Figure 6. The position allocation scenario.
For the first V2I part, in our simulation environment, 4 kinds of entities are 20 vehicles, 5 RSUs, 20 corresponding nodes, and an ENodeB cellular infrastructure. Figure 6 depicts the vehicles position al-location and Table 5 depicts the used configuration parameters. For the 20 vehicles, every vehicle has a unique corresponding node (CN) that is communicating with it. For the 5 RSUs, the main task of them is to help forward/download packets to/from the Internet. Each of the RSUs also has two interfaces, i.e., one IEEE 802.11p network interface, which is to interact with vehicles and the other one is the tra-ditional wired network interface, which is to connect to the Internet. The two interfaces of each RSU are bridged together such that it can forward data from one to the other one. In the part of LTE infra-structure of the cellular network, two components inside are ENodeB and PDN Gateway. ENodeB is the base station for signal sending and receiving; PDN gateway is for packet routing. ENodeB con-nects with the PDN gateway, and the PDN gateway concon-nects to the Internet. All of the simulation envi-ronment is under the LTE signal coverage.
Figure 7-(a) and (b) have the average throughput comparison of using the proposed scheme with 3 dif-ferent thresholds and the naïve scheme. It can be observed that the proposed scheme with 3 difdif-ferent thresholds work better than the naïve one. For different thresholds set in the proposed scheme, the situ-ation of threshold 0.3/0.6 works better than the situsitu-ation of threshold 0.6/0.9. The reason is that the proposed scheme has some control to allow vehicles using the RSU and the naïve one doesn’t have;
the proposed scheme with lower threshold allows fewer vehicles to enter into the RSU such that the network quality of the RSU can be better.
Figure 7. Comparison of the average throughput of the naïve scheme and the proposed control
scheme with threshold 0.3, 0.6, and 0.9 using the 1Mbps sending rate.
For the Figure 8 is the cumulative traffic data of using RSUs. The red lines indicate the IEEE802.11p cumulative data. The Figure 8-(a) represents the naïve scheme and the red line only reach 1300 MB in the end. Conversely, the Figure 8-(b) represents the proposed control scheme and the red line almost reach the 2300 MB cumulative data traffic finally. Obviously, the proposed control scheme can deliver more data through IEEE 802.11p that means offloading more data from cellular network to IEEE 802.11p network.
(a) (b)
Figure 8. Cumulative data of using RSUs, (a) the naïve scheme using the 1Mbps sending rate (b) the proposed control scheme with threshold 0.3 using the 1Mbps sending rate.
The Figure 9 indicates the packet loss rate of three different threshold proposed control scheme work versus naïve scheme. The naïve scheme is up to 48.5% and the proposed control scheme is only 6.3%.
The efficiency of the packet delivery improves a lot for our proposed work due to the effective net-work quality evaluation. The packets do not congest in the specific RSU. That is the reason of pro-posed control scheme works fine.
Figure 9. Average packet loss rate of IEEE 802.11p network packet loss rate using 1Mbps sending rate.
Figures 10 and 11 show the comparison of throughput and goodput between (i) the Wi Fi infrastruc-ture-based Wi Fi offloading and (ii) the D2D–based Wi Fi offloading using the proposed D2D offload-ing scheme respectively. There are 8 nodes, i.e., 4 pairs, and each transmitter mobile node transmits 500 packets. Figure 10 shows the comparison of the variety of the throughput during the transmission period. Both of the proposed D2D mode's throughput and the infrastructure mode's throughput sharply increase at the beginning. It is because most of the mobile nodes count backward the random backoff number to process channel access, both modes have very low collision rates at the beginning. The re-transmission causes by buffer full makes the contention windows of the infrastructure mode bigger than that of using the proposed scheme. Thus, the throughput of the infrastructure mode is lower than that of using the D2D mode. The D2D mode finishes their transmission at about 0.63th second, it is earlier than the infrastructure mode, i.e., at about 1.53rd second.
Figure 10. The comparison of the average goodput in the AP of both modes.
Figure 11. The comparison of the variety of the throughput during the transmission period.
Figure 12 shows the comparison of the collision rate in the used channel. Figure 12 shows that the lision rate is increased when the number of mobile nodes is increased in both modes. However, the col-lision rate using the proposed D2D scheme is always lower than that of using the infrastructure mode.
The reason is the same as that for the goodput comparison. That is, each paired mobile nodes only need to compete the channel access once for each data's transmission using our proposed D2D scheme and it needs to compete the channel access twice for each data's transmission using the Wi Fi infra-structure-based offloading.
Figure 12. The comparison of the collision rate in the used channel of both modes.