.2.2. Rules and Inference Engine

For simplicity, we utilized the fuzzy inference method of Mamdani’s min-max, since it facilitates the process of formulating the rule base. We designed the knowledge base rules according to the understanding of VANETs’ characteristics as well as utilizing a weight assignment technique for every input parameter. In this method, the fuzzy operator AND selects the minimum value of the antecedents. When adding multiple rules, the maximum value of the consequent is utilized as the fuzzy output. The fuzzy rules for Set 1and Set 2’s parameters are shown in

Table 2. Set 1’s fuzzy rules.

Rule Number	SINR	Number of Neighbors	Mobility	Rank
1	Small	Small	High	Very Small
2	Small	Small	Medium	Very Small
3	Small	Small	Low	Small
4	Small	Medium	High	Very Small
5	Small	Medium	Medium	Small
6	Small	Medium	Low	Small
7	Small	Big	High	Small
8	Small	Big	Medium	Medium
9	Small	Big	Low	Medium
10	Medium	Small	High	Small
11	Medium	Small	Medium	Medium
12	Medium	Small	Low	Medium
13	Medium	Medium	High	Medium
14	Medium	Medium	Medium	Medium
15	Medium	Medium	Low	Big
16	Medium	Big	High	Medium
17	Medium	Big	Medium	Big
18	Medium	Big	Low	Big
19	Large	Small	High	Medium
20	Large	Small	Medium	Big
21	Large	Small	Low	Big
22	Large	Medium	High	Big
23	Large	Medium	Medium	Very Big
24	Large	Medium	Low	Very Big
25	Large	Big	High	Very Big
26	Large	Big	Medium	Very Big
27	Large	Big	Low	Very Big

and 

Table 3. Set 2’s fuzzy rules.

Rule Number	CBT	Mobility	PDR	Rank
1	Small	High	Large	Very Small
2	Small	High	Medium	Very Small
3	Small	High	Small	Small
4	Small	Medium	Large	Very Small
5	Small	Medium	Medium	Small
6	Small	Medium	Small	Small
7	Small	Low	Large	Small
8	Small	Low	Medium	Medium
9	Small	Low	Small	Medium
10	Medium	High	Large	Small
11	Medium	High	Medium	Medium
12	Medium	High	Small	Medium
13	Medium	Medium	Large	Medium
14	Medium	Medium	Medium	Medium
15	Medium	Medium	Small	Big
16	Medium	Low	Large	Medium
17	Medium	Low	Medium	Big
18	Medium	Low	Small	Big
19	Large	High	Large	Medium
20	Large	High	Medium	Big
21	Large	High	Small	Big
22	Large	Medium	Large	Big
23	Large	Medium	Medium	Very Big
24	Large	Medium	Small	Very Big
25	Large	Low	Large	Very Big
26	Large	Low	Medium	Very Big
27	Large	Low	Small	Very Big

, respectively. Set 3 has two levels of fuzzy rules. The first level is shown in

Table 4. Set 3’s fuzzy rules (1).

Rule Number	CBT	SINR	PDR	Preliminary Congestion Rank
1	Small	Small	Large	Small
2	Small	Small	Medium	Small
3	Small	Small	Small	Small
4	Small	Medium	Large	Small
5	Small	Medium	Medium	Small
6	Small	Medium	Small	Small
7	Small	Large	Large	Small
8	Small	Large	Medium	Medium
9	Small	Large	Small	Medium
10	Medium	Small	Large	Small
11	Medium	Small	Medium	Medium
12	Medium	Small	Small	Medium
13	Medium	Medium	Large	Medium
14	Medium	Medium	Medium	Medium
15	Medium	Medium	Small	Big
16	Medium	Large	Large	Medium
17	Medium	Large	Medium	Big
18	Medium	Large	Small	Big
19	Large	Small	Large	Medium
20	Large	Small	Medium	Big
21	Large	Small	Small	Big
22	Large	Medium	Large	Big
23	Large	Medium	Medium	Big
24	Large	Medium	Small	Big
25	Large	Large	Large	Big
26	Large	Large	Medium	Big
27	Large	Large	Small	Big

and the second level in

Table 5. Set 3’s fuzzy rules (2).

Rule Number	Preliminary Congestion Rank	Number of Neighbors	Mobility	Final Congestion Rank
1	Small	Small	High	Very Small
2	Small	Small	Medium	Very Small
3	Small	Small	Low	Small
4	Small	Medium	High	Very Small
5	Small	Medium	Medium	Small
6	Small	Medium	Low	Small
7	Small	Big	High	Small
8	Small	Big	Medium	Medium
9	Small	Big	Low	Medium
10	Medium	Small	High	Small
11	Medium	Small	Medium	Medium
12	Medium	Small	Low	Medium
13	Medium	Medium	High	Medium
14	Medium	Medium	Medium	Medium
15	Medium	Medium	Low	Big
16	Medium	Big	High	Medium
17	Medium	Big	Medium	Big
18	Medium	Big	Low	Big
19	Big	Small	High	Medium
20	Big	Small	Medium	Big
21	Big	Small	Low	Big
22	Big	Medium	High	Big
23	Big	Medium	Medium	Very Big
24	Big	Medium	Low	Very Big
25	Big	Big	High	Very Big
26	Big	Big	Medium	Very Big
27	Big	Big	Low	Very Big

. In this fuzzy rule set, the output of the first level (preliminary congestion rank) is used as an input in the second level.

Depending on the fuzzy values and utilizing IF/THEN rules, the final network congestion rank (output) can be determined. The final network congestion rank is either Very Small, Small, Medium, Big, or Very Big.

5.2.3. Defuzzification

The defuzzification process uses the center of gravity (COG), a widely adopted method in real applications. The output membership function, demonstrated in Figure 10, represents the preliminary congestion rank which is related to Fuzzy Rule Set 3 only. The output membership function, shown in Figure 11, represents the final congestion rank for all fuzzy rule sets. High congestion ranks require lower beacon rates.

Rate Control Mechanism

Once the fuzzy logic system provides the congestion rank, the rate control function in our protocol determines the next beacon rate ($NBR$) by adjusting the current beacon rate ($CBR$) through the use of Equation (8).

$$NBR=\left\{\begin{array}{cc}CBR+2\ast stepsize,\hfill & VeryLow\hfill \\ CBR+stepsize,\hfill & Low\hfill \\ CBR,\hfill & Medium\hfill \\ CBR-stepsize,\hfill & High\hfill \\ CBR-2\ast stepsize,\hfill & VeryHigh\hfill \end{array}\right.$$

(8)

The final step is to send the beacon at the new rate ($NBR$). Algorithm 1 outlines the procedure of the HACEAB protocol.

Algorithm 1 HACEAB protocol.

1: Use the default beacon rate 2: Receive beacons from other vehicles 3: Determine the number of direct neighbors $ND$ 4: Calculate the density threshold $DT$ = (1000/(CCHI + SCHI))∗(CCHI-GI)∗BL∗(1/BR) 5: if $ND$ > $DT$ then 6:     Use Set 3’s parameters 7: else Initialize $n_m=0$ for $m=[0,N]$ 8:     Generate random point x,y within the nodes’ transmission range 9:     Set $m=0$; 10:     if The node itself or neighboring node is within R/5 of x,y then 11:         Increment m 12:     end if 13:     Increment $n_m$ 14:     Repeat steps 9–14 30 times 15:     Calculate the quadrat value $Q=V\left[n\right]/E\left[n\right]$ 16:     if Q > 1 then 17:         Use Set 1’s parameters 18:     else Use Set 2’s parameters 19:         Fuzzify the crisp input 20:         Trigger the inference engine and use the selected set fuzzy rules 21:         Defuzzify and get the final congestion rank 22:         Calculate the next beacon rate (NBR) 23:         Send a beacon at the new rate 24:     end if 25: end if

5.3. Transmission Power Adaptation

HACEAB further enhances performance by adjusting the transmission power according to the network density and node distribution. The network density status and node distribution are very important aspects to look at when a protocol needs to adjust its transmission power. Using the density estimation function and the quadrat method explained earlier, HACEAB adapts the transmission power. After getting the density status (dense or not dense) from the density estimation function, if the density status is dense, the new transmission power ($NTP$) is equal to the current transmission power ($CTP$) decreased by one step size, because in that case, we need to exclude more nodes to reduce the congestion. In case of nondense conditions, HACEAB determines the spatial distribution of vehicles using the quadrat method. If the quadrat value is more than 1, it means that the vehicles are clustered in a 1D-oriented uniform distribution. In that case, $NTP$ will be equal to $CTP$ increased by one step size, since the spaces between clusters are large. In case the quadrat value is near one, the vehicles are uniformly randomly distributed in 2D. In that situation, $NTP$ is equal to $CTP$. Otherwise, $NTP$ is equal to $CTP$ decreased by one step size, since it means that the nodes are more evenly spaced. Algorithm 2 shows the steps of adapting the transmission power in HACEAB.

Algorithm 2 Transmission power method.

1: if $ND$ > $DT$ then 2:     $NTP=CTP-$ step size 3: else 4:     if Q > 1 then 5:         $NTP=CTP$ + step size 6:     else 7:         if Q ≈ 1 then 8:            $NTP=CTP$ 9:         else 10:            $NTP=CTP-$ step size 11:         end if 12:     end if 13: end if

6. Simulation and Results

To evaluate the performance of HACEAB, an NS-3 network simulator was utilized, which is based on C++ [62]. The simulation parameters are shown in

Table 6. The simulation parameters.

Parameter	Value
Number of vehicles	10, 50, 100, 150, 200
Packet size	400 bytes
Data rate	6 Mbps
MAC/PHY protocol	IEEE 802.11p
Signal propagation	Nakagami
Reception threshold ($R{x}_{Th}$)	−91 dBm
CCA threshold	−95 dBm
Simulation time	300 s

. We carried out a comprehensive evaluation of HACEAB including covering both highway and urban environments to ensure the robustness and adaptability of HACEAB in various scenarios.

First, the performance of our protocol was compared against the schemes which were based only on Set 1, Set 2, or Set 3’s parameters. Then, the performance of HACEAB was compared with other protocols which were a fixed rate of 10 beacons per second (a requirement of many safety applications [5]), JSRC [19], and BRAIN-F [20]. Finally, it was shown how adaptive transmission power could improve the protocol performance. The performance was evaluated based on three metrics as follows:

• AverageWaiting Time (AWT): it is the duration during which CCH is noted as being occupied due to transmissions from nearby vehicles.
• Total Delay: It is the duration it takes for a packet to be delivered from the source to the destination. This metric indicates the freshness of the information.
• AverageNumber of Neighbors: It is the average number of discovered neighbors within the transmission range. This metric reflects the cooperative/neighborhood awareness.

6.1. Performance of HACEAB

We evaluated the performance (only the beaconing frequency adaptation with no transmission power adaptation) of HACEAB and compared it with protocols that used only Set 1, Set 2, or Set 3’s parameters to confirm that it could intelligently combine their strength in nondense and dense conditions in both clustered and uniformly distributed environments.

6.1.1. Highway Environment

The vehicles’ mobility was created based on a constant speed mobility model in NS-3. The position allocation in our simulation was based on the NS-3 random rectangle position model [63]. In this model, vehicles are distributed uniformly on a straight line (highway road scenario). We conducted simulations utilizing the WAVE model [64]. The WAVE model constitutes the comprehensive system architecture constructed in NS-3 for vehicular communications.

Figure 12 shows that HACEAB maintains a low average waiting time compared to Set 2 and Set 3’s protocols for 10, 50, 100, and 150 vehicles (not dense), and less than Set 1 and Set 2’s protocols for 200 vehicles (dense). This is because HACEAB uses the parameters of Set 1 in nondense clustered environment, while it uses the parameters of Set 3 in a dense highway environment (Table 1). Similarly, from Figure 13, we can notice that HACEAB has a smaller delay than Set 2 and Set 3’s protocols for 10, 50, 100, and 150 vehicles, while it has a smaller delay than Set 1 and Set 2’s protocols for 200 vehicles. The main aim of an adaptive beaconing scheme is to enhance cooperative awareness and maintain information freshness without inducing network congestion. Figure 14 illustrates that HACEAB achieves a closer level of awareness compared to Set 1 and Set 2’s protocols for 10, 50, 100, and 150 vehicles, and is closer to Set 2 and Set 3’s protocols for 200 vehicles.

6.1.2. Urban Environment

In order to obtain results for an urban environment, we used the network simulator, NS-3. We employed a Manhattan grid configuration of dimensions 3 × 3, featuring an edge length of 1000 m, with a spacing of 500 m between any two intersections. To obtain a realistic simulation, Simulation of Urban MObility (SUMO) [65] was used to generate the vehicles’ movement. At the onset of each simulation, vehicles were dispersed randomly across the road network. As the simulation advanced, vehicles started moving in accordance with the car-following model, where a vehicle’s speed is influenced by the speed of the vehicle ahead. Random routes were generated using the randomTrips utility within the SUMO framework. Subsequently, the resulting mobility traces were imported into NS-3 to simulate node mobility.

Figure 15 shows that HACEAB has a smaller average waiting time than Set 2 and Set 3’s protocols for 10 and 50 vehicles, and a smaller one than Set 1 and Set 3’s protocols for 100 and 150 vehicles. For 200 vehicles, HACEAB has less waiting time than Set 1 and Set 2’s protocols. From Figure 16, it can be observed that HACEAB performs closer to Set 1 and Set 2’s protocols for 50 vehicles and better than Set 1 and Set 3’s protocols for 100 and 150 vehicles (nondense urban) in terms of delay. For 200 vehicles (dense urban), HACEAB performs better than Set 1 and Set 2’s protocols, since it uses the parameters of Set 3 (Table 1) as the local density ($ND$) is larger than the density threshold ($DT$). Similarly, in Figure 17, we can see that HACEAB detects a higher number of neighbors than Set 3’s protocol for 10, 50, 100, and 150 vehicles, and a higher number than Set 1’s protocol for 200 vehicles.

6.2. Performance of HACEAB Compared to Other Protocols

In the previous section, the performance of HACEAB was shown to combine the strength of the protocols that used only Set 1, Set 2, or Set 3 in both dense and nondense highway and urban environments. Here, HACEAB, functioning as an adaptive beaconing rate protocol, is compared with a fixed rate of 10 beacons per second. Additionally, it is compared with two other adaptive beaconing rate protocols, namely JSRC and BRAIN-F, within both highway and urban environments.

6.2.1. Highway Environment

From Figure 18, we can clearly see that the average waiting time of the fixed rate of 10 beacons per second is the highest. This means fixed-rate beaconing is inefficient in terms of channel usage, since the beaconing rate is always the same regardless of the channel status. JSRC behaves closer to the fixed-rate protocol for low density and improves as density increases, since it reduces the beaconing rate as the number of vehicles increases. As for BRAIN-F, it uses the channel more efficiently than the fixed rate and JSRC for 10, 50, 100, and 150 vehicles. However, for 200 vehicles, JSRC performs better. This figure shows that the average waiting time for HACEAB is smaller than that for all other protocols for all densities. This is because HACEAB monitors the channel carefully and adapts intelligently according to its status. Figure 19 shows that HACEAB maintains a smaller delay than the other protocols, while the other protocols have almost the same performance pattern for channel waiting time. In Figure 20, we can see clearly that HACEAB achieves a much higher level of awareness than the other protocols, and it improves as the density increases (more scalable than the other protocols).

6.2.2. Urban Environment

Figure 21 proves that HACEAB maintains a smaller waiting time than the other protocols for all densities. This means that HACEAB uses the channel more efficiently than the other protocols. Figure 22 illustrates that a fixed beaconing rate causes a high delay that increases exponentially due to a high processing delay which increases as the number of packets becomes larger. The delay for HACEAB, shown in Figure 22, is closer to that of BRAIN-F for 10 and 50 vehicles and lower than that of all other protocols for 100, 150, and 200 vehicles. Figure 23 shows that the cooperative awareness of the fixed beaconing rate becomes relatively worse as the number of vehicles increases. The reason behind that is the increase in the number of collisions, which make other vehicles unable to receive packets. In this figure, we can see that HACEAB achieves a comparable level of cooperative awareness to BRAIN-F for 10 and 50 vehicles, with a slight improvement for 100 and 150 vehicles, and a noticeable improvement for 200 vehicles.

HACEAB exhibits superior performance compared to other protocols due to its careful consideration of crucial factors, including channel conditions, traffic situations, and link quality. This capability allows HACEAB to dynamically adjust its beaconing rate, effectively addressing essential aspects for efficient operation. Moreover, HACEAB decides to adapt its beaconing frequency and uses the most suitable factors based on different scenarios in a novel way.

6.3. Performance of HACEAB with Transmission Power Adaptation

Here, HACEAB’s performance with power adaptation (referred to as power on) is evaluated and compared to its performance without power adaptation (referred to as power off).

6.3.1. Highway Environment

Figure 24 shows that the average waiting time of HACEAB when power adaptation is on is better than when power adaptation is off for all densities. This is because HACEAB decreases the transmission power in dense conditions (when the local density is larger than the density threshold) or whenever it detects that the vehicle distribution is more evenly spaced (Q < 1). Reducing transmission power in these conditions causes more vehicles to be excluded, thus alleviating the load on the channel. In Figure 25, we notice that the packet delay, when the transmission power adaptation is on, is slightly higher than when it is off. This is expected since HACEAB increases the transmission power according to the vehicle distribution, which causes more vehicles to be covered. Consequently, the number of packets that are processed locally increases. A smart protocol is the one that increases the transmission power to improve cooperative awareness when the channel is freer to accommodate more vehicles and vice versa; this is what HACEAB does. In Figure 26, we can obviously see that goal has been accomplished. Transmission power adaptation in HACEAB increases cooperative awareness.

6.3.2. Urban Environment

Figure 27 illustrates the efficient use of the channel by HACEAB when the transmission power adaptation is on compared to when it is off, since HACEAB determines the node distribution and the network density status before it adjusts the transmission power. In Figure 28, the packet delay when the transmission power adaptation is on is almost the same as when it is off for 10 and 50 vehicles and better for 100 and 150 vehicles. Yet, for 200 vehicles, when the transmission power adaptation is off, the protocol maintains a smaller delay than when it is on, since more nodes are covered when HACEAB increases the transmission power. Figure 29 shows that HACEAB achieves a higher level of cooperative awareness when transmission power adaptation is on than when it is off.

7. Conclusions

A beacon is a periodic message that is sent to all one-hop neighbors and contains basic information about the sending vehicle, which is needed to support safety and nonsafety applications. The rate at which this information is updated has to be high enough to maintain its freshness without unnecessarily congesting the network. This motivates the need for adaptive beaconing to cope with mobility and network density variations in VANETs.

Our preliminary work led us to establish, through extensive simulation experiments, that specific input parameters played a more crucial role in adapting beacon rates to certain environments and conditions. Building upon that insight, we introduced HACEAB, an innovative adaptive beaconing scheme based on fuzzy logic, designed for heightened awareness and channel efficiency. This scheme was tailored to adapt to various environments and conditions. Initially, HACEAB estimates network density using an adaptive threshold function to discern the prevailing network condition, distinguishing it as either dense or not. Afterwards, the protocol assesses the spatial distribution of nodes through the quadrat statistical method to classify the network environment as either clustered or uniformly distributed. Subsequently, it adjusts the beaconing rate by utilizing the relevant fuzzy logic system input parameters based on the determined network condition and environment. Additionally, it enhances performance by adapting the transmission power to accommodate changes in node density and distribution.

Our simulation results confirmed the ability of HACEAB to intelligently adapt to different densities and spatial distribution of nodes. We compared HACEAB’s performance to that of fixed-rate beaconing, JSRC, and BRAIN-F. HACEAB achieved remarkable improvements compared to the fixed-rate beaconing in terms of neighborhood awareness, channel usage, and packet delay for both highway and urban environments. Moreover, HACEAB outperformed BRAIN-F across all density levels, while substantially outperforming JSRC for lower density levels in terms of channel usage and packet delay. As for neighborhood awareness, HACEAB detected a higher number of neighbors than JSRC and BRAIN-F across all density levels in both highway and urban environments. The results also showed that when transmission power adaptation was used, further improvement in neighborhood awareness and channel usage was achieved in both urban and highway environments. In the future, and as many applications in vehicular networks depend heavily on beaconing, we plan to investigate the impact adaptive beaconing may have on some of these applications.