FL-ASB: A Fuzzy Logic Based Adaptive-period Single-hop Broadcast Protocol

Abstract

In vehicular ad hoc networks, vehicle-to-vehicle–based broadcast can fast disseminate safety messages between vehicles within the whole network and hence expand drivers perception vision, which will reduce the accident probability and ensure the transportation reliability. As for fixed-period single-hop broadcast protocol, disseminating safety messages frequently can cause excessive network load. However, increasing period purely does not guarantee the real-time performance. In addition, exiting adaptive-period single-hop broadcast protocols also have limitations without considering synthetically various impact factors. Thus, how to design a single-hop broadcast protocol that can dynamically adjust the broadcast period according to the actual road condition is a pressing issue. A Fuzzy Logic Based Adaptive-period Single-hop Broadcast Protocol in vehicular ad hoc networks is designed in this article, which provides a new solution for the dissemination of period safety messages. In this article, the impact of various factors (such as the number of one-hop neighbor nodes, vehicle speed, received signal strength index, and visibility) on the single-hop broadcast period has been analyzed. In view of each impact factor, we design corresponding membership function and fuzzy rules according to the specific scenarios and parameters. It realizes the adaptive changes of period safety messages broadcast period through the simulation of the proposed fuzzy logic inference system. Finally, we verify the performance of the Fuzzy Logic Based Adaptive-period Single-hop Broadcast Protocol in a bidirectional four-lane highway scenario. Simulation results show that the proposed Fuzzy Logic Based Adaptive-period Single-hop Broadcast Protocol has obvious advantages in terms of network load ratio, average one-hop delay, and delivery ratio.

Keywords

Single-hop broadcast protocol fuzzy logic routing vehicular ad hoc networks

Introduction

Vehicular ad hoc networks (VANETs) have been attracted much attention in both industry and academia recently.^1,2 One of the important applications of VANETs is to transmit safety messages between vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I), so as to reduce traffic accidents, avoid traffic jams, and save fuel consumption.^3–5 Different from the unicast messages (point-to-point messages, for example, the data packets transmitted over the Internet) transmitted in a conventional network, the traffic safety messages are transmitted by the way of broadcast. It is the reason that the traffic messages are related to public advantages and most vehicles on the road may be interested in that. Thus, compared with the unicast routing protocol, the broadcast routing protocol is more suitable to undertake the task of safety messages dissemination in VANETs.⁴

Safety messages are divided into two categories in VANETs.^6,7 One is periodic safety message (PSM), including the speed, position, and direction of the vehicle. In the form of Beacon, PSMs are disseminated periodically by means of single-hop broadcast. Therefore, vehicles on the road can acquaint with the real-time status of neighbor vehicles to prevent accidents. The other is event-driven emergency warning message (EWM), which is the real-time emergency alarming messages resulting from traffic accident. The EWMs could to be disseminated in VANETs immediately. Due to the limitation of the communication range of a vehicle, the EWMs should be disseminated by means of multi-hop broadcast. The EWMs can be forwarded by the relay nodes to the target vehicles in a further distance, thus avoiding a chain-collision-accident. For the above reasons, the safety message dissemination mechanisms in VANETs can be divided into single-hop broadcast and multi-hop broadcast. Thereinto, according to beaconing category, the single-hop broadcast protocol can be divided into message frequency control (MFC),^8,9 transmit power control (TPC),^10,11 miscellaneous,^12,13 and hybrid^14,15 situations.⁷ This article takes the MFC into account. With MFC, the single-hop broadcast protocol can be divided into two categories: fixed-period single-hop broadcast protocol^16,17 and adaptive-period single-hop broadcast protocol.^18–21

As for fixed-period single-hop broadcast protocol, Zhong et al.¹⁶ proposed the TrafficInfo protocol that selects and broadcasts the most relevant messages rather than all the messages to reduce the information redundancy. Nakorn et al.¹⁷ proposed a single-hop broadcast protocol based on Bloom filter. Based on the characteristics of the Bloom filter in data storage to reduce the size of the message, their proposed protocol can reduce the network load greatly. Due to the fact that the fixed-period single-hop broadcast protocol can broadcast safety messages frequently, when the vehicle density is too large, it will occupy too much channel resource and make the network overloaded. Therefore, researchers have proposed the adaptive-period single-hop broadcast protocol which could adjust the broadcast period adaptively. It can dynamically adjust the frequency of sending safety messages and has the advantage of lower network load. Based on two key parameters, that is, message availability and channel quality, Sommer et al.¹⁸ proposed the adaptive traffic beacon (ATB) protocol to adjust broadcast period. The successive rate and power adaptation (SuRPA) protocol proposed by Zemouri et al.¹⁹ adopts a joint adaptive approach combining transmission rate with transmission power to solve the problem of channel congestion. Chaabouni et al.²⁰ proposed the collision-based beacon rate adaptation (CBA) protocol, which takes congestion time as a parameter to adjust the frequency of message generation. Hassan et al.²¹ carried out a fuzzy processing of the data packet carried time, the number of single-hop neighbor nodes, and the vehicle speed according to fuzzy logic theory. The performance of the single-hop broadcast protocols with corresponding adaptive period is better than those with fixed period.

Theoretically, the period of one-hop should be as short as possible to ensure the real-time characteristic of messages. However, the PSMs can share the channel resource when they are broadcasted. Therefore, too high frequency to broadcast messages could result in channel overloading and cause channel congestion. It is a tradeoff between real-time characteristic of messages and channel occupancy. Traditional fixed-period single-hop broadcast protocols have poor scalability. The network load will increase sharply with the traffic density, because they only considers the real-time characteristics of PSMs. Most adaptive-period one-hop broadcast protocols usually do not consider the restrictive relation between various factors, including channel state, weather condition, traffic density, and vehicle speed. It is complicated to establish a mathematical model which can think about all kinds of factors. However, the fuzzy logic inference method can consider the restrictive relation among various factors, without developing the complicated mathematical model. Therefore, in this article, in view of fuzzy logic inference method, we design an adaptive-period single-hop broadcast protocol, named after Fuzzy Logic Based Adaptive-period Single-hop Broadcast Protocol (FL-ASB). We verify the effectiveness of FL-ASB in network load ratio, average one-hop delay, and delivery ratio through simulations, compared with the fixed-period single-hop broadcast protocols and the adaptive-period single-hop broadcast protocol (ASB²¹).

The rest of this article is organized as follows. Section “Fuzzy logic theory” briefly introduces the concepts, principles, and inference process of fuzzy logic. Section “Adaptive-period method based on fuzzy logic” synthetically considers several important impact factors, designs the corresponding membership function and fuzzy rules set, and realizes the adaptive changes of PSMs broadcast period by simulating the fuzzy logic inference system. Section “Simulation results and discussions” presents the simulation results to verify the performance improvement of the FL-ASB protocol in network load ratio, delivery ratio, and average one-hop delay. Finally, conclusions are drawn in section “Conclusion.”

Fuzzy logic theory

The traditional mathematical logic is a classical two-value logic, that is, “true” or “false,” usually expressed as “1” or “0.” However, in the fuzzy logic, a proposition could be “partial true” or “partial false.” The degree that a proposition belongs to “true” or “false” can be measured by degree of membership. On account of the transition states between two values expressed by degree of membership, researchers can use the accurate mathematical model to describe fuzzy problems.²²

Fuzzy set and membership function

The two-value logic is usually used for classical sets. Given any set $A$ , any element $x$ in the domain either does or does not belong to $A$ . In fuzzy logic, we consider a kind of fuzzy set different from the classical set. We define an ordinary set as $U$ . Any mapping $f$ from $U$ to [0, 1] interval can determine a fuzzy subset of $U$ , named after fuzzy set $A$ of $U$ , and the mapping $f$ is called the membership function of fuzzy set $A$ . For any element $x$ of $U$ , $f (x)$ is called the degree of membership of $x$ for fuzzy set and can also be expressed by $A (x)$ . The degree of membership represents the degree that the element $x$ belongs to fuzzy set $A$ . Therefore, larger the degree of membership is, higher the degree that the element $x$ belongs to $A$ is.

The fuzzy set $A$ can be described by the membership function. For discrete domain $U = {x_{1}, x_{2}, \dots, x_{n}}$ , the representation of fuzzy set $A$ is expressed as follows:

Zadeh representation

A = \frac{f_{A} (x_{1})}{x_{1}} + \frac{f_{A} (x_{2})}{x_{2}} + \dots + \frac{f_{A} (x_{n})}{x_{n}}

(1)

Ordered couple representation

A = {(x_{1}, f_{A} (x_{1})), (x_{2}, f_{A} (x_{2})), \dots, (x_{n}, f_{A} (x_{n}))}

(2)

Vector representation

A = [f_{A} (x_{1}), f_{A} (x_{2}), \dots, f_{A} (x_{n})]

(3)

For continuous domain $U$ , the fuzzy set $A$ can be expressed by Zadeh method as

A = \int_{U} \frac{f_{A} (x)}{x}

(4)

In general, for different problems, we should choose different kinds of membership functions, according to the analysis of experts in related field. The frequently used membership function is triangular membership function, trapezoidal membership function, and sigmoid membership function.

Fuzzy inference

Fuzzy inference means a process which derive a new fuzzy proposition through the known fuzzy proposition and exiting definite rules. The known fuzzy propositions are called premises, the rules used by inference are called fuzzy rules, and the fuzzy propositions inferred are called conclusions.

Fuzzy inference is a kind of approximate inference, which is the process that corresponds the input fuzzy set to a specific output fuzzy set according to those certain fuzzy rules. Fuzzy rules play an important role in the process of fuzzy inference, and they are described in natural language based on experts’ experience, which expression form is usually “If $x$ is $A$ , then $y$ is $B$ .” Thus, it is also called the “IF-THEN” rule.

The fuzzy inference process usually includes four parts: fuzzification, fuzzy rules base, inference method, and defuzzification, as shown in Figure 1:

Fuzzification. The process of converting specific input values into degree of membership of fuzzy sets via membership functions.

Fuzzy rules base. The “brain” of fuzzy logic inference, which stores all “IF-THEN” rules that fuzzy inference required. These rules are usually described by experts using natural language based on their experience.

Inference method. The method of obtaining the final fuzzy conclusion, according to the degree of membership of input variables to fuzzy sets and the detailed fuzzy rules. There are many ways about fuzzy inference. The conclusions may be different with different methods. Among the methods, the Mamdani inference method based on Max–Min compositional rule is usually used.

Defuzzification. The process of converting the fuzzy conclusions into detailed output values. The most common method is center of gravity (CoG) method.

Figure 1.

The four parts of fuzzy logic inference.

Adaptive-period method based on fuzzy logic

In this section, we will follow the steps of fuzzy logic inference and introduce how to use fuzzy logic inference method to obtain the adaptive single-hop broadcast period, considering several factors, such as traffic density, vehicle speed, channel state, and objective weather condition.

Scenario and problem description

At present, the main scenarios of message dissemination mechanism in VANETs can be divided into highway, urban expressway, and urban intersection. Compared with intersections and expressways in urban scenario, it is easier to analyze vehicle movement patterns in highway scenario. The lane shows a one-dimensional straight line shape in highway, where the vehicles move in only one lane. Therefore, in order to reduce complexity, this article focuses on the scenario of bidirectional four-lane highway with design speed of 120 km/h. As shown in Figure 2, the vehicles in top two lanes are eastbound, and the vehicles in bottom two lanes are westbound. In addition, we ignore the emergency parking lane and ramp.

Figure 2.

Bidirectional four-lane highway scenario of design speed of 120 km/h.

Theoretically, larger the single-hop broadcast period is, smaller the broadcast frequency of safety messages is, smaller the updating frequency of safety messages is, worse the real-time performance is, but smaller the channel load is. In fact, the single-hop broadcast period can be affected by many factors, such as traffic density, vehicle speed. For example, when the traffic density is large, if the broadcast period is short, a large number of vehicles will broadcast safety messages frequently, resulting in the excessive channel load and congestion. Therefore, when the traffic density is large, the broadcast period should be extended, so as to decrease the broadcast frequency and reduce the channel congestion. Conversely, we allow a short broadcast period. Considering the impact factors of the single-hop broadcast period, Hassan et al.²¹ focused on the number of neighbor nodes in one-hop, vehicle speed, and packet carried time and derived the adaptive one-hop broadcast period through fuzzy logic method. However, it has several limits below.

Statistical method for the number of neighbor nodes in one-hop

Hassan et al.²¹ calculated the number of one-hop neighbor nodes of target vehicle to represent the current local traffic density, which is simple and effective. However, there are some limitations that it only considers the vehicles which are in front of the target vehicle in the same moving direction. Therefore, the traffic density is smaller without taking into account the vehicles behind the target vehicle. It could affect the design of the one-hop broadcast period.

Necessity of packet carried time

Hassan et al.²¹ used the packet carried time to represent the time that a packet is carried by some vehicles before it is forwarded. However, in the single-hop broadcast scenario of this article, each vehicle periodically transmits and receives PSMs from its neighbor vehicles, without using carry-forward mechanism. Therefore, this article does not consider the affect of packet carried time on the one-hop broadcast period.

Comprehensiveness of impact factor

Hassan et al.²¹ only took into account three factors which effect the one-hop broadcast period, which are incomplete. There are other significant factors to consider. First, the state of the wireless channel has a great impact on the quality of communication. In VANETs, the large vehicle speed could have serious impact on the state of wireless channel. Hence, it is necessary to analyze the impact of wireless channel state. The better wireless channel state is, the higher frequency of messages dissemination that can be allowed is. Second, the weather factor could have a impact on the drivers behavior and judgment in the actual driving environment. The visibility will directly affect the drivers visual range and indirectly affect the drivers driving behavior (drivers are more likely to travel at a low speed when the visibility is poor). Thus, the frequency of safety messages dissemination could be different with visibility. When the visibility is poor, the broadcast frequency of safety messages should be increased to help drivers obtain real-time traffic information.

Selection of impact factor

From the analysis above, we can find that there are some limitations in the adaptive adjustment of single-hop broadcast period. In order to solve them, we will think over the following factors comprehensively and make a more reasonable adaptive adjustment for the single-hop broadcast period.^23–25

Number of one-hop neighbor nodes N (veh)

In this article, the calculation of the number of neighbor nodes in one-hop range is based on the following two principles, as shown in Figure 3. First, the communication radius $R$ of vehicle A can cover vehicles in all lanes, assuming that the communication radius $R$ is greater than the total lanes width. We assume that only the vehicles which are in the same direction as the vehicle A will receive the safety message. The vehicles which are in the opposite direction of A have no affect on A due to the limits of highway lane. Thus, the vehicles in the opposite direction do not need to receive the safety messages from vehicle A. Second, for the vehicles which move in the same direction as A, we should calculate all vehicles that are located in front of and behind vehicle A within the one-hop communication range. As shown in Figure 3, vehicles C, D, E, and F can be regarded as the one-hop neighbor nodes. Therefore, $N (A) = 4$ .

Figure 3.

Neighbor node numbers in one hop.

Vehicle speed V (m/s)

Vehicle speed is one of the important information of vehicle operating status. It can indirectly reflect the traffic density and the changing frequency of vehicle status. According to traffic flow theory, when the traffic density is too large, the vehicles tend to have a low speed. The vehicles move almost smoothly, and there is no need to update PSMs frequently. However, when the traffic density is small, the vehicles tend to have a high speed. The status of vehicles will change violently in a short time. Therefore, it is necessary to increase the frequency of updating PSMs, so that the vehicles can obtain the information of neighbor vehicles timely. In summary, the speed of vehicles must be also considered.

Received signal strength index

In wireless communication environment, the high speed of vehicles will further aggravate the Doppler effect and multipath effect, having a great impact on the wireless channel. Thus, it is very important to consider the wireless channel when we calculate the broadcast frequency of the PSMs. The broadcast frequency of safety messages could be increased in a good wireless channel state. While the wireless channel state is bad, we should decrease the broadcast frequency of safety messages to reduce the channel load and avoid the channel congestion. In this article, we adopt received signal strength index ( $R S S I$ ) to measure the quality of the wireless channel. When vehicle $Y$ receives the messages from its neighbor vehicle $X$ , the current $R S S I$ can be calculated by the following formula

\begin{matrix} R S S I (Y) = (1 - α) \times R S S I (X) \\ + α \times (1 - R x_T h r / R x_P r) \end{matrix}

(5)

where $α$ is a smoothing coefficient, belonging to [0, 1]. In this article, the value of $α$ is 0.7.²⁵ $R x_P r$ is the received signal power and $R x_T h r$ is the threshold of the received signal power. The signal whose power is less than the threshold cannot be received. $R S S I (X)$ is the $R S S I$ of its neighbor vehicle $X$ , and its initial value is 0. Thus, when $Y = X$ , we can obtain

R S S I (X) = α \times (1 - R x_T h r / R x_P r)

(6)

which means that in the initial conditions (there is only one vehicle $X$ ), vehicle $X$ will receive the signal sent by itself and calculate its own $R S S I$ .

Visibility D (km)

Different from traditional wireless communication scenarios, in the actual environment of road traffic, weather is also an objective factor to consider, in addition to several factors that can directly affect the wireless communication. In some bad weather conditions, such as rain, snow, fog, it can cause the road slippery and hard to drive on it, which will even lead to traffic accidents. And it can also make the drivers have a poor visibility to bring dangerous consequences. In VANETs, the PSMs can expand drivers visual field and reduce the impact of poor visibility on drivers. Thus, in this article, we consider the objective environment factor, that is, visibility, as a significant factor to affect the safety messages broadcast period and quantify the visibility for analysis. Visibility is the maximum distance that a person with normal vision can see the outline of a target in some weather condition. In this article, visibility is denoted by $D$ (km). When the visibility is poor, it is necessary to increase the broadcast frequency of safety messages to help drivers acquaint the situation around them. While the visibility is good, it should decrease the broadcast frequency of safety messages to reduce the channel load.

Fuzzification

In fuzzy logic inference, fuzzification is the process of mapping each specific input variable to the degree of membership through the membership function. In this process, the determination of the membership functions of input variables is crucial. In this article, through repeated tests, we select four membership functions as below, which correspond to each input variable, respectively.

Membership function of the number of one-hop neighbor nodes N (veh)

According to the classification of highway service level in China “Technical Standard of Highway Engineering”²⁶ and traffic flow theory, we can calculate the number of vehicles per lane under different traffic densities, with the one-hop communication range $R = 300 m$ . As shown in Table 1, we can obtain the range of $N$ under different traffic densities, combined with the statistical method of $N$ mentioned above. According to the data in the table, the number of neighbor nodes $N$ is between 0 and 26. With different values of $N$ , the fuzzy sets of the number of one-hop neighbor nodes are divided into four ranks (VeryFew, Few, Many, VeryMany), as shown in Figure 4.

Table 1.

Service level classification of highway with design speed of 120 km/h.

Service level	Maximum service volume $Q_{m}$ (pcu/h/ln)	Traffic density $K$ (pcu/km/ln)	Average speed $v$ (km/h)	Number of vehicles within $R$ (veh/ln)	Number of one-hop neighbor vehicles $N$ (veh)
I	1200	(0, 14]	85.7	(0, 4]	(0, 8]
II	1650	(14, 22]	75	(4, 6]	(8, 12]
III	1980	(22, 30]	66	(6, 9]	(12, 18]
IV	2200	(30, 45]	48	(9, 13]	(18, 26]

Figure 4.

Membership function of the number of one-hop neighbor nodes.

Membership function of vehicle speed $V$ (m/s)

According to the related traffic laws in China, the speed of the vehicles should be between 60 and 120 km/h, when they are on a highway designed to run at 120 km/h. With different values of $V$ , the fuzzy sets of vehicle speed are divided into three ranks (Slow, Medium, Fast), as shown in Figure 5.

Figure 5.

Membership function of vehicle speed.

Membership function of RSSI

We normalized the value of $R S S I$ between 0 and 1 as shown in Figure 6. The received signal strength is divided into three levels (Bad, Medium, Good), corresponding to the $R S S I$ intervals ([0,0.8], [0.6,1] [0.8,1]), respectively. Closer the value of $R S S I$ is to 0, worse the channel quality is. Closer the value of $R S S I$ is to 1, better the channel quality is. Moreover, 0.6 and 0.8 are the critical points to measure the channel quality.

Figure 6.

Membership function of RSSI.

Membership function of visibility D (km)

Visibility is usually distinguished by the maximum distance where the target contour can be seen clearly. Generally, it can be divided into eight levels. In order to facilitate the analysis, the ranges of visibility below 5 km are consolidated into the category of bad visibility. Thus, we integrate eight levels into four levels (VeryGood, Good, Ordinary, Bad), as shown in Table 2. The membership function about $D$ is shown in Figure 7.

Table 2.

Classification of visibility.

Rank	Visibility $D (km)$	Visibility level	Visual clarity
I	[20, 30]	VeryGood	Very clear
II	[15, 25]	Good	Clear
III	[10, 20]	Ordinary	Ordinary
IV	[0, 15]	Bad	Not clear

Figure 7.

Membership function of visibility.

“IF-THEN” fuzzy rules set

In this article, we make the fuzzy rules according to the relationship between the four input variables ( $N$ , $V$ , $R S S I$ , $D$ ) and the output variable (broadcast period $T$ ), as shown in Table 3. Because the number of fuzzy rules is large, we only show part of rules in Table 3. The total rules can refer to Supplemental material. The formulation of the fuzzy rules follows several principles as follows:

The number of one-hop neighbor nodes reflects the traffic flow density. In view of traffic flow theory, the speed of vehicles decreases with the increase of traffic flow density. While the traffic flow density is small, the vehicles tend to be independent of each other and the speed of vehicles is large. However, it is also possible that the conservative drivers choose a low speed. Therefore, in the determination of fuzzy rules, $N$ and $V$ have the following relationship, as shown in Table 4.

Poor visibility $D$ can affect the drivers’ sight. Thus, most drivers could choose a low speed for safety. However, if the vehicle flow density is low, the aggressive drivers are likely to choose a high speed. If the vehicle flow density is large, the aggressive drivers have no choice but move with a low speed. Therefore, when the level of $N$ is “VeryFew,” $D$ could be selected from the four levels (VeryGood, Good, Ordinary, Bad) with all kinds of levels of vehicle speed. When the level of $N$ is not “VeryFew,” $D$ could be selected “Bad” only with a slow speed.

The channel state is independent of the other three input variables. Thus, regardless of the values of other variables, the $R S S I$ can be chosen from the three levels (Bad, Medium, Good).

The value of broadcast period $T$ is related to four input variables, and the selection of $T$ follows the following rules:

When $N$ is large and $V$ is small, it means that the density of vehicles is large and they move in a low speed. In addition, the vehicle statue also changes slowly. Thus, we should increase broadcast period $T$ to reduce the channel congestion and improve reliability. On the contrary, when $N$ is small and $V$ is large, we should decrease the broadcast period $T$ to increase the updating frequency of safety messages and improve the real-time performance.

When $R S S I$ is large, it means that the channel state is good. Thus, we could decrease the broadcast period $T$ . On the contrary, if $R S S I$ is small and the channel state is bad, we should increase the broadcast period $T$ to reduce the channel congestion and guarantee a good delivery ratio.

When $D$ is small, it means that the drivers have a poor visibility. Thus, we should decrease the broadcast period $T$ to increase the updating frequency of safety messages. It can help drivers expand the visual field in a bad weather and let drivers know the situation of surrounding vehicles timely.

Table 3.

Partial fuzzy rules.

Detailed rules	$N$	$V$	$R S S I$	$D$	$T$
1	VeryFew	Fast	Good	VeryGood	VeryShort
2	VeryFew	Fast	Good	Good	VeryShort
3	VeryFew	Fast	Good	Medium	VeryShort
4	VeryFew	Fast	Good	Bad	VeryShort
5	VeryFew	Fast	Medium	VeryGood	VeryShort
6	VeryFew	Fast	Medium	Good	VeryShort
7	VeryFew	Fast	Medium	Medium	VeryShort
8	VeryFew	Fast	Medium	Bad	VeryShort
9	VeryFew	Fast	Bad	VeryGood	MediumShort
10	VeryFew	Fast	Bad	Good	MediumShort
…	…	…	…	…	…

Table 4.

$N - V$ relationship table.

Number of neighbor nodes in one-hop $N$	Vehicle speed $V$ : Fast	Vehicle speed $V$ : Medium	Vehicle speed $V$ : Slow
VeryFew	$\sqrt$	$\sqrt$	$\sqrt$
Few	$\sqrt$	$\sqrt$	$\sqrt$
Many		$\sqrt$	$\sqrt$
VeryMany			$\sqrt$

Defuzzification

Defuzzification can transform the fuzzy inference results into specific numerical results. This process is realized by proper defuzzification method, based on the output membership function. There are many kinds of defuzzification methods, and the CoG method is adopted in this article. Zhang et al.²⁷ gave the detailed description about CoG method. According to the fuzzy rules, the output variable, that is, broadcast period $T$ , could be divided into several levels, as shown in Figure 8.

Figure 8.

Output membership function.

Fuzzy inference process

This article used the fuzzy toolbox in MATLAB to establish the fuzzy logic inference system. We import the membership functions of input variables, membership function of output variable, and fuzzy rules into that system. When we input four variables mentioned above to the inference system, we can obtain corresponding output value of broadcast period. According to FL-ASB protocol, the broadcast period $T$ will change adaptively with the four input variables. We cannot obtain the curve of $T$ as these input variables change simultaneously. Therefore, we observe how the broadcast period $T$ changes with one of input variables under the condition that the other three variables are constant.

T changes with different N

As shown in Figure 9, as $V = 90$ , $R S S I = 0.5$ , $D = 15.1$ , the value of the broadcast period $T$ increases with the increase of the number of one-hop neighbor nodes as a whole. When the number of one-hop neighbor vehicles increases to 16, the broadcast period can increase to about 4.5 s. When the number of one-hop neighbor vehicles $N$ increases, the traffic flow density could be increased as well. Thus, the broadcast period should become larger, in order to reduce the broadcast frequency of the safety messages and avoid the channel congestion.

Figure 9.

Graph of broadcast period changing over the number of one-hop neighbor nodes ( $N = 90$ , $R S S I = 0.5$ , $D = 15.1$ ).

T changes with different V

As shown in Figure 10, as $N = 13$ , $R S S I = 0.5$ , $D = 15.1$ , the broadcast period $T$ will decrease with the increase of vehicle speed $V$ as a whole. When the vehicle speed increases to 110 km/h, the broadcast period can decrease to 2.5 s. Larger the vehicle speed is, less the traffic flow density is. Thus, we could decrease the broadcast period without causing channel congestion. What is more, larger the vehicle speed is, faster the moving state changes. Therefore, we should increase the updating frequency of safety messages, in order to inform the surrounding vehicles of its own situation. Based on the two points above, the broadcast period should be decreased with the increase of vehicle speed to guarantee the real-time characteristic of safety messages.

Figure 10.

Graph of broadcast period changing over vehicle speed ( $N = 13$ , $R S S I = 0.5$ , $D = 15.1$ ).

T changes with different RSSI

As shown in Figure 11, as $N = 13$ , $V = 90$ , $D = 15.1$ , the broadcast period will decrease with the increase of $R S S I$ , especially in the $R S S I$ interval [0.5, 1]. When the value of $R S S I$ increases to 1, the broadcast period can decrease to about 2.3 s. Larger the $R S S I$ is, better the current channel state is, more safety messages can be broadcasted simultaneously. Thus, the broadcast period $T$ could be decreased in order to increase the broadcast frequency of safety messages and improve the real-time performance.

Figure 11.

Graph of broadcast period changing over RSSI ( $N = 13$ , $V = 90$ , $D = 15.1$ ).

T changes with different D

As shown in Figure 12, as $N = 13$ , $V = 90$ , $R S S I = 0.5$ , the broadcast period $T$ will increase with the increase of visibility $D$ , especially in the interval [10, 12] km. When the value of visibility increases to 12 km, the broadcast period can increase to 3.6 s. Less the visibility is, less the drivers’ visual field is. Thus, we should decrease the broadcast period of safety messages, in order to update safety messages timely and help the drivers know the situation of surrounding vehicles. With the visibility increasing, it means the weather condition changes better and the drivers can have a larger visual field. Therefore, the safety messages should not be broadcasted frequently. The broadcast period could be increased to avoid channel congestion.

Figure 12.

Graph of broadcast period changing over visibility ( $N = 13$ , $V = 90$ , $D = 15.1$ ).

In summary, through the fuzzy logic inference, the broadcast period $T$ could change adaptively according to the input variables. And the values of $T$ accorded with the theoretical logic realize the function of adjusting the broadcast period dynamically.

Simulation results and discussions

In this section, we will verify the performance of the FL-ASB protocol through simulations. The simulations are based on the following assumptions:

The bidirectional four-lane highway in this article is an ideal highway to meet the following conditions:

The width of lanes is between 3.75 and 4.50 m. In this article, we define the width of lanes is 3.75 m.

All vehicles are cars.

The drivers have good driving skills and obey the traffic rules.

All vehicles are equipped with GPS and WAVE/DSRC equipment.

The basic parameters of the vehicles are the same, such as acceleration, maximum speed, minimum speed.

The communication range of vehicles is 300 m.

Mobility model and simulation parameters

In this article, in order to ensure the realness and reliability of broadcast protocol simulation results, we adopt VanetMobiSim²⁸ to simulate the vehicle following model in a bidirectional four-lane highway scenario with a vehicle speed of 120 km/h. After using the VanetMobisim software to execute the vehicle mobility simulation file in the format of .xml, a file in the format of .tcl will automatically be generated in the storage path. This .tcl file can be imported into the NS2²⁹ software through the “source” command to set the network topology and the corresponding vehicle mobility situation.

In this article, the simulator platform is NS2, which operated on the Ubuntu system installed in Vmware Workstation 10.0. For simulating the broadcast protocol, the C++ language is used to modify the corresponding Mac802_11Ext and WirelessPhyExt modules in NS2, according to the IEEE 802.11p protocol. The other simulation parameters are shown in Table 5.

Table 5.

Simulation parameters.

Parameters	Values
PHY/MAC protocol	IEEE 802.11p
Channel model	Nakagami (m = 3)
Propagation model	Two-ray ground
Size of PSM (bytes)	256
Simulation range (m²)	15 × 3000
Simulation times (s)	500
Number of vehicles (veh/lane)	20, 40, 60, 80, 100, 120
Vehicle speed (km/h)	[60, 120]
Visibility (km)	>15
RSSI	0.8

PSM: periodic safety message; RSSI: received signal strength index.

After the simulation process of using the NS2 software to execute the broadcast protocol, a trace file in the format of .tr can be automatically generated, which includes the packets transmission record during the execution of the broadcast protocol. By extracting the records in the .tr file, we can analyze the performance of our proposed broadcast protocol and compare with other related protocols.

Simulations and analysis

To verify the advantages of FL-ASB protocol, this section will compare FL-ASB with fixed-period single-hop broadcast protocol (i.e. FSB-0.05s and FSB-0.25s) and adaptive-period single-hop broadcast protocol ASB.²¹ The performance index are network load ratio, delivery ratio, and average one-hop delay.

Network load ratio

The network load ratio is equal to the ratio of the number of PSM packets delivered in the network to the total number of packets delivered in the network. It is used to measure the condition that PSM messages occupy the network resources. Figure 13 shows the tendency of the network load ratio of each protocol with the increase of the number of vehicle nodes $N$ .

Figure 13.

Network load ratio comparison between FL-ASB and other protocols.

From Figure 13, we can see that the fixed-period single-hop broadcast protocol FSB-0 05s and FSB-0.25s have the highest network load ratio. This is because they cannot adaptively adjust the broadcast period according to the network density. The ASB protocol can adjust the broadcast period according to network density; therefore, its network load ratio is much lower than the FSB protocols. However, with the increase of the number of vehicle nodes, the network load ratio increases obviously. The FL-ASB protocol proposed in this article has optimized the statistical method of the number of neighbor nodes, based on the ASB protocol. As a result, the network load ratio of FL-ASB is less than that of ASB, and the network load ratio of FL-ASB is basically stable as the number of vehicles increases.

Delivery ratio

The delivery ratio is the ratio of the number of PSM packets received successfully to the total number of PSM packets transmitted in the network. It is used to measure the validity of the broadcast protocol. Figure 14 shows the tendency of delivery ratio of each protocol with the increase of the number of vehicle node $N$ .

Figure 14.

Delivery ratio comparison between FL-ASB and other protocols.

From Figure 14, we can see that the delivery ratio of FSB-0.05s, FSB-0.25s, and ASB show obvious downward trend with the increase of number of vehicle nodes. It is because that as the vehicle density increases, the broadcast period cannot be well adjusted, resulting in broadcasting packets frequently. Thus, packets collision increases and the delivery ratio decreases. The FL-ASB protocol proposed in this article can guarantee a high delivery ratio. On one hand, the FL-ASB protocol can adaptively adjust the broadcast period according to the number of vehicle nodes. Moreover, the FL-ASB protocol takes a wider range of considerations when calculating the number of one-hop neighbor vehicles. Thus, it takes into account more one-hop vehicle nodes and makes the broadcast period larger to decrease the packets collision probability. On the other hand, the FL-ASB protocol also considers the channel state. Adjusting the broadcast period according to different channel states also decreases the packet collision probability and increase the delivery ratio to a certain degree.

Average one-hop delay

Average one-hop delay means the time from a PSM packet being sent to being received in one hop, which is used to measure the timeliness of the broadcast protocol. Figure 15 shows the tendency of average one-hop delay of each protocol with the increase of the number of vehicle nodes $N$ .

Figure 15.

Average one-hop delay comparison between FL-ASB and other protocols.

From Figure 15, we can see that fixed-period single-hop broadcast protocol FSB-0.05s and FSB-0.25s have a higher average one-hop delay, and the average one-hop delay of FSB-0.05s is higher than that of FSB-0.25s. This is because the fixed-period single-hop broadcast protocols cannot adjust the broadcast period according to the number of vehicle nodes. When the vehicle density is small, it has a high delivery delay due to the low network connectivity. While the vehicle density is large, broadcasting packets frequently can lead to packet collisions and increase delay due to the reason that the broadcast period cannot be adjusted timely. The adaptive-period single-hop broadcast protocol ASB and FL-ASB both have low and stable average one-hop delay, because they can adaptively adjust the broadcast period according to the traffic density. In addition, the FL-ASB protocol proposed in this article has lower and more stable average one-hop delay than the ASB protocol. This is because the FL-ASB protocol optimizes the statistical method of the number of neighbor nodes on the basis of the ASB protocol and takes the channel state and other factors into consideration. Therefore, the FL-ASB protocol has a better adaptive performance of broadcast period and can effectively decrease the collision probability caused by frequent packet transmission and decrease the delay.

From the analysis above, the FL-ASB protocol has shown a better performance in the network load ratio, average one-hop delay, and delivery ratio. What is more, in addition to the traffic density, the FL-ASB protocol also considers the channel state and weather condition. Thus, the FL-ASB protocol has better scalability and adaptability compared with other protocols.

Conclusion

In this article, we design an adaptive-period single-hop broadcast protocol based on fuzzy logic theory. We take into account various factors which could affect broadcast period, such as the number of one-hop neighbor nodes, vehicle speed, received signal strength, and visibility. Furthermore, we analyze the impact of various factors on the single-hop broadcast period. Using the fuzzy logic inference method, we realize the adaptive changes of the PSMs broadcast period. Finally, through simulations, we verify that our proposed FL-ASB protocol has shown a better performance in network load ratio, average one-hop delay, and delivery ratio. In addition, the FL-ASB protocol has better scalability and adaptability compared with other protocols. There are still open issues in the future. First, the design of broadcast protocol should accord with the actual traffic environment. In this article, we have made some assumptions to simplify the problem. Second, the research should take into account more complex road scenario. How to design a broadcast protocol in urban scenario needs to be investigated.

Footnotes

Handling Editor: Carlos Calafate

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article is supported by the State Key Laboratory of Rail Traffic Control and Safety (contract no. RCS2016ZT015), Beijing Jiaotong University. And it is supported in part by the Chinese National Programs for High Technology Research and Development 863 project (no. 2015AA016005).

References

Liang

Zhang

et al . Vehicular ad hoc networks: architectures, research issues, methodologies, challenges, and trends. Int J Distrib Sens N 2015; 11(8).

Jeong

. Survey on protocols and applications for vehicular sensor networks. Int J Distrib Sens N 2016; 12(8).

Zhuang

Pan

Luo

et al . Time and location-critical emergency message dissemination for vehicular ad-hoc networks. IEEE J Sel Area Comm 2011; 29(1): 187–196.

Panichpapiboon

Pattara-Atikom

. A review of information dissemination protocols for vehicular ad hoc networks. IEEE Commun Surv Tut 2012; 14(3): 784–798.

Yan

Zhang

Wang

et al . Access points planning in urban area for data dissemination to drivers. IEEE T Veh Technol 2014; 63(1): 390–402.

Hartenstein

Laberteaux

. A tutorial survey on vehicular ad hoc networks. IEEE Commun Mag 2008; 46(6): 164–171.

Shah

SAA

Ahmed

Xia

et al . Adaptive beaconing approaches for vehicular ad hoc networks: a survey. IEEE Syst J 2018; 12(2): 1263–1277.

Schmidt

Leinmuller

Schoch

et al . Exploration of adaptive beaconing for efficient intervehicle safety communication. IEEE Network 2010; 24(1): 14–19.

Rezaei

Sengupta

Krishnan

. Reducing the communication required by DSRC-based vehicle safety systems. In: Proceedings of the intelligent transportation systems conference, Seattle, WA, 30 September–3 October 2007, pp.361–366. New York: IEEE.

10.

Torrent-Moreno

Mittag

Santi

et al . Vehicle-to-vehicle communication: fair transmit power control for safety-critical information. IEEE T Veh Technol 2009; 58(7): 3684–3703.

11.

Kloiber

Harri

Strang

. Dice the TX power: improving awareness quality in VANETs by random transmit power selection. In: Proceedings of the vehicular networking conference, Seoul, South Korea, 14–16 November 2012, pp.56–63. New York: IEEE.

12.

Settawatcharawanit

Choochaisri

Intanagonwiwat

et al . V-DESYNC: desynchronization for beacon broadcasting on vehicular networks. In: Proceedings of the 75th vehicular technology conference, Yokohama, Japan, 6–9 May 2012, pp.1–5. New York: IEEE.

13.

Park

Kim

. Collision control of periodic safety messages with strict messaging frequency requirements. IEEE T Veh Technol 2013; 62(2): 843–852.

14.

Huang

Fallah

Sengupta

et al . Adaptive intervehicle communication control for cooperative safety systems. IEEE Network 2010; 24(1): 6–13.

15.

Rawat

Popescu

Yan

et al . Enhancing VANET performance by joint adaptation of transmission power and contention window size. IEEE T Parall Distr 2011; 22(9): 1528–1535.

16.

Zhong

Wolfson

. Disseminating real-time traffic information in vehicular ad-hoc networks. In: Proceedings of the intelligent vehicles symposium, Eindhoven, 4–6 June 2008, pp.1056–1061. New York: IEEE.

17.

Nakorn

Rojviboonchai

. Bloom filter for fixed-size beacon in VANET. In: Proceedings of the 79th vehicular technology conference, Seoul, South Korea, 18–21 May 2014, pp.1–5. New York: IEEE.

18.

Sommer

Tonguz

Dressler

. Traffic information systems: efficient message dissemination via adaptive beaconing. IEEE Commun Mag 2011; 49(5): 173–179.

19.

Zemouri

Djahel

Murphy

. Smart adaptation of beacons transmission rate and power for enhanced vehicular awareness in VANETs. In: Proceedings of the 17th international conference on intelligent transportation systems, Qingdao, China, 8–11 October 2014, pp.739–746. New York: IEEE.

20.

Chaabouni

Hafid

Sahu

. A collision-based beacon rate adaptation scheme (CBA) for VANETs. In: Proceedings of the international conference on advanced networks and telecommunications systems, Kattankulathur, India, 15–18 December 2013, pp.1–6. New York: IEEE.

21.

Hassan

Ahmed

Rahman

. Adaptive beaconing system based on fuzzy logic approach for vehicular network. In: Proceedings of the 24th international symposium on personal indoor and mobile radio communications, London, 8–11 September 2013, pp.2581–2585. New York: IEEE.

22.

Keller

Liu

Fogel

. Basic fuzzy set theory. New York: Wiley-IEEE Press, p.400.

23.

Bae

. A dynamic HI-CAST protocol for safety message dissemination in VANET. In: Proceedings of the 5th international conference on ubiquitous and future networks, Da Nang, Vietnam, 2–5 July 2013, pp.827–831. New York: IEEE.

24.

Bae

Olariu

. Design and evaluation of HI-CAST and its variants for safety message dissemination in VANET. In: Proceedings of the 13th international conference on control, automation and systems, Gwangju, South Korea, 20–23 October 2013, pp.800–806. New York: IEEE.

25.

Ohzahata

Kato

. Fuzzy logic based multi-hop broadcast for high-density vehicular ad hoc networks. In: Proceedings of the vehicular networking conference, Jersey City, NJ, 13–15 December 2010, pp.17–24. New York: IEEE.

26.

China Construction Standard Highway Committee. Technical standard of highway engineering (JTG B01-2014). Beijing: China Communications Press, 2014, p.120.

27.

Zhang

Liu

. Fuzzy risk analysis based on COG method. J Liaoning Tech Univ (Nat Sci Ed) 2012; 31(2): 240–243.

28.

VanetMobiSim, https://sourceforge.net/projects/vanetmobisim/

29.

Network Simulator version 2 (NS2), https://sourceforge.net/projects/nsnam/

FL-ASB: A Fuzzy Logic Based Adaptive-period Single-hop Broadcast Protocol

Abstract

Keywords

Introduction

Fuzzy logic theory

Fuzzy set and membership function

Fuzzy inference

Adaptive-period method based on fuzzy logic

Scenario and problem description

Statistical method for the number of neighbor nodes in one-hop

Necessity of packet carried time

Comprehensiveness of impact factor

Selection of impact factor

Number of one-hop neighbor nodes N (veh)

Vehicle speed V (m/s)

Received signal strength index

Visibility D (km)

Fuzzification

Membership function of the number of one-hop neighbor nodes N (veh)

Membership function of vehicle speed V (m/s)

Membership function of RSSI

Membership function of visibility D (km)

“IF-THEN” fuzzy rules set

Defuzzification

Fuzzy inference process

T changes with different N

T changes with different V

T changes with different RSSI

T changes with different D

Simulation results and discussions

Mobility model and simulation parameters

Simulations and analysis

Network load ratio

Delivery ratio

Average one-hop delay

Conclusion

Footnotes

Declaration of conflicting interests

Funding

References

Membership function of vehicle speed $V$ (m/s)