Proper Handover Between VANET and Cellular Network Improves Internet Access

Transcription

1 Proper Handover Between VANET and Cellular Network Improves Internet Access Shumin Bi, Cailian Chen, Rong Du, inping Guan Key Laboratory of System Control and Information Processing, Ministry of Education of China and Department of Automation, Shanghai Jiao Tong University, P.R. China {luolan-8, cailianchen, durong, Abstract There are currently several ways of wireless access to support vehicular communication, including vehicular ad-hoc networks (VANETs) and cellular networks (off-the-shelf 3G and LTE). It is necessary to make a seamless handover decision to guarantee quality of service (QoS) of communications for a vehicle moving in the regions covered by more than one access networks. In this paper, we provide a performance guaranteed optimized handover decision algorithm. With this algorithm, the communication of vehicles can handover through heterogeneous wireless access networks not only to reach overall load balance among all access points, but also to maximize the data rate of the whole networks as well as the vehicles fairness. In addition, in the process of decision making, the data rate of handover vehicles are estimated. Simulations are performed to demonstrate the efficiency of the proposed algorithm. I. INTRODUCTION With the increasing demands for drive-thru internet connections and safe driving, intelligent transportation systems [1] are developed to support wireless communications. Vehicular ad hoc networks (VANETs) formed by vehicles which are equipped with sensors and communication devices have attracted the interest of network operators and service providers for the provision of safety and emergency services, as well as informational and entertainment applications [2]. The Wireless Access in Vehicular Environments (WAVE) protocols (IEEE 82.11p/169) [3] provide interoperability between wireless devices on board of vehicles (OBU) and infrastructures located near the roads (RSU). Thus, vehicleto-vehicle (V2V) communication and vehicle-to-infrastructure (V2I) communication can be established in the vehicular network. The performance of IEEE DCF in the highly mobile vehicular networks has been investigated in [4]. And, the downlink scheduling performance of an IEEE based roadside access point serving a number of moving vehicles has been analysed in [5]. Both of them study the easy internet access for a vehicle in single network. However, RSUs could not be available to provide the ubiquitous internet access to vehicles. The high-rate ubiquitous RSU connections is not practical or very expensive due to the huge deployment and maintenance cost [6]. In order to ensure vehicles to access network at will, even in places uncovered by RSUs, existing radio access networks such as cellular networks (off-the-shelf 3G and LTE) and WiFi may be employed to enhance vehicular communications. The potential impact of heterogeneous wireless networks has been confirmed by an ever-increasing amount of mobile internet traffic, which can not solely be absorbed by cellular data communication networks [8]. Then, we can form vehicular heterogeneous networks which can be combinations of VANETs and cellular networks for vehicular communications. As same as VANETs, cellular networks can provide services for vehicles. When internet connection needs to migrate between heterogeneous networks for performance and high-availability reasons, seamless handover is a necessary first step. With the seamless handover decision, the continuous services can be supported and quality of service (QoS) of communications can be guaranteed for a vehicle in region covered by more than one access networks. How to make handover decision is a challenging problem. It requires that demands of vehicles should be satisfied while the overall network performance is optimized. Several interworking mechanisms for combining WLANs and cellular networks into integrated wireless environments have been proposed. Nasser [9] proposed a vertical handoff decision (VHD) method that simply estimates the service quality for available networks and selects the network with the best quality. As it is shown in [9], the known vertical handoff algorithms are not adequate in coordinating the QoS of many individual mobile vehicles or adapting to newly emerging performance requirements for handoff and changing network status. So, Ref. [11] proposed the VHD controllers (VHDCs) to tackle this problem. In the process of handoff decision making in [11] the obtained data rate of vehicles in handoff is assumed to be constant. However, as we known, the most important characteristic of VANET is its inherent high mobility. The current traffic status has a great impact on the achievable data rate. Thus, we can not just make an assumption of the data rate. It is necessary to estimate the data rate through theoretical analysis. In our work, we provide a controller to operate an algorithm which guarantees the performance of optimized handover decision. The controller collects realtime traffic information, then it informs vehicle of an appropriate access point through solving an joint optimization. The optimization is a welldefined objective function which includes consideration of the data rate overall networks and load balancing across access points. Each vehicle s demand should be satisfied to ensure their fairness to a certain extent. Additionally, the gained data rate of each handover vehicle from VANET is theoretically analysed to solve the optimization. Our main contributions are summarized as follows: With the application of the vehicular heterogeneous network, we 1) design a performance guaranteed optimized handover decision algorithm to realize the choice of access point. 2) theoretical analyse the data rate that vehicle can get in VANET. Based on the above two steps, we can improve the internet access with a vehicular heterogeneous wireless network. The rest of this paper is organized as follows. Section II describes our heterogeneous wireless networking system model. Section III describes the details of optimization algorithms to select an appropriate access point. In section IV, simulation results are presented. Conclusions are stated in section V.

2 II. SYSTEM DESCRIPTION The considered framework for the study of handover is shown in Fig. 1. We consider a section of Shanghai-Jinshan expressway (S4 A4) in Shanghai, China. The highway runs through our campus (Shanghai Jiao Tong University). We introduce the following three core components of the framework to highlight the characteristics for internet access. 1) Infrastructure: In the framework, the access infrastructure consists of RSUs and BSs. The BS covers the whole campus, including the section of highway. RSU is equipped with a wireless transceiver operating on Dedicated short Range Communication (DSRC) which is 75MHz radio spectrum in the 5.9GHz band allocated by the United States Federal Communication Commission (FCC). The transmission range of RSU is much smaller than BS. There is a gap between the coverage of two adjacent RSUs. RSUs and BSs are connected to different routers which are combined through the controller to get internet services. 2) Vehicle:Vehicles are equipped with both DSRC and cellular communication hardware. They can connect to VANET infrastructure (RSU) and other vehicles through DSRC and connect to cellular network (BS) through off-the-self 3G/LTE. For simplification, the vehicle to vehicle communications are not considered. Although the communication link between vehicles and BS is more stable than between vehicles and RSU, vehicles would like to use much cheaper and higher rate DSRC communication. 3) Controller: The controller is connected to BSs, RSUs and Internet backbone. It allocates the network radio resources and services demands requested to vehicles based on the realtime road traffic information. And, we do not specific analyse the process of traffic information uploading and downloading. Based on the above framework, a vehicular heterogeneous wireless network which includes VANET and cellular network is formed. Vehicles that move in the highway may be covered by more than one access network. Due to vehicular mobility and characteristics of networks, vehicles may experience weak or degrade received signal strength (RSS) from their current access point (RSU or BS). If any vehicle wants to get service or maintain an underway service, an appropriate access point (RSU or BS) needs to be chosen. This introduces a problem about handover. With a proper handover, the continuous service and QoS experience of vehicle can be significantly enhanced. We consider following two cases of network selection: 1) While in service at an RSU, the RSS for vehicle has dropped below a specified threshold (e.g., car A in Fig. 1). 2) While in service at a BS, the RSS from the next RSU has just exceeded a specified threshold (e.g., car B in Fig. 1). Then, how can the controller find appropriate access points for vehicles. After the collection of real time road traffic information, the controller optimizes the well-defined joint objective function. Then, proper access points are found and the performance of internet access is improved with the heterogeneous network. In the next section, a performance metric is introduced to show the weight of the two criteria of data rate and load balance on the handover decision making and show the calculation of the data rate vehicle can get. III. ANALYTICAL MODEL A. Problem Formulation In this section, the details of the optimization techniques used in our handover decision algorithm are provided. We define the type of networks (RSU and BS) as the network k (k = 1, 2), respectively. Let A = {a 1,, a M } and Fig. 1. The considered framework for the study of handover. B = {b 1,, b N } be the sets of RSUs and BSs covering the considered road sections. In the sections, V = {u 1,, u K } is defined as the set of all vehicles. Each vehicle is either requesting a handover (or just turned on) or is currently serviced by an RSU or BS at one time, with no need for mobility at the time of optimization decision. Thus, the set V can be divided into two subsets at a certain time. One is U = {u 1, u 2,, u j,, u m }, where m is the number of vehicles requesting handover currently, and j, 1 j m is the corresponding indexes of those vehicles. The other is U = V U which means the set of vehicles that have a good connection (i.e.,not requiring handover) to an RSU or BS. If all vehicles assess to the same network, it may cause the congestion of this network. To prevent this case from happening, we assume that network k has maximum bandwidths B k. The prices or weights for the bandwidths of network k is predefined as w k. It is known that an RSU is able to maintain the average bit rate information for vehicles that are currently associated with it [1]. Thus, each RSU a i (1 i M) or BS b i (M + 1 i M + N) can maintain the effective data rates e a ij and eb ij for vehicle u j that belongs to Ū a or Ū b, respectively. However, we haven t the data rate information of vehicle which absents active signal from RSUs, such as u j U. To solve the optimization, r kj which is defined as the data rate access points k can provide should be estimate for u j U. The estimation of r kj is shown in the next subsection. To formulate the optimal handover decision, a binary variable x kj is defined to denote which network vehicle u j connects to. That is, { 1, x kj =, vehicle u j connect to the network k otherwise Then, we define an association matrix consisting of x kj as follows: Definition 1: Let = {x kj } 2 K be an association matrix for the road covered by RSUs and/or BSs, such that we have: 2 k=1 x kj 1, for 1 j K, x kj {, 1} (2) (1)

3 x kj (1 k 2, 1 j K) are binary indicators, each of which has a value of one when vehicle u j hands off to RSU or BS. Thus, a vehicle either connects and only connects to RSU ( or BS) or not connects to both of them, so we have Eq. (2). Moreover, let be the set of all association matrices. Definition 2: Let γ k (1 k 2) denote the total provided data rate of RSU a i (1 i M) and BS b i (M + 1 i M + N). Let r kj denote the data rate provided to vehicle u j (1 j K) by network k. Then, for any = {x kj } Fig. 2. Spacial division for the region between two consecutive RSUs. γ k () = u j U r kjx kj (3) In case that each RSU a i (1 i M) and BS b i (M + 1 i M +N) can maintain the effective data rate e a ij and eb ij for vehicle u j Ū, we can get data rate for the whole network. R() = 2 k=1 (γ k() + u j Ū e kjx kj ) (4) where e kj denote the known data rate for vehicle u j Ū, e 1j = M i=1 ea ij and e 2j = M+N i=m+1 eb ij. We formulate the vertical handover decision problem to maximize the data rate (the whole network) as follows: max s.t. R() γ k () + u j Ū e kjx kj B k r kj x kj c j, u j U 2 k=1 x kj 1, u j V x kj {, 1} where the first constraint in Eq. (5) ensures that the total load on access point can not exceed the maximum bandwidth supported by RSUs and BSs. We define c j as the data rate demand of the vehicle with handover, then the second constraint means the reasonable demands for data rate gained from network k should be meet for handover vehicles. The last two constraints define the association matrix, as shown in Eq. (2) and Eq. (1), and = {x kj } 2 K. In order to prevent too many vehicles from accessing the same network which has a better transmission, vehicles should pay some different prices when they serve through different kinds of networks. We define w(k)e kj x kj /B k as the cost function of vehicle u j from the network k. Thus, we have the overall cost function as 2 k=1 w(k)((γ k() + u j Ū e kjx kj )/B k ) and formulate the following optimization for the distributedness of the load. min 2 k=1 (5) γ k ()+ e kj x kj u j Ū w(k)( B k ) (6) with the constraints of Eq. (5). Minimizing Eq. (6) results in preventing the RSUs and BSs with already higher load from being more congested. To accomplish a joint optimization of the total data rates and the fairness of load in the region we considered, we formulate a combined function F with parameters a and b as follows: γ k () + e kj x kj 2 u j Ū F (, a, b) = ar() + b/ w(k)( ) B k k=1 (7) Minimizing the function in Eq. (6) is equivalent to maximizing the reciprocal of the same cost function. Parameters a and b could be tuned to represent only data rate over the whole network or only load balancing across all RSUs/BSs or a weighted combination of the two. Thus, we have the joint optimization statement of the total data rates and the fairness of the load as follows: max with the constraints of Eq. (5). F (, a, b) (8) B. Calculation of the Provided Data Rate by RSU 1) Mobility Model: Consider the network depicted in Fig. 1. For simplicity, we assume that the inter-distance of two RSUs is L. Let the first RSU (RSU1) be the coordination reference and the second RSU (RSU2) be the located at distance L from RSU1. We divide the range [, L] in smaller segments Z = {1, 2,, Z}, where each zone z Z is of length d z as shown in Fig. 2. A vehicle that moves along the highway iteratively transits through the segments. Thus, we model the mobility of the vehicle using a Markov chain model, inspired by [4], where the states correspond to the different segments in [, L]. Nonetheless, in our model, we define the spacial zones as segments placed between two adjacent RSUs. The mobility of vehicles is represented by the zone transitions using a Markov chain model as shown in Fig. 3 in which each state corresponds to one spatial zone. The time that vehicles stay in each zone z Z is considered to be geometrically distributed with mean t z, which is determined by the length of the partition zone and the average velocity, v, of vehicles as t z = d z /v. Thus, in a small duration, e.g.,, the vehicle either move to the next zone with probability /t z, or remain in the current zone with probability 1 /t z. The set of vehicles in the section of road is represented by V = {u 1,, u K }. Given the transition probability matrix P, the steady state probability matrix of vehicle u j V can be represented by π j = {π j (1), π j (2),, π j (z), π j (Z)} of the Markov chain, in which π j (z) means the probability that vehicle u j V stays at the segment z Z. We can derive π j by solving the following set of linear equations: { πj P = π j Z z=1 π (9) j(z) = 1 The number of vehicles in segment z is defined as n z, and n z = K j=1 π j(z). (1) 2) Access Connectivity Probability: The access probability is the probability that an arbitrary vehicle u j V is directly connected to an RSU. As assumed in many papers [4] [6] [7], vehicles enter the road according to a Poisson distribution, i.e., the distance between vehicles follows an exponential distribution. For any vehicle u j V in zone z Z, we use the distance from the middle point of zone z to the reference

4 Fig. 3. Markov chain model for a vehicle s mobility. point, ω z, to represent the location of vehicle in that zone. According to [7], we have P j (ω z ) = 1 (1 g ς R (ω z))(1 g ς R (L ω z)) (11) where P j (ω z ) is the probability that the vehicle whose distance to the coordination reference is ω z being directly connected to either RSU1 or RSU2; g ς R (ω z) is the probability that a vehicle and RSU separated by distance ω z are directly connected under channel model ς. 3) Packet Collision Probability: The packet collision probability, p col, which is the probability of packet losses due to collisions occurring between two or more vehicles transmitting at the same time. As we known, for VANET, the IEEE 82.11p [3] protocols and Wireless Access in Vehicular Environments (WAVE) technologies have together defined a for vehicular communications. The 82.11p employs carrier sense multiple access with collision avoidance (CSMA/CA) as the fundamental access method to the wireless media. Thus, we can calculate the collision probability of vehicle u j in zone z as follows. p col,j (ω z ) = 1 (1 P j (ω z )) n z 1 (1 P j (ω z )) n z z Z(z),z z (12) where Z denotes the set of zones that fall into the radio range of the vehicle u j and n z is the number of vehicles in zone z calculated by Eq. (1). For simplicity, we assume that if the middle point of the zone falls into the radio range (r) of the vehicle, that zone is considered in Z. Therefore, we have Z(z) = {z ω z r < ω z < ω z + r } (13) The gained effective data rate of vehicle u j from RSU is given by the following. r 1j = P j r j (1 p col,j (ω z )) (14) where r j is the input data rate by the transmitter (RSU). Thus, the gained data rate from RSU is estimated and data rate r 2j provided by BS can be known from experience. Finally, the handover scheme is solved by substitution of r kj from Eq. (14) in Eq. (8). With the estimated data rate r kj, the controller can solve the joint optimization Eq. (8). The proper access points is found for vehicles as well as the improved performance of internet access. IV. PERFORMANCE EVALUATION First, we describe the simulation setup. Then we evaluate the performance of the whole network with different conditions. The provided bit rate from RSU to those vehicles with handover can be calculated using the above analytical results. Additionally, we evaluate the performance of data rate gain from RSU with single vehicle. The provided data rate r 2j from BS will be randomly chosen from 6 1Kbps [11]. A. Network Settings We calculate the analytical results using the derived formulas in previous sections, and conduct the simulation in Matlab. In our simulation scenarios, we just consider the section of S4 covered by two adjacent RSUs which are distributed along the highway with inter-distance of L=125m, and with an effective radio coverage of 5m each. Of course, one BS is able to cover the considered highway. 4 vehicles are dispersed in this road, with radio coverage of 25m each. The bandwidth capacities of RSU and BS are set to 6 and 3Mbps, respectively. We set the weights (prices) associated with the RSU and BS bandwidth usages to w = [1, 1]. To calculate the provided data rate from RSU, the log-normal shadowing parameters are used. The divided small zone is of length d z = 5m. Vehicles average speed range from 6 to 12Km/h, and the data rate by the RSU is set to 1K 2Mbps. B. Results In this section, we present and discuss simulation results. We compare the results under different conditions. The comparisons are presented in terms of the overall system data rate and costs. To better understand the data rate gained from RSU, we evaluate the data rate performance with single vehicle. First, we show the performance of tradeoff and the improved performance of the application of vehicular heterogeneous network. To set different value of a and b in solving the joint optimization problem in Eq. (8) can put different emphases on data rate and cost. When RSU inter-distance L=125, a and b are set as b/a = 1, 1, 5, 1, which are denoted as,, and, respectively. Fig. 4 represents the corresponding data rate and cost under different value of b/a, respectively. Compared Fig. 4(a) with Fig. 4(b), we can see that the more emphasis on cost equal to a higher value of b/a, the lower data rate and lower cost we get. It is because network should have to pay for lower cost, i.e., it performs worst in terms of data rate. For the same graph, we compare the performance with and. Obviously, some handover vehicles can not access to network, the date rate and cost is smaller. However, the cost does not reach 1times compared with. The ratio of the number of satisfied vehicles is much smaller than that with BS as shown in Fig. 4(c). More vehicles are satisfied, the fairness of vehicles is improved. This is an advantage of our system, a vehicular heterogeneous wireless network have more choice for vehicles to access internet. Second, we study the access scheme. The access ratio of the number of vehicles which connect to different networks with different situation is shown in Fig. 5. Overall, the Access to RSU among all vehicles represents the proportion of vehicles that access to RSU among all vehicles. Obviously, the ratio increases with the increasing of b/a. The Not access to network among handovers represents the proportion of vehicles that can not access to network among those handover vehicles. The Access to RSU among handovers represents the proportion of vehicles that access to RSU among those handover vehicles. Compared with this two curves, the value of the ratio of vehicles does not access to network is smaller when the ratio of vehicles access to RSU is smaller. This means a proportion of vehicles which want to handover connect to BS. This is an advantage of the architecture we considered. Different from the traditional system, the heterogeneous network improved the performance of internet access with the cellular network to join. Vehicles can keep continuous communication with network as soon as possible.

5 Data rate (Kb/s) Cost (1 3 ) The ratio of the number of satisfied vehicles (a) (b) (c) Fig. 4. The different performance with different value of b/a. Access ratio Access to RSU of all vehicles Not access to network of handoff vehicles Access to RSU of handoff vehicles Different value of b/a Transmission rate(kb/s) L=75 L=1 L= Average speed(km/h) Transmission rate(kb/s) v=35 v=5 v=65 v= Vehicle density(vpm) Fig. 5. scheme. Comparison of four different association Fig. 6. speeds. Transmission rate for different average Fig. 7. densities. Transmission rate for different vehicle Last, we analyse the performance of data rate gained from RSU when observing single vehicle. From Fig. 6, we can observe that due to the number of vehicles are solid, the larger inter distance of RSU means the less density of vehicles which will results to the bigger data rate under the same speed. In order to analysis how the density impact on the transmission rate, we get the following simulation result. As shown in Fig. 7, it is inevitable that the larger the vehicle density of the road, the lower the transmission rate the vehicle probable get from RSU, it is result from the competition of vehicles in the road. V. CONCLUSION In this paper, we present a vehicular heterogeneous wireless network which is the combination of VANETs and cellular technologies to improve the performance of internet access for vehicles. Under this situation, we provide a method to solve the handover problem which is ubiquitous in heterogenous networks including vehicular heterogeneous wireless network. A performance guaranteed optimal handover decision algorithm is applied at the controller which is used to mange all vehicles connection. In the process of handover decision making, a joint optimization is provided to maximize the data rate of overall networks and keep a load balance across all access point. The data rate demand of each with handover is satisfied. ACKNOWLEDGMENT This work was partially supported by Ministry of Science and Technology of China under grant 21CB73183, by NSF of China under the grants , , and , by Program for New Century Excellent Talents in University, Ministry of Education of China, and by Science and Technology Commission of Shanghai Municipal, China under the grant 13QA1419. REFERENCES [1] United states department of transportation, Intelligent transportation systems, [Online]. Available: [2] H. T. Cheng, H. Shan, and W. Zhuang, Infortainment and road safety sevice support in vehicular networking: From a communication perspective, Mechanical Systems and Signal Processing, 21. [3] Wireless LAN medium access control (MAC) and physical layer (PHY) specifications amendent: Wireless access in vehicular environment, IEEE Std, P82.11p, Jul, 15, 21. [4] T. H. Luan,. Ling, and. Shen, MAC in motion: impact of mobility on the MAC of drive-thru internet, IEEE Trans. Mobile Comput., vol. 11, no. 2, pp , Feb [5] T. Hui, W. C. Lau, and O. Yue, VECADS: Vehicular Context-Aware Downstream Scheduling for Drive-Thru Internet, Proc. IEEE VTC-Fall 212, Sep, 212. [6] Y. Zhuang, J. Pan, V. Viswanathan, and L. Cai, On the uplink MAC performance of a drive-thru internet, IEEE trans. Vehicular Technology, vol. 11, no. 4, pp , Apr [7] S. C. Ng, W. Zhang, Y. Zhang, Y. Yang, and G. Mao, Analysis of access and connectivity probabilities in vehicular relay networks, IEEE J. Sel. Areas Commun., vol. 29, no. 1, pp , Jan [8] J. N. Cao and C. S. Zhang, Seamless and secure communications over heterogeneous wireless networks, Springer, 214. [9] N. Nasser, A. Hasswa, and H. Hassanein, Handoffs in fourth-generation multi-network environment, IEEE Commun. Mag., vol. 44, no. 1, pp , Oct. 26. [1] H. Zhai, J. Wang, and Y. Fang, Providing statistical QoS guarantee for voice over IPad in the IEEE Wireless LANs, IEEE Wireless Commun. Mag., vol. 13, no. 1, pp , Feb. 26. [11] S. Lee, K. Sriram, K. Kim, Y. H. Kim, and N. Golmie, Vertical handoff decision algorithms for providing optimized performance in heterogeneous wireless networks, IEEE trans. Vehicular Technology, vol. 58, no. 2, pp , Feb. 29. [12] S. Céspedes, N. Lu,. Shen, VIP-WAVE: On the feasibility of IP communications in 82.11p vehicular networks, IEEE trans. Intell. Transp. Syst., vol. 14, no. 1, pp , Mar [13] M. Asefi, J. W. Mark, and. Shen, A mobility-aware and qualitydriven retransmisssion limit adaptation scheme for video streaming over VANETs, IEEE Wireless Commun. Mag., vol. 11, no. 5, pp , May. 212.