Sage Journals: Discover world-class research

Abstract

In VANET (vehicular ad hoc network), RSUs (road side units) have limited coverage and high deployment cost, so they are deployed sparsely in urban area, which leads to blind zone (BZ) between adjacent RSUs, where vehicles cannot connect to Internet by RSUs. In this paper, we study how to use RSUs and vehicles to download big files cooperatively in BZ. In order to choose the cooperative vehicles with minimum total delay, we propose a sequential decision vehicle selection method based on the residual file (SSRF). The method divides the process of selection into several stages and selects cooperative vehicles from the candidates; the decision sequence generated by SSRF determines the set of cooperative vehicles. Simulation and data analysis show that our method is effective in terms of delivered ratio and file delivered delay.

1. Introduction

VANET is a type of mobile ad hoc networks consisting of vehicles and RSUs. As the deployment of RSUs, many recent researches use 802.11p standard to realize interaction between vehicles and Internet by RSUs [1]. The typical interactive information includes not only road condition and traffic safety text information, but also multimedia files. In VANET, RSUs have limited coverage range and high deployment cost [2, 3], so they are deployed sparsely especially in out-of-hotspots, for example, residential areas and industrial parks, which leads to the existence of BZ between RSUs. In addition, vehicles with high mobility and frequent network topology change [4] make it difficult to maintain uninterrupted Internet access via RSUs. This is especially critical when they want to download big files. In urban VANET, road topology is more complex. So the scenario with two-way road and double crossings is studied in this paper. We aim to reduce the delay of downloading the whole big file (called the total delay) by making full use of two-way vehicles. The interactive protocol between RSUs and vehicles is 802.11p standard [5]. Here, we use the two-way vehicles to help the downloading vehicle (De for short) download big file. It is natural to use the vehicles with the same direction as De, because they have enough time to forward part of the big file. It is also possible to use the reverse vehicles, since the link duration time is more than 11.25 s (based on a maximum speed of 80 km/h and transmission range of 250 m), so they also have time to forward cooperatively under the 802.11p standard [6, 7]. As shown in Figure 1, when De leaves the coverage of RSU1, the communication between two RSUs by Internet will distribute the rest of the big file to the vehicles between them with the same and reverse direction as De such as vehicle 1 and vehicle 2 to help De to download the file. The selected vehicles use store-and-forward mechanism [7] and deliver the part of big file to De by opportunistic contacts. Therefore, the most important issue in the cooperative downloading is how to select the appropriate vehicles to reduce the total delay.

Figure 1

Vehicles store-and-forward in BZ.

Sequential Decision Problem (SDP) is composed of multidecision stages related to time order, which adapts to the dynamic joining and leaving of vehicles. A target solution for each stage will be obtained by SDP. However, we mainly consider the residual file in this paper, so an improved SDP method called SSRF is proposed in Section 4. The earlier the link that between cooperative vehicles and the De starts, the earlier the big file can be delivered. The shorter link duration time may lead to more frequent handover. Therefore, the link start time and link duration time between the De and cooperative vehicles are the primary factors to consider. In other words, the link start time should be close to the link end time of the last cooperative vehicle and the file block size carried by the selected vehicle should be as large as possible. In the proposed SSRF, the selection process is divided into several stages, and the generated decision sequence is exactly the cooperative vehicles.

The rest of the paper is as follows. Related work is presented in Section 2. Section 3 gives the model description. Section 4 presents the proposed SSRF method, and the performance evaluation of cooperative vehicles in urban scenarios is reported in Section 5.

2. Related Work

A number of cooperative mechanisms have been proposed in VANET: Mob-Torrent was presented in [8], which uses 3G network, where the AP (Access Point) prefetched data to mobile helpers. Therefore the total amount of data transferred is increased. In [9], a DC-ARQ mechanism to reduce packet losses in transmissions and a carry-and-forward method to improve delivery of information were proposed. Cooperative distribution of safety information in highway VANET was studied in [10]. The DSRelay in [11] proposed vehicle selection algorithm in highway VANET to ensure the connection probability, but it had long delay. The typical cooperative download schemes in [8–11] considered the highway VANET scenario without complex topology and changeful vehicle behaviors, which was quite different from urban VANET. All [8, 9, 11] only consider undirectional vehicles to cooperative download and cannot take full advantage of two-way traffic, especially in the urban VANET. In recent years, some works studied transferring big file by stopped vehicles [12, 13], but parked vehicles were limited by power and other factors.

Our study is different from the work listed above in two points. (1) Different scenario: we study the scenario of two-way road with double crossings. Complex topology and changeful vehicle behaviors are different from highway. (2) Different objects: in urban VANET, the road is mostly two-way. Although vehicles moving in the same direction may have longer link time, they cannot ensure their connection with De. Reverse vehicles connect to the destination vehicle with probability 1, but the link time may be shorter. Therefore, we take full advantage of two-way vehicles.

3. System Problem Formulation

3.1. Problems and Assumption

In VANET, there are three types of communication: RSU-RSU, Vehicle-RSU, and Vehicle-Vehicle. Internet is the resource provider and the bridge of communication among RSUs [14]. The assumptions in our model are as follows: (1) by using GPS technologies, accurate information (speed and location of vehicles) can be achieved (error < 1 m) [15]; (2) energy consumption and computation resources are enough in VANET; (3) vehicles and RSUs can build connection and transfer and download file once they move into each other's coverage. Figure 2 shows that De moves from RSU1 to RSU2, and it downloads part of a big file when traveling across the coverage of RSU1 quickly. Then, RSU1 will choose some vehicles as candidates from straight-moving vehicles (e.g., b) and right-turning vehicles (e.g., c); RSU2 chooses candidates from straight-moving, right-turning, and left-turning vehicles e, d, and f as well. The goal is to guarantee the successful delivery to De. RSU1 does not use the left-turning vehicle a, because the traffic light will reduce the probability of a meeting De. The key problem we should solve is that how to reduce the total delay by designing a series of analysis and algorithm, in order to guarantee the file delivery. To achieve this purpose, we assume that each vehicle will register information ( $V_{i}^{n}$ , $t_{i}^{k}$ , D, and L) once it moves into RSU's coverage; here $V_{i}^{n}$ is the current speed; $t_{i}^{k}$ means register time when vehicle i come into the RSUk; D represents the driving directions; and $D_{d}$ , $D_{l}$ , and $D_{r}$ indicate straight-moving, right-turning, and left-turning, respectively; L is the location and $L_{u}$ , $L_{d}$ , $L_{l}$ , and $L_{r}$ represent vehicle coming from up, down, left, and right of crossing, respectively.

Figure 2

System scenario with double crossings of two-way vehicles.

3.2. Channel Model

In this paper, $V_{k t}^{i}$ represents the downloading rate when vehicle i crosses RSUk at time t. Therefore, the channel capacity is given by [16]

\begin{matrix} V_{k t}^{i} = W_{k} \log (1 + \frac{P}{N_{0} W_{k} d_{t}^{γ}}), \end{matrix}

(1)

where

W_{k}

is the channel bandwidth of RSUk, P is transmitting power,

d_{t}

is the distance between the RSUs and vehicle at time slot t, γ is the path loss exponent and,

N_{0}

is number of vehicles. The RSU knows the value of

N_{0}

since vehicles need to establish and terminate their connections when they enter and leave the coverage range.

4. Select Cooperative Vehicles by SSRF

4.1. Selection of Candidates

The key to select candidates is to know in advance, whether one or more vehicles currently within the coverage of the RSU will meet the target De. Since network topology changes fast and the effectiveness of the historical data cannot be guaranteed, in order to select the reliable candidates, we should combine historical and current data. In addition, it is difficult to keep a constant velocity for each vehicle during the driving, so it is not accurate to take the registered velocity as the velocity of vehicles driving between the two RSUs. We divide candidates into two classes: vehicles moving after or before the De, which are added into sets Veh_after and Veh_before, respectively. The candidate selection process ends once De drives out of coverage of RSU1.

There are two scenarios in calculating the distance between vehicles in Veh_before and De: (1)

straight road:

\begin{matrix} d_{D e}^{i} = V_{i} (t_{D e}^{k} - t_{i}^{k}), \end{matrix}

(2)

where

V_{i}

is the speed of vehicle i; that is,

\begin{matrix} V_{i} = α V_{i}^{n} + \frac{(1 - α) (d_{k - 1} + 2 R)}{(t_{i}^{k} - t_{i}^{k - 1})}, \end{matrix}

(3)

where α is weighted factor;

d_{k - 1}

is the distance between RSUk-1 and RSUk; R is the coverage radius of RSU.

(2)

Right-turning: when De enters into RSU1, any right-turning vehicle i in straight road; then the distance is given by

\begin{matrix} d_{D e}^{i} = R + \frac{d_{0}}{4} - V_{D e} (t_{D e}^{k} - t_{i}^{k} + \frac{(R - d_{0} / 4)}{V_{i}}) . \end{matrix}

(4)

Otherwise,

\begin{matrix} d_{D e}^{i} = R + \frac{d_{0}}{4} + V_{i} (t_{D e}^{k} - \frac{(R - d_{0} / 4)}{V_{i}}), \end{matrix}

(5)

where

d_{0}

is the width of road.

Also, there are also two cases for the distance between any vehicle in Veh_after and De: (1)

straight road:

\begin{matrix} d_{D e}^{j} = V_{j} (t_{j}^{k} - t_{D e}^{k}); \end{matrix}

(6)

(2)

right-turning:

\begin{matrix} d_{D e}^{j} = V_{D e} [\frac{(R - d_{0} / 4)}{V_{i}} + t_{j}^{k} - t_{D e}^{k}] - R - \frac{d_{0}}{4} . \end{matrix}

(7)

The time of right-turning vehicle i in Veh_before arriving at straight road is

\begin{matrix} t^{i} = {\begin{cases} t_{D e}^{k}, & i has arrived at straight \\ road when D e register, \\ t_{i}^{k} + \frac{(R - d_{0} / 4)}{V_{i}}, & otherwise . \end{cases} \end{matrix}

(8)

Also, the right-turning vehicle j,

t^{j}

is given by

\begin{matrix} t^{j} = {\begin{cases} t_{j}^{k}, & j has arrived at straight \\ road when {Veh}_{D e} register, \\ t_{j}^{k} + \frac{(R - d_{0} / 4)}{V_{j}}, & otherwise . \end{cases} \end{matrix}

(9)

In addition,

t_{start}^{i}

and

t_{end}^{i}

are link start time and link end time, respectively:

\begin{matrix} t_{start}^{i} (k) = \max (\frac{2 R}{V_{D e}} + t_{D e}, t^{i} + \frac{(d_{D e}^{i} - r)}{| V_{D e} - V_{i} |}), \\ t_{end}^{i} (k) = \min (\frac{(2 R + d_{k})}{V_{D e}} + t_{D e}, t^{i} + \frac{(d_{D e}^{i} + r)}{| V_{D e} - V_{i} |}), \end{matrix}

(10)

where r is the transmission range of vehicle.

The candidates selected by RSU2 are all to the left of De and oncoming. We ignore the effect of traffic lights nearby RSU2, because the traffic lights in different crossings are not synchronized. Therefore, any vehicle j that can be a candidate should be oncoming and meet De between two adjacent RSUs. The computation is the same as above and $t_{start}^{i}$ and $t_{end}^{i}$ :

\begin{array}{l} t_{start}^{j} (k + 1) = \frac{(4 R + d_{k} + b * d_{0} / 4 - | t_{j}^{k + 1} - t_{D e}^{k} | - r)}{(V_{D e} + V_{j})}, \\ t_{end}^{j} (k + 1) \\ = \min (\frac{2 R + d_{k}}{V_{D e}} + t_{D e}^{k}, \\ \frac{4 R + d_{k} + b * d_{0} / 4 - | t_{j}^{k + 1} - t_{D e}^{k} | + r}{V_{D e} + V_{j}} + t_{D e}^{k}) . \end{array}

(11)

Next, we suppose that the cache of all candidates is large enough for cooperative downloading file. Vehicles have the information ( $t_{start}^{i}$ , $t_{end}^{i}$ , $t_{i}^{j}$ ). We use $T_{j}^{i}$ to present the driving time of i in RSUj:

\begin{matrix} T_{j}^{i} = {\begin{cases} \frac{2 R + β * d_{0} / 2}{V} - t_{D e}^{j} + t_{i}^{j}, & t_{D e}^{j} \geq t_{i}^{j}, \\ \frac{2 R}{V_{i}}, & t_{D e}^{j} < t_{i}^{j}, \end{cases} \\ β = {\begin{cases} 0, & straight-moving, \\ 1, & left-turing, \\ - 1, & right-turing. \end{cases} \end{matrix}

(12)

Therefore, when i comes across RSUj, the maximum file block it can carry is

\begin{matrix} N_{j}^{i} = \int_{t_{D e}}^{T_{j}^{i} + t_{D e}} V_{j t}^{i} d t . \end{matrix}

(13)

We use $V_{Veh i}$ as the transmission rate between i and De, so the file that i can transfer to De is given by $N_{V}^{i} = V_{Veh i} * (t_{end}^{i} - t_{start}^{i} - t_{con})$ , where $t_{con}$ is link building time. So, the maximum file block a candidate can carry is

\begin{matrix} N_{i} = \min (N_{j}^{i}, N_{V}^{i}) . \end{matrix}

(14)

4.2. SSRF

Cooperative vehicle selection is a dynamic decision process. Vehicles in the set $V = {{Veh}_{after}, {Veh}_{before}}$ is sorted ascendingly according to link start time, and RSU knows all information ( $t_{start}^{i}$ , $t_{end}^{i}$ , $N_{i}$ ) of vehicles in its converge.

SDP is to find a decision sequence among the decision space, which can optimize target function value. In the model, we assume that $x_{t}^{n}$ is state variable; $X (t) = {x_{t}^{0}, x_{t}^{1}, \dots, x_{t}^{n}}$ is a feasible state set; $U (t) = {u_{t}^{0}, u_{t}^{1}, \dots, u_{t}^{m}}$ is a feasible decision variable set; $z_{t}^{p}$ is a strategy and $Z (t) = {z_{t}^{0}, z_{t}^{1}, \dots, z_{t}^{p}}$ is the set of feasible strategies; $D (l) = {d_{1}^{l}, d_{2}^{l}, \dots, d_{t}^{l}, \dots, d_{N}^{l}}$ is a feasible strategy set for SDP; X, U, and Z present state space, decision space, and strategy space; $\hat{U} = {{\hat{u}}_{1}, {\hat{u}}_{2}, \dots, {\hat{u}}_{V}}$ is a feasible bound set; f is State Transition function; and F is target function. If the problem is to minimize the target function, SDP can be described as

\begin{matrix} \min F (t) = F (t, x_{t}^{n}, u_{t}^{m}, z_{t}^{p}), t = 1,2, \dots, N, \\ x_{t + 1}^{n} = f (t, x_{t}^{n}, u_{t}^{m}, z_{t}^{p}), \\ x_{t}^{n} \in X (t) \subset X \subset R^{n}, \\ u_{t}^{m} \in U (t) \subset U \subset R^{m}, \\ z_{t}^{p} \in Z (t) \subset Z \subset R^{p} . \end{matrix}

(15)

Here the set of state decisions is decided by feasible bound set

\hat{U}

. With the evolution of constraint condition, the estimate of decider for delay will change. For example, we cannot always select vehicle with earlier link start time as cooperative vehicle; otherwise, it could lead to transmission collision. In the same way, we cannot always select vehicle with earlier link end time as cooperative vehicle, because both of the two ways will increase total delay.

In this paper, we define the size of residual file $N_{r}$ as a state viable; the decision viable set is the candidate set $V (r) = {V_{1}, V_{2}, \dots, V_{r}}$ . The candidates are not selected as cooperative vehicles yet; so, the feasible strategy is

\begin{matrix} x_{i} = {\begin{cases} 0, & i not selected, \\ 1, & i selected . \end{cases} \end{matrix}

(16)

Definition 1.

Suppose that there is a structure $G = (C, W, Ø)$ that satisfies the following characteristics as follows (see Figure 3): (1)

$C = {c_{0}, c_{1}, \dots, c_{n - 1}, c_{e}}$ includes $n + 1$ layers, where $c_{0}$ is the begin layer, $c_{e}$ is the terminal layer, and the set of nodes in layer $c_{i}$ is $B_{i} = {B_{i (1)}, B_{i (2)}, \dots, B_{i (k)}}$ .

(2)

$w_{i j} (r, h)$ is the weight of nodes r in $B_{i (r)}$ and h in $B_{i (h)}$ , where $w_{i j} = {w_{i j} (r, h) ∣ r \in B_{i}, j \in B_{j}}$ , is the weight set of layer $c_{i}$ and layer $c_{j}$ .

(3)

W is a set of route ω in G, which satisfies the conditions: ω starts from layer $c_{0}$ once and only once it passes through $c_{0}, c_{1}, \dots, c_{n - 1}$ , successively, and arrives at $c_{e}$ finally.

(4)

Function F is a mapping of set W for feasible solution to solve the problem. That is, for each route ω fulfilling the conditions, we can find a corresponding feasible solution to solve the problem; similarly, it is the same in reverse.

Figure 3

Structure chart of SSRF.

From the definition above we can see that the structural map converts the feasible solution into a kind of path coding. The length of the path in Figure 3 is the target function value of the problem and the minimum delay corresponding to the shortest route $ω^{*}$ fulfilling the conditions in G.

By the previous analysis, as layer $c_{0}$ only has the destination vehicle De, the link end time of layer $c_{0}$ is $t_{D e} = 2 R / V_{D e}$ . The rules for each layer to select nodes are as follows: if there is overlap between adjacent layers such as 2 and 3, as shown in Figure 4, we should select one node which has the latest link end time as the next cooperative vehicle, because without blank area vehicles should download more file cooperatively to reduce delay. So we use the difference of two nodes' link end time as the weight of the two nodes. If all nodes in the next layer do not overlap with the current selected node, the node which has the smallest difference between its link start time and the link end time as the next cooperative vehicle should be selected. According to the rules above, the vehicle sequence we choose in Figure 4 is 1, 3, and 4. Therefore, we use the difference between the link start time of next node and the link end time as the weight of the two nodes. So smaller weight values are better.

Figure 4

Weighted value analysis between two layers.

In addition, we should consider another case: in the dynamic selection, vehicles entering into RSU later may make up the blank area described above, because selection of cooperative vehicles is dynamic and real-time. Here, we presume that De can receive unordered file blocks.

After selecting node from layer $c_{i}$ , we will select next node in layer $c_{i + 1}$ , as shown in Figure 5. There is blank area between vehicles 1 and 2, 3, so we select 2 and 5 as cooperative vehicles according to the rules above. Now we pay attention to vehicle 4. Although vehicle 4 entered into RSU later, it has a faster speed and will build link with De before 2. Therefore, we should use vehicle 4 to make up the blank area between vehicle 1 and vehicle 2 to reduce the total delay.

Figure 5

Transmission blank area.

From the above, the value of $w_{i j} (r, h)$ is defined as follows: if $t_{end}^{r} \geq t_{start}^{h}$ , $w_{i j} (r, h) = t_{end}^{r} - t_{start}^{h}$ , otherwise $w_{i j} (r, h) = t_{start}^{h} - t_{end}^{r}$ . Therefore, the target function of SSRF is given by

\begin{array}{l} \min F (N_{r}) = \sum_{i = 0}^{n - 2} w_{i j} (r, h) + t_{con} (n - 1) \\ s . t . j = i + 1 \\ r \in B_{i}, j \in B_{j}, \\ x_{r} = 1, x_{h} = 1 . \end{array}

(17)

In the improved method, there is a main selection in each layer as the solution of target function, and there may exist a secondary selection to make up blank area (see Algorithm 1).

Algorithm 1: SSRF algorithm.

Set $N_{r}$ = N, $r = De$ , $j = i + 1$

While $N_{r}$ > 0 and $t_{end}^{r} < t_{De} + (2 R + d_{k}) / V_{De}$

Select the minimum one from the lower:

$w_{i j} (r, h)$ = min $(w_{i j} (r, h) ∣ h \in B_{j})$ ;

Vehicle h is selected as cooperative vehicle:

$x_{h} = 1; r = h$ ;

if no overlap $w_{i j} (r, h) \geq 0$

{file size vehicle cooperative download

$N_{h}^{D} = N_{h}$ ;

for $(i = i + 1; i < n; i + +)$

${j = i + 1$ ;

$w_{i j + 1} (r, s) = \min (w_{i j + 1} (r, s) ∣ s \in B_{j + 1})$ ;

if $t_{start}^{s} < t_{start}^{h}$ s overtakes h

$N_{h}^{D} = N_{h}^{D} + V_{Veh h} * (t_{start}^{h} - t_{start}^{s})$ ;

}else $N_{h}^{D} = \min (N_{k}^{h}, V_{Veh h} * (t_{end}^{h} - t_{end}^{s}))$ ;}

$N_{r} = N_{r} - N_{h}^{D}; i + +; j = i + 1$ ;

end while

In conclusion, no matter whether the file is downloaded completely or not between the two RSUs, SSRF will reduce the total delay: (1) if the file is downloaded completely before De enters into the next RSU, the set of cooperative vehicles ${Veh}_{end} = {{Veh}_{1}, {Veh}_{2}, \dots, {Veh}_{p}}$ obtained from SSRF will ensure the minimum total delay, because in this case, the key to ensure the minimum total delay is to reduce the blank area. More precisely, $t_{start}^{s} < t_{start}^{h}$ and $N_{h}^{D} = N_{h}^{D} + V_{Veh h} * (t_{start}^{h} - t_{start}^{s})$ in the algorithm are to ensure minimum blank area. (2) Otherwise, if De cannot completely download the whole file, we should ensure that De can download file as much as possible. In the algorithm, $N_{h}^{D} = N_{h}^{D} + V_{Veh h} * (t_{start}^{h} - t_{start}^{s})$ and $N_{h}^{D} = \min (N_{k}^{h}, V_{Veh h} * (t_{end}^{h} - t_{end}^{s}))$ ensure the maximum transmission.

5. Simulation Analysis

In this section, the real taxi trace data within 30 days in San Francisco from [17] is used. We used the IEEE 802.11p physical layer, which defines an international standard for wireless access in vehicular environments [18]. In the experiments, given that the area of the simulated network is fixed, we use different numbers of vehicles varying between 50 and 250 to represent dense network and sparse network. We do the experiments in two aspects: to prove that our method is reasonable and correct and to compare it with other methods. The parameters are listed in Table 1 and the evaluating metrics are as follows. (1)

The total delay is a metric for evaluating the system performance. The relationship between the number of vehicles and total delay is described in Figure 6(c). With the number of vehicles changing from 50 to 100, the total delay decreases obviously, because there will be more cooperative vehicles as the number of vehicles increases. The total delay remains relatively stable when the number of vehicles is in the range of 100–150. Then as the number of vehicles increases further, the delay has a modest increase. The increased vehicles lead to vehicles moving slowly, so that vehicles must spend more time to meet De. As a result the total delay increases. From the longitudinal analysis of Figures 6(a) and 6(b), we can see the same result as Figure 6(c). To transversely analyze Figures 6(a) and 6(b), we can see that the total delay increases with the increase in file size and distance of BZ. Figure 6(d) proves that the SSRF we proposed has significant advantage. Therefore, under the same condition, our total delay is the minimum. The results of the simulation prove that the SSRF lowers the total delay obviously.

(2)

The number of cooperative vehicles is shown in Figure 7. Figure 7(a) indicates that the shorter the BZ distance, the smaller the number of cooperative vehicles. It is because on one hand, if the BZ distance is shorter, the number of vehicles we can choose is lower. On the other hand, shorter BZ distance leads to shorter cooperative download time, so the number of cooperative vehicles also reduces. But on the whole, the number of cooperative vehicles does not change too much. The research on the impact of file size and number of cooperative vehicle is studied in Figure 7(b). Obviously, as the size of file increases, the number of cooperative vehicles will increase. Longitudinal comparison in Figures 7(c) and 7(d) shows that when DSRelay chooses the vehicles with the same moving direction, the number of cooperative vehicles is the lowest. It is because the probability of vehicles in the same direction meeting De is small and the link duration time is longer. For the same reason, reverse vehicles will meet De with probability 1 and the link duration time is shorter, so the number increases. However, the SSRF we propose utilizes the characteristics of two-way vehicles. So the average number of cooperative vehicles is between those two other methods. It can be seen from Figure 7(d) that more cooperative vehicles are needed when the file size increases. However, the cooperative vehicles with the same direction can download cooperatively more files, so the number of cooperative vehicles increases slowly in the same case. From the whole data in the figures, it can be seen that the SSRF method considers both lower number of cooperative vehicles and the total delay.

(3)

In fact, since the selected cooperative vehicles may not meet De or collision with ongoing transmission happens, the delivered ratio will be decreased, as shown in Figure 8. It is reasonable that the more vehicles we can select, the higher the delivered ratio will be. On the other hand, we can see from Figure 8 that our method gets a higher delivered ratio than DSRelay, which proves that SSRF can guarantee the delivered ratio. It is the result of candidates' selection and the number of cooperative vehicles.

Table 1

System parameters.

Parameter	Quantity	Value
W	Bandwidth of RSU	10 MHz
$P / (N_{0} * W)$	Transmit signal-to-noise ratio	60 dB [13]
γ	Path loss exponent	3 [13]
R	RSU's transmission	500 m [4]
r	Vehicle's transmission	250 m [19]
$V_{i}$	Speed of vehicle i	20~80 km/h

Figure 6

Influence factors of total delay.

Figure 7

Average number of cooperative vehicles.

Figure 8

Delivered ratio.

6. Conclusion

In this paper, we mainly study how to take full advantage of two-way mobile vehicles to cooperatively download file in urban VANET with BZ of two adjacent RSUs. We improve the sequential method and propose a SSRF method, which divides the selection process into several stages, and decision sequence generated by SSRF is exactly the set of cooperative vehicles, which reduces the total delay and improves the delivery ratio. In the future work, we will mainly consider the study of multihop relay cooperative download and the management of links between vehicle-vehicle and RSU-vehicle.

Footnotes

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

This work is supported by National Natural Science Foundation of China (61370201), the Scientific Research Foundation for the Returned Overseas Chinese Scholars (45), Liaoning Provincial Natural Science Foundation of China (2013020019), and High-Tech 863 Program (2012AA111902).

References

Ott

Kutscher

A disconnection-tolerant transport for drive-thru internet environments

Proceedings of the IEEE INFOCOM

March 2005

Miami, Fla, USA

1849 1862

2-s2.0-25844469166

10.1109/INFCOM.2005.1498464

Shieh

W.-C.

Sou

S.-I.

Tsai

S.-Y.

A study of video frame sharing in sparse vehicular networks

Proceedings of the 17th IEEE International Conference on Parallel and Distributed Systems (ICPADS '11)

December 2011

444 448

2-s2.0-84856593950

10.1109/ICPADS.2011.15

Campolo

Molinaro

On vehicle-to-roadside communications in 802.11p/WAVE VANETs

Proceedings of the IEEE Wireless Communications and Networking Conference (WCNC '11)

March 2011

1010 1015

2-s2.0-79959306704

10.1109/WCNC.2011.5779273

Bitam

Mellouk

QoS swarm bee routing protocol for vehicular ad hoc networks

Proceedings of the IEEE International Conference on Communications (ICC '11)

June 2011

1 5

2-s2.0-80052166924

10.1109/icc.2011.5963424

Sathiamoorthy

Dimakis

A. G.

Bai

Distributed storage codes reduce latency in vehicular networks

Proceedings of the IEEE INFOCOM

March 2012

2646 2650

Bai

ETP: Encounter Transfer Protocol for opportunistic vehicle communication

Proceedings of the IEEE INFOCOM

April 2011

Shanghai, China

2201 2209

2-s2.0-79960858102

10.1109/INFCOM.2011.5935034

Fall

A delay-tolerant network architecture for changed Internets

Proceedings of the ACM SIGCOMM

August 2003

27 34

Chen

B. B.

Chan

M. C.

MobTorrent: a framework for mobile internet access from vehicles

Proceedings of the 28th Conference on Computer Communications, IEEE INFOCOM

April 2009

1404 1412

2-s2.0-70349658487

10.1109/INFCOM.2009.5062056

Trullols-Cruces

Morillo-Pozo

Barcelo

J. M.

Garcia-Vidal

A cooperative vehicular network framework

Proceedings of the IEEE International Conference on Communications (ICC '09)

June 2009

1 6

2-s2.0-70449488308

10.1109/ICC.2009.5198586

10.

Fallah

Y. P.

Huang

C.-L.

Sengupta

Krishnan

Analysis of information dissemination in vehicular ad-hoc networks with application to cooperative vehicle safety systems

IEEE Transactions on Vehicular Technology 2011 60 1 233 247

2-s2.0-78751677465

10.1109/TVT.2010.2085022

11.

Balasubramanian

Mahajan

Venkataramani

Interactive Wi-Fi connectivity for moving vehicles

Proceedings of ACM SIGCOMM

August 2008

427 438

12.

Liu

Lou

Chen

Cao

The sharing at roadside: vehicular content distribution using parked vehicles

Proceedings of the IEEE INFOCOM

March 2012

2641 2645

13.

Liu

Lou

Chen

Cao

PVA in VANETs: stopped cars are not silent

Proceedings of the IEEE INFOCOM

April 2011

Shanghai, China

431 435

2-s2.0-79960855955

10.1109/INFCOM.2011.5935198

14.

Zhou

Zhang

Song

Jing

Vasilakos

A. V.

Distributed media services in P2P-based vehicular networks

IEEE Transactions on Vehicular Technology 2011 60 2 692 703

2-s2.0-79951901852

10.1109/TVT.2010.2102782

15.

Abumansoor

Boukerche

A secure cooperative approach for nonline-of-sight location verification in VANET

IEEE Transactions on Vehicular Technology 2012 61 1 275 285

2-s2.0-84856187703

10.1109/TVT.2011.2174465

16.

Cheung

M. H.

Hou

Wong

V. W. S.

Huang

Dynamic optimal random access for vehicle-to-roadside communications

Proceedings of the IEEE International Conference on Communications (ICC '11)

June 2011

1 6

2-s2.0-80052175522

10.1109/icc.2011.5962703

17.

http://crawdad.cs.dartmouth.edu/data.php

18.

Saleet

Langar

Naik

Boutaba

Nayak

Goel

Intersection-based geographical routing protocol for VANETs: a proposal and analysis

IEEE Transactions on Vehicular Technology 2011 60 9 4560 4574

2-s2.0-83655191345

10.1109/TVT.2011.2173510

19.

Hassan

M. I.

H. L.

Sakurai

Andrew

L. L. H.

Effect of retransmissions on the performance of the IEEE 802.11 MAC protocol for DSRC

IEEE Transactions on Vehicular Technology 2012 61 1 22 34

2-s2.0-84856170113

10.1109/TVT.2011.2172964

Cooperative Downloading by Multivehicles in Urban VANET

Abstract

1. Introduction

2. Related Work

3. System Problem Formulation

3.1. Problems and Assumption

3.2. Channel Model

4. Select Cooperative Vehicles by SSRF

4.1. Selection of Candidates

4.2. SSRF

Definition 1.

Algorithm 1: SSRF algorithm.

5. Simulation Analysis

6. Conclusion

Footnotes

Conflict of Interests

Acknowledgments

References