Modern vehicles are equipped with detection technologies like ultrasonic sensors used for parking assistance systems, video cameras employed to monitor the lane or detect pedestrians, and radars used to detect and measure the distance from a vehicle to nearby obstacles [Cai+14]. However, the proper performance of these detection technologies can be a ected by natural factors such as snow, rain, and non-line of sight, which are very common in vehicular environments. Fortunately, these problems can be overcome with vehicleto-vehicle (V2V) communication.

V2V communication o ers a platform for the deployment of cooperative road safety and tra c e ciency applications. The goal of safety applications is to alert drivers about potentially hazardous situations with su cient time to take proper actions. Road safety can be increased by means of periodic exchange of status messages, called \beacons" [ETS14], which contain data such as the position, speed, acceleration, and direction of transmitting vehicle, among others. With the information provided by the beacons, the vehicles create a map of their surroundings, which is used by safety applications for a variety of purposes.

Drivers are vulnerable to tra c in intersections. Without a clear map of the vehicles located in the nearby area, they may be not aware of the danger coming from vehicles driving in the perpendicular direction, resulting in a high possibility of car crashes [GCG17]. In this urban scenarios, the line of sight between vehicles is often a ected by obstacles such as buildings, and parked or moving vehicles [SED14]. At the moment when the vehicles have communication or line of sight between them, it might be very late because due to the speed of the vehicle, the safety distance could be surpassed and an accident could take place. At intersections, the e ect of shadowing caused by buildings a ects drastically the communication range of vehicles. This issue impacts negatively on the capacity of road safety systems to detect neighbours that approach to the intersection [MKH11]. In [SED14], the authors examined the use of parked vehicles as relay nodes for improving cooperative awareness and road tra c safety in urban and suburban environments. The authors showed the use of parked vehicles as relay nodes to be e ective and to improve the cooperative awareness among all nearby nodes if the node density is high. This, however, requires both a high tra c density and a high percentage of equipped vehicles (i.e., substantial market penetration of DSRC devices). A possible solution to the low percentage of DSRC-equipped vehicles, as well as the shadowing problem at the intersection, is to use dedicated RoadSide Units (RSUs) as relay nodes. Placing an RSU at a strategic position can strengthen the connectivity between vehicles traveling on perpendicular roads of an obstructed intersection (see Figure 1). In this work, we analyze how an RSU can improve connectivity at urban intersections blocked by a building. We use a realistic simulation framework to evaluate the connectivity provided by the system in terms of the following performance metrics: noti cation position, number of beacons received, and link life time. We focus on the worst case scenario, which considers that a building totally obstructs the communication link between vehicles. Further, we study the impact of using di erent values of the path loss exponent for direct line of connectivity. We aim at demonstrating that, by improving connectivity, we also improve the operation conditions of road-safety applications, hence, providing an increase in road safety for the participant vehicles. 2

We conducted the experiments using the Veins framework [SGD11], which bidirectionally couples the OMNeT++ network simulator and the SUMO road tra c simulator. We have employed as test scenario an urban intersection located on the following streets: Beauchef with Blanco Encalada, in Santiago, Chile.

This intersection is obstructed by a building, which can be noted inspecting Fig. 2. The scenario seen from Google Earth and SUMO tra c simulator is illustrated in Fig. 2a and Fig. 2b, respectively. At this intersection, we have designed two experiments. In the rst one, two vehicles are moving on perpendicular roads blocked by a building, as shown in Fig. 2c. In the second one, we placed an RSU on the intersection to retransmit the beacons received from Node 1, as shown in Fig. 2d. In these experiments only Node 1 regularly broadcast beacons. The idea is to evaluate the capacity of Node 1 to notify its presence to Node 0 without and with an RSU. Both vehicles and RSU employ the IEEE 802.11p EDCA model [ES12] of the Veins framework to represent the MAC/PHY layer. This is an open source implementation, which fully captures the distinctive properties of IEEE 802.11p radio access technology. Node 1 broadcast beacons to the communication channel using a rate of 10 beacon/s. The radio signal propagation is simulated using the Simple-Obstacle Shadowing model [Som+11]. With the Simple-Obstacle Shadowing model, the building obstructs the communication link between vehicles until they arrive at intersection.

The path loss exponent values = f2:8; 3:0; 3:5g were selected according to [Fer+14]. The communications are established on control channel (CCH) without considering the e ect caused by multi-channel operation. The beacons' size is 250 bytes and are transmitted with a priority corresponding to voice access category (AC VO). Each vehicle is 5 m long, 2 m wide, and has maximum acceleration of 0.8 m/s2, maximum (a) (c) (b) (d) deceleration of 4.5 m/s2, and maximum speed of 50 km/h. The antenna height of vehicles is 1.5 m, whereas the height of the dedicate RSU is 2.2 m. Table 1 includes the additional simulation parameters. 3

In order to evaluate the connectivity provided by the system without and with RSU, we computed in Node 0 the following metrics: - Number of Beacons Received (NBR): The NBR is directly related with the knowledge collected by the vehicle when approaching to the intersection. - Beacon reception time (BRT): The BRT register the times at which beacons are received for the duration of the link. - Link Life Time (LLT): The LLT related with BRT; it and corresponds to the time duration for a link between two vehicles.

Negative values of the noti cation position mean that the vehicle has not arrived at the intersection, whereas zero means that the node is at the intersection, and positive values mean that the vehicle passed the intersection.

Figure 3 shows the potential bene ts of using an RSU on obstructed intersection in terms of NBR by Node 0. Without an RSU, the building blocks direct transmissions between vehicles, reducing drastically their communication range. In this situation, the connectivity is only possible when the vehicles are very close from intersection. In fact, the Node 0 is aware of the presence of Node 1 when it is located at 15.61 m from the crossing point. This noti cation position may not be su cient to react to potentially dangerous situations. Note that the path loss exponent does not have an impact in this case. However, the noti cation position increases signi cantly with the presence of an RSU. Fig. 3a shows that Node 0 receives the rst beacon from Node 1 when it is located at 155.43 m from the crossing point for = 2:8. Moreover, with a poor reception (i.e., =3.5) the noti cation position is still twice in comparison to the situation without the RSU, as shown in Fig. 3c.

Figure 4 shows the BRT computed by Node 0, which implicitly includes the LLT of the vehicles while approaching the obstructed intersection. The RSU signi cantly increases LLT and reduces the distance between the point the rst beacon was received respect to the intersection, especially for lower path loss exponents. Fig. 4a shows an increase in LLT of 17 s when the RSU is used as a relay node for = 2:8. With the most aggressive path loss exponent = 3:5, the RSU still provides a gain of 5 s in LLT, and the rst noti cation is realized 3 s in advance. This time of anticipation is vital for the performance of vehicle collision avoidance systems, which need to alert drivers with su cient time and distance to take proper actions. Table 2 shows a summary of the metrics studied for the di erent path loss exponents. As future work, we will aim to present a speci c relay strategy scheme. Intelligent algorithms need to be applied in these solutions to control the moment that the RSU should participate as a relay node or stay silent for safety applications at intersection. De ne an e cient procolo to enable the relays in the obstruction zones, is not only permitted improvement the use and channel load but to the the bene ts of improving communication distance and link life time for the security applications and so these parameters of the o ered reaction extends to the drivers or the vehicles directly.

We plan to evaluate the performance of the radio signal propagation in obstructed intersections with a high density of vehicles. Experimental validation in real conditions will be carried out by installing OBUs in vehicles. The experimental results will be compared with data obtained by simulations.

Path Loss exponent α = 3.0 Path Loss exponent α = 2.8 With RSU

Without RSU Path Loss exponent α = 2.8 With RSU

Without RSU

This work has been partially funded by ERANETLAC ELAC2015/T10-0761.