3.1 Problem discussed
Based on mobile cloud services in the car networking architecture, vehicles request to upload data to the server before the gateway service has been acquired by the surrounding GSW of the relevant information and they are stored in the GSW list. How to select a suitable gateway service provider from the list according to the data type and gateway server’s relevant parameters, and how to quickly upload the data to the cloud are the focuses of this section.
Using communication resources as an example, in the current urban environment there may be massive distribution to service gateway nodes, such as those equipped with the 4G network interface for buses or taxis, or through a fiber-optic cable connected to the Internet in the Roadside Unit (RSU) etc. Through these nodes, the vehicle can pass data to the cloud and can also receive data from the cloud. Some nodes, in addition to providing gateway services, also have the ability to process and analysis the received data.
The following section presents a composite gateway selection vector, which comprehensively considers link stability (LET), network signal strength (RSS), transmission bandwidth (BW), service cost (CC), packet error rate (e
p
) and transmission load (CL) to calculate the vector value. A small vector value is selected as the service GSW. The gateway selection vector calculation method is as follows:
$$ \begin{array}{l}F={a}_1\times \left(1-\frac{LET}{LE{T}_{\max }}\right)+{a}_2\times \left(1-\frac{RSS}{RS{S}_{\max }}\right)+\hfill \\ {}{a}_3\times \left(1-\frac{BW}{B{W}_{\max }}\right)+{a}_4\times \frac{CC}{C{C}_{\max }}+\hfill \\ {}{a}_5\times \frac{e_p}{E_{\max }}+{a}_6\times \frac{CL}{C{L}_{\max }}\hfill \end{array}, $$
(1)
-
(1)
Link stability (LET) LET is a measure of the link stability between the current vehicle and the GWS. Based on vehicle location and travel direction, the speed of the vehicle can be calculated, and it can be used to explain if two vehicles are in the mutual communication range in the next period of time. Suppose the coordinates of i and j of the two vehicles are, respectively, (x
i
, y
i
) and (x
j
, y
j
) and v
i
and v
j
show the speed of the vehicle, θ
i
and θ
j
, respectively, are the slope of the road with respect to the X-axis of the vehicle node, and R is the transmission radius. The formula for calculating LET is then as follows:
$$ LE{T}_{ij}=\frac{-\left( ab+ cd\right)+\sqrt{\left({a}^2+{c}^2\right)R-{\left( ad-bc\right)}^2}}{a^2+{b}^2}, $$
(2)
here,
$$ \begin{array}{l}a={v}_i \cos {\theta}_i-{v}_j \cos {\theta}_j,\hfill \\ {}b={x}_i-{x}_j,\hfill \\ {}c={v}_i \sin {\theta}_i-v \sin {\theta}_j,\hfill \\ {}d={y}_i-{y}_{j.}\hfill \end{array}. $$
-
(2)
Packet error rate e
p
e
p
reflects the quality of V2V communication based on DSRC and can be obtained from the measured signal to noise ratio of the physical layer. By means of Binary Phase-shift Keying (BPSK) modulation, the packet error rate, e
p
, of data transmission by white Gaussian noise channel is as follows:
$$ {e}_p=1-{\left[1-Q\left(\sqrt{\frac{2{P}_r}{R_b\times {N}_0}}\right)\right]}^L, $$
(3)
here, \( Q(x)=\left(1/\sqrt{2\pi}\right){\displaystyle {\int}_x^{\infty }{e}^{-{t}^2/2}dt} \),
in the above, N
0 is the noise energy spectral density, P
r
is the received signal power and R
b
is the base rate of data transmission.
This section mainly relates to the theoretical analysis of GWS access delay and GWC data transmission delay. In order to analyze the performance of this method, the mathematical model of this section is based on the following three points:
-
(1)
The vehicle from the end of the road comes into the road at a speed of v. The number of vehicles entering the road is subject to a Poisson distribution of λ
-
(2)
The vehicles are equipped with a global positioning system and data transceiver equipment
-
(3)
The physical channel is reliable and does not generate errors
TC(G
i
, G
i + 1)(i = 0, 1, ⋯, n) represents the transmission delays of the cloud sending messages to the GWS. TD(G
i
, G
i + 1)(i = 0, 1, ⋯, n) represents the broadcast delay from the GWS sending-beacon to all target area vehicles. T(G
i
, G
i + 1)(i = 0, 1, ⋯, n) indicates that the time delay of the service information is received by all the target vehicles in the subsection. Then:
$$ T\left({G}_i,{G}_{i+1}\right)=TC\left({G}_i,{G}_{i+1}\right)+TD\left({G}_i,{G}_{i+1}\right), $$
(4)
and
$$ \begin{array}{l}{T}_{ac}= \max \Big\{T\left({G}_0,{G}_1\right),T\left({G}_1,{G}_2\right),\\ {}\kern1.5em \cdots, T\left({G}_n,{G}_{n+1}\right)\Big\}.\end{array} $$
(5)
Suppose the random variable x represents the number of vehicles entering the road within the time interval (0, t]. The density function and mathematical expectation of x can then be expressed as follows:
$$ P\left(x=n\right)=\frac{{\left(\lambda l\right)}^n\times {e}^{-\lambda l}}{n!} $$
(6)
$$ E(x)=\lambda l. $$
(7)
If d represents the distance between two adjacent vehicles, then d = vt. If two cars can communicate with each other, then d ≤ R. The conditional probability density function and mathematical expectation of d is then as follows:
$$ P\left(d=k\Big|d\le R\right)=\frac{1-{e}^{-\left(\lambda k/v\right)}}{1-{e}^{-\left(\lambda R/v\right)}} $$
(8)
$$ E(d)=\frac{v}{\lambda }-\frac{R}{e^{\lambda R/v}-1}. $$
(9)
The number s of vehicles in the vehicle transmission distance can be expressed as follows:
$$ E(s)=\Big|R/E(d)\Big|=\left|\frac{R\lambda \left({e}^{\lambda R/v}-1\right)}{v\left({e}^{\lambda R/v}-1\right)-R\lambda}\right|. $$
(10)
The average number of hops for data transmission in a section of l length is as follows:
$$ h(l)=\left|\frac{E(x)-1}{E(s)}\right|=\left|\frac{\lambda l-1}{E(s)}\right|. $$
(11)
Beacon single-hop delay T
b
is defined as the time interval from which the beacon is successfully received and forwarded from the beacon to the receiving node. Because the beacon sends no back-off process and message confirmation, T
b
can be expressed as follows:
$$ {T}_b={T}_{aifs}+\delta +{L}_b/{r}_b, $$
(12)
where T
aifs
is the arbitrary frame interval. δ is the average time when the timer changed to 0, L
b
is the size of beacon and r
b
is the speed of transmission.
$$ TD\left({G}_i,{G}_{i+1}\right)={T}_{aifs}+\delta +{T}_b\times \left|\frac{h\left({l}_{rs}\right)}{2}-1\right|, $$
(13)
where, l
rs
is the length of road section. With GWC as the selected service gateway, the data transmission of a single hop transmission time T
d
can be expressed as follows:
$$ \begin{array}{c}\hfill {T}_d={\displaystyle \sum_{m=0}^{\infty }{e}^m\left(1-e\right)\left[{T}_{sifs}+\omega +\delta +{L}_d/{r}_b+\right.}\hfill \\ {}\hfill \left.m\times \left({T}_{sifs}+\omega +\delta +{L}_d/{r}_b\right)\right]\hfill \end{array}, $$
(14)
where, T
sifs
stands for the short frame interval, L
d
is the transmission data size and ω is the time of the back-off process.
$$ \omega ={\displaystyle \sum_{j=0}^{CW\left[3\right]}\left(\omega \Big|j\right)\frac{1}{CW\left[3\right]+1}} $$
(15)
$$ \omega \Big|j=\left\{\begin{array}{c}\hfill {\displaystyle \sum_{k=1}^j{\overline{Y}}_k},j\in \left[1,CW\left[3\right]\right]\hfill \\ {}\hfill 0,\kern5.25em j=0\hfill \end{array}\right., $$
(16)
where, CW[3] represents the contention window size of the third types of channel access type messages, Y
k
represents the time length of the first time slices in CW[AC], AC is the channel access type and \( {\overline{Y}}_k \) is the average of Y
k
.
Suppose the distance from the vehicle to the service gateway is l
c
, then:
$$ \begin{array}{l}{T}_{tr}={\displaystyle \sum_{m=0}^{\infty }{e}^m\left(1-e\right)\left[{T}_{sifs}+\omega +{L}_d/{r}_b+\right.}\hfill \\ {}m\times \left.\left({T}_{sifs}+\omega +{L}_d/{r}_b\right)\right]+{T}_d\times \left[\frac{h\left({l}_c\right)}{2}-1\right]\hfill \end{array}. $$
(17)
3.2 Experiments
We wrote the simulation program on the OMNeT++ and SUMO platforms. OMNeT++ is mainly a simulated vehicle node communication protocol. SUMO is used to simulate traffic scenarios.
In this paper, through simulation experiments based on data-upload strategy-related properties of mobile cloud services, including GWS information service coverage, channel occupancy rate, access delay, data transmission delay and transmission rate, we propose to verify our theoretical analysis and other related agreements presented herein.
3.3 Service information coverage and channel occupancy
Figure 1 shows the relationship between the coverage rate of GWS service information and traffic density. P is the proportion of GWS accounting for the total number of vehicles.
Because of the delay in receiving service information, this time-lag will cause the service information to be out of date. In the experiment, if the vehicle only receives the service information for which the transmission time delay is more than 5 s, the vehicle is not counted as being in the service range of the vehicle.
As can be seen from Fig. 1, CDUL can guarantee a high coverage rate of service information in the case of different GWS ratios and traffic densities. When the vehicle density is too low, the connectivity of the vehicle is relatively low, and service information delay is too high. At this time the service message coverage is relatively low. However, the cloud can dynamically adjust the number of service information broadcast hops so that even if the GWS ratio is low, it can actually guarantee the service information coverage rate to be greater than 95%.
Figure 1 also shows the use of the traditional beacon broadcasting method of service information coverage. At this point, because all GWS only broadcasts one hop for service information, when the GWS number is too small it is difficult to ensure service coverage.
Figure 2 shows the relationship between channel occupancy rate and vehicle density. It shows that GWS service information does not take up too much of the channel resources under different GWS ratios. This is because the cloud in the choice of service gateway can effectively avoid GWS. This is compared with the channel occupancy rate without the use of CDUL and it is clear that when all the GWS are involved in the service, a large amount of service information broadcasting leads to excessive channel resource occupancy and, especially when the request for gateway service is less, channel resource waste is serious and not conducive to the rapid broadcast of other urgent messages.
3.4 Access delay
GWS access delay is composed of two parts: cloud transmission delay (TC); and service information broadcast delay (TG). TC is related to network type and transmission bandwidth factors. In the experiment, the 4G network environment, which covers a wide range of urban traffic environments, is selected to test the transmission delay of the TC network. One hundred sets of tests were performed at different locations during the test. The measured results are shown in Fig. 3. From Fig. 3 we can see that time delay (TC) is concentrated over 7–15 ms.
Figure 4 shows the relationship between GWS service information broadcast delay (TG) and traffic density.
When the traffic density is low at sections, the gateway number is too small. Information service is needed through multi hop manner to pass to the road of the vehicle. Due to the inter vehicle lower connectivity, the vehicle cannot directly send service information to the surrounding vehicles and work on the carrying forwarding mode, which leads to the traffic density is 0.02 and the access delay is relatively high. If the GWS ratio is 5% at this time, the entire section is only 2 GWS. Access delay is greater than 1s, and it is difficult to ensure that the vehicle quickly receive service information. In addition, from the experimental results, it is shown that the delay of transmitting information from the cloud to the GWS is much longer than that of the GWS to broadcast the service information to the surrounding vehicles. So as far as possible, the choice of some static service gateway can effectively reduce the access delay. From Fig. 4 it can be seen that the traffic density is 0.02 units/meters, and the vehicle can provide information broadcast service through carrying forward. On the whole, the experimental results have a high consistency with the theoretical curve.
3.5 Data transmission delay
Figure 5 shows the GWC data transmission delay under different traffic densities. It can be seen that when the GWC node reaches a certain number, data transmission delay is very low, and in different traffic densities when the transmission delay environment is very stable, the experimental curve and the theoretical curve have a high degree of consistency.
Figure 6 compares the cloud data upload largely (CDUL) with the gateway selection method clustering multi-gateway management (CMGM): (1) in view of the sparse data, we can see that the emergency message transmission delay, which is higher than that of the proposed method in this paper, can be seen from the experiment and that comparisons with CMGM delay time can be made. This is because in the CMGM protocol, a certain range of vehicles passes through a gateway to upload data, resulting in the gateway communication load being too high, whereas the choice of service gateway in CDUL is more dispersed and the communication load of each GWS is balanced. At the same time, the communication quality of the GWS is fully considered and the delay of data transmission is reduced when an emergency message is delivered; (2) in view of the relatively large data, the traffic information transmission delay with higher link stability is compared. It can be seen that when the traffic density is high, the CMGM will lead to high data transmission delay. In the choice of GWS, CDUL is considered for the link stability, communication load and other factors. Such a large data in the transmission of traffic information can ensure the lower transmission delay.
3.6 Data transmission rate
Figure 7 compares the transmission loss rate of CDUL and CMGM under different densities. CMGM will lead to a large number of vehicles in the same range selecting the same gateway, so that the gateway to send buffer queue packets overflows. In addition, if the mobile gateway is too concentrated this will lead to a large number of data transmission collisions. And CDUL in the gateway selection to avoid the GWS too concentrated, while considering the various GWS communication loads in order to avoid a large number of vehicles choosing the same gateway. As can be seen from Fig. 7, CDUL data transmission in the low packet has a loss rate, and with the increase in traffic density, packet loss rate increases slowly.
As can be seen from the experiment, when the GWS density reaches a certain level, the service information broadcast method proposed in this paper has higher coverage rate and lower channel occupancy rate. Compared with the existing methods, the transmission delay and packet loss rate of data upload is lower which is reflected in the very stable performance curve. If GWS density is too low, the performance parameters of the five have a certain effect, but in the current traffic environment it is very easy to meet the demand of GWS density, especially in many city buses and taxis with the deployment of such a large number of 4G network interfaces. A number of private cars have also begun to use such a network interface. It can be predicted that with the development of the car networking and mobile cloud services, some of the static GWS will also begin to be significantly deployed. Additionally, according to the mobile cloud services based on mobile network architecture and using a data-upload method to ensure that reliable data are rapidly uploaded to the cloud, this will not make too high a demand on the channel and communication resources.
