2.1

Network discovery

Vertical handover process can be roughly divided into three phases: network discovery phase, network selection phase and network handover phase⁶. In network discovery phase, vehicle scans the surrounding networks to find the network that can be used. Vehicles with multiple network interfaces must activate these interfaces to obtain the surrounding available network information. If interfaces are always active, it will consume a lot of terminal power. Therefore, for Internet of vehicles users, timely discovery of available networks and reduction of energy consumption of vehicle terminals are two important aspects to be considered. In this paper, we proposed a dynamic scanning mechanism based on vehicle speed. We determine the scanning period as

where T_min is the minimum scanning period, which is set as 1s in this paper. T_max is the maximum scanning period, which is set as 9s in this paper. V is the current vehicle speed, V_max is the maximum speed of the vehicle, which is set as 120km/h in this paper.

2.2

Network selection

Network selection phase is the second phase of vertical handover process. In this phase, by analyzing the surrounding network information, the most suitable network for Internet of vehicle users is selected. A good handover decision algorithm can determine the most appropriate handover time, avoid unnecessary handover, and improve the user’s quality of service.

Received signal strength (RSS) is an important evaluation basis of wireless network link quality; It is also an important judgment standard that users need to consider in the process of vertical handover. If the vehicle terminal can know the RSS of current access network in advance, the vehicle can determine the best switching time and switch to a better network to prevent call drop, switching failure, etc. In this paper, we use grey prediction algorithm to predict the RSS of the current access network.

2.3

Grey prediction method to predicts RSS in Internet of Vehicle

GM (1,1) prediction model is widely applied in grey theory system. The model can well predict the future signal strength, and the steps are as follows:

By using the grey prediction algorithm, Internet of vehicles users can determine the best time to leave the current network, so as to avoid the problems such as call drop caused by not switching in time. Next, the decision method needs to be used to determine which candidate network to switch to. This paper selects Multi-Attribute decision making (MADM) algorithm, which is simple, efficient and widely used in vertical handover algorithm⁷.

2.4

Analytic hierarchy process (AHP) to determine the attribute weight

The idea of AHP method is to decompose the complex decision-making problem into sub-problem hierarchies that are easier to understand⁸. The typical hierarchical structure model for AHP is shown in Figure 1. According to the requirements of Internet of vehicle users for the network, delay (D), bandwidth (B), packet loss rate (P) and price (C) are selected in this paper. Assuming that the Internet of vehicle user is running the information business type, this type of service has high requirement for price while low requirement for delay and bandwidth.

Figure 1.

Hierarchical structure model for AHP.

According to the steps of the AHP method, the corresponding weight vector can be obtained as:

2.5

TOPSIS method to select the network

TOPSIS method is an MADM method that use the idea of finding the candidate network that is closest to the positive ideal network, and furthest from the negative ideal network. After getting the attribute weights by AHP method, the weight vector is used in TOPSIS method to evaluate the most appropriate network for the current service type, and the best network is selected.

3.1

Simulation scenario

We verify the feasibility and accuracy of the proposed algorithm on OPNET simulation platform. The simulation scenario consists of two LTE networks and two WLAN networks. The overlapping coverage scenario is shown in Figure 2. LTE 1 covers the whole simulation area with a coverage radius of 3000m. The coverage radius of LTE 2 is 1000m. The coverage radius of WLAN 1 is 300m and that of WLAN 2 is 500m. Internet of vehicle user cross these four networks according to the trajectory shown in the figure. Assuming that the current business type of the vehicle is information service, the vehicle moves at the speed of 30km/h before point A, after point A, it moves at the speed of 60km/h.

Figure 2.

Simulation scenario.

The parameters of the networks are set as shown in Table 1.

Table 1.

Parameters of the candidate networks.

3.2

Analysis of simulation results

In this paper, we use the activate times of terminal interface to represent terminal power consumption, and compares the dynamic scanning period method with the fixed scanning period of 5s and 7s. The results shown in Figure 3 are the comparison of interface activate times when the vehicle moves at variable speed shown in Figure 2. Figure 4 shows the comparison of network discovery time under three scanning methods. We can see from the result that the dynamic scanning period method has less interface activate times than the fixed scanning period of 5s. Although it has more interface activate times than the 7s scanning period, it performs significantly better in network discovery time than 7s scanning period. Therefore, the method based on dynamic scanning period proposed in this paper achieves the balance between discovering the network in time and reducing the activate times of terminal interface.

Figure 3.

Interface activate times.

Figure 4.

Network discovery time.

The simulation results in Figure 5 compare the network selection results during the whole simulation period when the vehicle uses the grey prediction algorithm and does not use the prediction algorithm. We can see from the figure that after using the Grey prediction algorithm, the vehicle terminal can predict the received signal strength of current access network effectively, and switch to a better network in advance. At the same time, due to the application of prediction algorithm, the vehicle terminal can prepare to switch in advance, and the handover call drop rate is improved. Figure 6 shows the simulation result.

Figure 5.

Network selection result.

Figure 6.

Handover call drop rate.