V2V Communication Quality: Measurements in a ... - Semantic Scholar

Downloaded from SAE International by Carl Bergenhem, Monday, March 17, 2014 10:17:16 AM 2014-01-0302 Published 04/01/2014 Copyright © 2014 SAE International doi:10.4271/2014-01-0302 saepcelec.saejournals.org

V2V Communication Quality: Measurements in a Cooperative Automotive Platooning Application Carl Bergenhem

Qamcom Research and Technology AB

Erik Coelingh

Volvo Car Corp.

Rolf Johansson

SP Technical Research Inst of Sweden

Ali Tehrani

Qamcom Research & Technology ABSTRACT This paper presents measurements on Vehicle to Vehicle (V2V) communication between participants in a platooning application. Platooning, according to the SARTRE concept, implies several vehicles travelling together in tight formation, with a manually driven heavy lead vehicle. The platoon being studied consists of five vehicles; two trucks in the lead and three passenger cars. The V2V-communication node in each vehicle contains an 802.11p radio at 5,9 GHz. It is used to send messages between vehicles to coordinate movements and maintain safety in the platoon. Another cooperative application that relies on V2V-communication is multiple UAVs flying in formation; as investigated in KARYON. This project also investigates cooperative autonomous vehicles. In both applications, V2V-communication is an enabling technology. Two metrics are studied to quantify the V2V-communication quality: system packet error rate and consecutive packet loss. These two metrics characterize the communication quality in the different tests (speed, antenna position and two tracks). The paper draws general conclusions on the performance of V2V-communication. The presented test results supports comparison of the tested antenna placements on the trucks and the communication quality related to speed and track.

CITATION: Bergenhem, C., Coelingh, E., Johansson, R., and Tehrani, A., "V2V Communication Quality: Measurements in a Cooperative Automotive Platooning Application," SAE Int. J. Passeng. Cars – Electron. Electr. Syst. 7(2):2014, doi:10.4271/2014-01-0302.

INTRODUCTION In this paper we present measurements on Vehicle to Vehicle (V2V) communication for a cooperative application. In such an application, multiple entities cooperate to solve a common goal such platooning in road vehicles or formation flying in airborne vehicles. Each individual vehicle has local sensors but must also share data from the sensors of other vehicles. This is because local sensors may be e.g. obscured by other vehicles or have limited range. Additionally, commands for coordination, such as joining or emergency brake, need to be exchanged. This is done with V2V-communication which hence becomes an enabling technology for cooperative systems. To design the cooperative function, the reliability and Performance of the V2V-communication must be known. The variability of communication quality should be quantified to understand how it will affect e.g. control algorithms of vehicle movements.

Two examples of cooperative application are SARTRE [1]: vehicle platooning on public roads; and KARYON [2]: Kernelbased architecture for cooperative vehicles promoting efficiency and safety. This paper presents results of tests done on the V2Vcommunication system that was implemented in the SARTREproject. The tests were performed at the APPLUS IDIADA facilities in Spain on multiple combinations of test tracks, platoon speed and antenna placements: Two different test tracks: a high speed oval circuit and a suburban general road track; three different top speeds: 50, 70 and 85 Km/h; and two different antenna placements on the trucks: Dual antennas on the cabin roof and antennas on both rear view mirrors (RVM). In the tests each vehicle communicates and measures the communication quality to each other vehicle. We study two metrics to quantify the V2V-communication quality: system

Downloaded from SAE International by Carl Bergenhem, Monday, March 17, 2014 10:17:16 AM

Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014) packet error rate and consecutive packet loss. These are studied at the system level, i.e. between the pair of communicating nodes. A sending or receiving node will be configured with either one or two radios (with antennas). Hence there are four possible pair configurations, i.e. systems. A measurement of communication quality, expressed in the two metrics, will apply to a particular system, i.e. the configuration of the communicating pair of nodes. Several groups have, in recent years, studied different aspects of V2V-communications using IEEE 802.11p, as is also used here. In [3, 4, 5] communication is characterized in a mix of Line-Of-Sight and Non-Line-Of-Sight conditions focusing on parameters related to the physical layer. Examples of such parameters are channel gain, path loss, coherence time, power delay profile and Doppler spread. In [6] the focus is on channel measurements at street intersections for V2V safety applications. Paper [7] presents range measurements for vehicles driving towards or away from each other. A survey of V2V-communication is done in [8]. It is shown that the channel characteristics vary depending on the type of road: (highway, rural, suburban and urban). This can be explained by the different velocities, surroundings (number of and distance to scatterers differs in open areas versus small cities), traffic intensity etc. All these variations affect the multi path propagation environment for the radio waves which in turn affects the channel response. A measurement campaign that is similar to the one presented here, is found in [9]. Here, experiments involve two V2V-equiped vehicles and one antenna position. Each vehicle communicates with one radio (and antenna). Consecutive packet loss, round trip delay of packets and communication range is also investigated here. Results with measurements and tests of a V2V-system for a platooning application, including a comparison of antenna placement, have been published [10]. A similar V2V-setup as in this paper is used, e.g. the truck communicated with two radios (and antennas). In this paper we contributes further to the understanding of how to design a reliable V2V-system for platooning. More realistic scenarios are used and more configurations are tested in this paper. The rest of the paper is organized as follows: a presentation of cooperative applications which use V2V-communication is presented in the next section. This is followed by a description of the measurement setup. Then the results of the test are presented and finally conclusions and reflections are made. At the time of measurements the main author, Carl Bergenhem, was employed at SP - Technical Research Institute of Sweden.

COOPERATIVE APPLICATIONS Cooperative applications involve multiple entities that cooperate to solve a common goal. The entities can be vehicles and the goal can be a specific task such as platooning (road vehicles) or formation flying (airborne). The cooperative aspect of the application necessitates wireless communication

among the participants of the application. In addition to vehicles, entities that participate in the application may also be road-side units that e.g. provide external information and high-level control. An example is traffic control that provides information on road condition and congestion. The vehicles involved in a cooperative application can be equipped with local sensors which provide data to the application. Through cooperation with other entities external data is also received, see Figure 1. Sensor fusion is used to meld the data from the two sources. Local data is mostly more trustworthy than external. Also external data is attributed different levels of trust depending on the source e.g. from a vehicle that is participating in the application or is unknown. However, external data can describe the environment from a standpoint that is e.g. obscured to local sensors such as the speed of the vehicle in front of the preceding vehicle in a platoon. Local sensors can also be limited e.g. in range or field of vision. Weighting of the data sources according to trust and data quality is typically used in sensor fusion algorithms. External data is sent with V2V-communication from the source, e.g. other participants, and is received at the local vehicle. Data can be periodically sent (time-triggered) or sent only when an event occurs (event-triggered). Two examples of cooperative application are given below; platooning according to SARTRE and formation flying of UAVs as studied in KARYON. A brief survey of other vehicle platooning systems is given in [11].

The SARTRE Platooning Application SARTRE (SAfe Road TRain for the Environment) was a project (2009 - 2012) that developed technology for platooning [1]. It was a European Commission Co-Funded FP7 project that sought to support a step change in transport utilization. The project vision was to develop and integrate solutions that allow vehicles to drive in platoons. SARTRE defined a platoon (or road train) as a collection of vehicles led by a manually driven heavy lead vehicle. The following vehicles (trucks and passenger cars) follow the lead vehicle automatically; both laterally and longitudinally. Vehicles may join or leave the platoon dynamically e.g. leave on arrival at the desired destination. The demonstrator consisted of platooning with five vehicles: Two trucks and three passenger cars. SARTRE aimed to explore technology for platooning on roads without changes to the infrastructure and that was safe enough to allow mixing with other users of public roads. Advantages of platooning included a reduction in fuel consumption, increased safety and increased driver convenience and comfort. The technical challenges in the project were many and interesting such as the design of control algorithm and sensorfusion. Another challenge is the V2V-communication system. The V2V node is a wireless gateway between the network in the local vehicle to the networks in the other vehicles. The V2V-node allows sharing of local vehicle signals such as speed and sensor data among the vehicles in the platoon. The shared


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014) signals are used in the control algorithms of the platoon. The platoon forms a cooperative system where sensing, control algorithm and actuation are distributed throughout the platoon and data is communicated between vehicles (V2V). Automatic control over an individual following vehicle is partly external from the lead vehicle and partly internal from the systems in the following vehicle itself. The following vehicles automatically strive to maintain the specified gap to the vehicle in front, and the trajectory as specified by the lead vehicle. The local systems in the following vehicle can also take over in emergency situations and during loss of communication. The platooning application requires that V2V-communication is used in addition to local sensors in each vehicle. Using V2V-communication implies that data can be sent directly from the source rather being indirectly measured locally with sensors. Detecting platoon movements via only local sensors is prone to lag and to accumulate errors. This is because local sensor measurements are only based on the adjacent vehicle, i.e. there is no “look ahead” e.g. of intended movements. For example the lead vehicle can directly send requested acceleration as measured at the pedal rather than having a following vehicle measure the acceleration with its local sensors. With local vehicle sensors a change in acceleration has to “propagate” through the platoon from the lead vehicle to each of the following vehicles and be detected. This affects, for example, the minimum gap size that can be safely achieved.

according to the condition of data. The on-line reliability of V2V-data is assessed with timestamps. The sender of a message adds a timestamps with the current time. The receiver can hence detect that the signals in the message have not been updated as expected and are hence old and less credible. When V2V-data is not sufficiently updates a fall-back strategy aims to gracefully degrade platoon operation while maintaining safety. A possible strategy can have four operational classes. Each class implies gradual loss of service that depends on the length time since the last update, i.e. due to a consecutive packet loss. This denoted BOP (“black-out” period). 1. If BOP is shorter than 100 ms, nothing is done; 2. If BOP is between 100 ms and shorter than 500 ms, modified control parameters are used based on the last valid V2V data. Safe assumptions are made e.g. about acceleration until V2V communication is resumed; 3. If BOP is between 500 ms and two seconds, the inter vehicle gap is smoothly increased; 4. Eventually, after two seconds BOP dissolution of the platoon is initiated. Platoon dissolution implies manual take over by the drivers in the following vehicles. The fall-back scheme and time parameters given above are examples and reasonable assumptions. Actual times will depend on the control algorithm and environment and need to be investigated in each case. Characterization of V2VPerformance, as in this paper, contributes to the reliability of the platooning system since e.g. control algorithms can be designed appropriately by taking into account the properties of the V2V-Performance.

KARYON Figure 1. An ego vehicle view of the cooperative application

Without V2V-communication, the gap between vehicles has to be larger to allow for the slower response. Using only local sensors can lead to lateral and longitudinal instability, increased oscillations, and unsafe behavior of the platoon. The goals of longitudinal control, e.g. the speed of the following vehicles, are to make coordinated movements that are accurate and adequately safe. Two examples are keeping a fixed gap between vehicles and being able to perform an evasive maneuver such as emergency brake or lane change. Lateral control (steering) has similar goals, solutions and also faces similar challenges. The local and V2V-data is combined by sensor fusion. The default weight of each source is decided offline; The weighting is adapted during operation e.g. with regard to operating conditions such as speed of the platoon. V2V-data is more prone to transient communication failure compared to the local sensor data. Sensor fusion is therefore used and is adaptive

KARYON (Kernel-based ARchitecture for safetY-critical cONtrol) [2] is an on-going project investigating how to solve the paradox of getting cooperative vehicles that are functionally safe without sacrificing Performance or price. It is co-funded by EU FP7 and consists of partners from academia and from automotive and avionic industry, respectively. The scenarios address several possible cooperative functions in the automotive domain (platooning, coordinated lane change, coordinated intersection) and in the avionics domain (coordinated common trajectory traffic, coordinated level crossing trajectories and coordinated flight level change). The problem addressed in KARYON is that proving that cooperative functions are functional safe, for 100% of the mission time might, become very costly and/or restrictive to the functional Performance. The proposed route forward is to move part of the safety assessment from design-time to runtime. Each cooperative function is defined as a number of different levels of Performance - each having an associated Hazard analysis and Risk assessment. This implies that depending on which level of Performance a function is in, different safety


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014) integrity levels are on the top level requirements and hence also different integrity levels required for the signals inside the system architecture. Typically a functional safety standard states how to argue for that the output of a certain component fulfils a certain safety integrity level for a given unwanted failure. The KARYON approach is to partly collect evidence in run-time to find what safety integrity levels that are actually met. This is then used for deciding a matching level of Performance where functional safety is reached. In order to find such evidence redundant sources of information are used. This is extending the scope of data via sensor fusion. The incoming values of data itself and related values of trust and quality are not only used to calculate the fused data value, but also to calculate a safety integrity attribute for each failure to avoid according to the applicable safety requirements. The measured realistic Performance characteristic of V2Vcommunication presented in this paper are valuable input when implementing the general design pattern of KARYON for the scenarios addressed in the scope.

MEASUREMENT SETUP The SARTRE demonstrator was used as a measurement platform with the difference that all vehicles had manual longitudinal and lateral control except for cruise control in the lead vehicle. The target gap between vehicles is 13m and was maintained based on driver judgment.

antenna configuration on the cars did not change during the tests. In the rest of the paper, when any antenna position is referred, such as roof and RVM, then this refers to the antenna position of the trucks.

Figure 3. The platoon. Two trucks in front - The lead truck(LV) and following truck (FV1). Three following passenger cars towards the rear (FV2..4).

Figure 4 shows the two placements of antennas on the trucks. At position 1 and 2 there are Mobile Mark RM3-5500 surface mount antennas with 5 dBi gain. These are mounted, with small metal brackets, on top of the side rear view mirrors. This position is denoted RVM in the paper. The height to the top of each mirror is 276cm. At position 3 and 4 there are Mobile Mark ECOM6-5500 stick antennas with 6 dBi gain. These are mounted with the magnet base on the cabin roof; each approximately 75 cm from the center line of the cabin and 80 cm from the read edge of the cabin, i.e. edge of the container. The height of the cabin roof where the antenna is mounted is 3,8 meters above the ground. The antenna is 26cm high and thus barely reaches over the container at the rear (height 4m).

Track The tests were done at the APPLUS IDIADA test grounds in Spain. The general road track (5.333km long) is entirely within the large oval high speed circuit, see Figure 2. On the high speed track (7.493 km long) constant speed was possible with all platooning vehicles. The inner lane was used. On the general road the speed denotes the maximum speed for the test, i.e. the vehicles had to slow down (to 30km/h lowest) in the curves. 50, 70 and 85 km/h targeted top speed was tested on both tracks. The environment at the track was: 16,5 C° and 1023 mbar air pressure. There was no rain and the tracks were dry.

Figure 2. The two test tracks.

Antennas, Placement Figure 3 shows the SARTRE platoon. The order of the vehicles did not change during the tests. The trucks have the same exterior dimensions. Both have double rear axles and carry a metal container. The antennas of the cars were placed on the center of the roof; approx 1,5 meters above the ground. The

Figure 4. The antenna placements on the trucks as seen from above.

V2V-Test Nodes The V2V-test nodes and software were developed at SP Technical Research Institute of Sweden and used in the SARTRE project. They are based on the ×86 single board PC ALIX 3D3 [12]. The radios in the V2V-nodes are based on IEEE 802.11p which is an amendment to IEEE 802.11 [13] to add Wireless Access in Vehicular Environments (WAVE). It proposes small modifications to the PHY and MAC layers (compared to IEEE 802.11a) in order to achieve a robust connection and a fast setup-time for moving vehicles. The radios use the Atheros AR5414 chipset. The nodes use ITS-G5 [14] to transmit messages. This frequency band (5855 - 5925 MHz) is reserved for ITS applications (Intelligent Transport Systems) in both Europe and US. The G5CC (Control Channel) is used; which is allocation (5,9GHz) of the ITS-band. The output power from the node is configured to 20 dBm. This level complies with the EIRP power limit (33 dBm) of the channels that are used, i.e. G5CC of the


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014) ITS-G5A band provided that the antenna gain is less than 13 dBi. An overview of ITS communication and the various involved European standards such as ETSI ITS-G5 is given in [15].

suffer failure during the test. The measurements hence reflect the V2V-communication performance and differences of the antenna positions, tracks and speeds.

Each node is connected to a GPS-receiver and therefore logs position etc. and synchronize the local clocks to UTC. The packets that are broadcast from each V2V-node contain e.g. the GPS position of the sender. It has a payload of approx. 20 bytes. The V2V-node uses UDP/IP for transport over the wireless link, implying 28 bytes overhead. The wireless link (802.11p) adds 34 bytes to overhead. The frame size over the wireless link is hence 82 bytes (MPDU) and is not varied during the tests.

Performance of the V2V-system is defined by two metrics; System Error Rate (SPER) and Consecutive Loss of packets (CPL). Lost packets can be e.g. received but corrupted packets or transmitted but not received packets. This is not differentiated between in the tests. All V2V-nodes transmit the same number of packets during the tests.

The V2V-test nodes broadcast packets every 25ms at a fixed rate, i.e. 40 packets per second. The number of sent packets is logged each second. Each node logs how many packets that are received from individual senders during one second. The V2V-nodes in the trucks have two separate radios, each connected to one antenna. The same packet is sent independently by both radios, i.e. two identical packets “in the air”. Truck V2V-nodes hence send twice as many packets as car V2V-nodes. The normal CSMA/CA multiple access protocol in 802.11 implies that the packets are not “in the air” simultaneously, but are interleaved. On reception of redundant packets only the first correctly received packets is accepted. This scheme with dual antennas was devised as a simple method to increase probability of correct reception of packets e.g. from the lead vehicle. The drawback of the method is that it uses twice as much channel capacity and that the scheme does not confirm to ETSI ITS standards. Similar V2V-nodes are also described in [10]. The tested configurations are summarized in Table 1. In total 12 different configurations, of three speeds, two antenna position and two tracks, were tested. Table 1. The tested configurations

SPER is a statistical measure of the ratio of lost packets to sent packets from a communications system. Here, the system can contain multiple redundant communication channels between the communicating pair of nodes. For example, sending from truck to truck implies a system of four communication channels in the system: Left to left, right to right, left to right and right to left antennas. A truck sending and receiving from a car implies two channels in the system. The trucks V2V-nodes send each a packet over two antennas, i.e. two packets with identical information are sent from the vehicle; one from each radio. Any redundant (duplicate) packets that are received are disregarded. Car to car implies one channel in the system. In the tests there will hence be four different pair configurations, i.e. systems. CPL is a measure of the burstiness of packet loss, i.e. the highest number of consecutively lost packets during one second. A CPL of zero implies that, during a one second interval, no packets were lost at all. A CPL of one implies that single packets were lost during the interval. Note that during the interval there could be several instances of single lost packets, but none in a row. A CPL of two implies that, during the interval, a maximum of two packets in a row were lost. This implies that there was no correct packets between the two lost. Further, there could also be instances of lost single packets. Only the highest CPL is recorded. CPL is relevant since it gives a measure of the number of lost packets in a row that the end system has to tolerate. When packets are lost, the end system is not updated with data and must possibly find an alternate strategy to handle these “black-out” periods. CPL denotes the length of these “black-out” periods. A measurement of communication quality, expressed in the two metrics, will apply to a particular system, i.e. the configurations in the communicating pair of nodes. The SPER and CPL of the different systems, i.e. configuration of the communication pairs of nodes, cannot be directly compared, i.e. they are “apples and oranges”.

RESULTS Measurements are done to assess performance of the V2V-system during steady-state platooning. Therefore maneuvering, such as acceleration/deceleration and lateral changes (turning) to form the platoon, is not included. Each test starts with the vehicles in platoon-formation. Further we assume that the V2V-communication equipment does not

In the SARTRE platooning application control of the platoon is coordinated from the LV. Therefore the results in the next sections are focused on communication from the LV to the FVs and from each FV to the LV.


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014)

System Packet Error Rate Figure 6 shows SPER for two different antenna placements on the trucks at the three different speeds. Around 10000 unique packets were broadcast from each node during the six test. The tests were done on the general road track with one lap per test. Each group of four measurements (with a blank), one group for each speed, is (from left to right): 1. Lead vehicle sending to following vehicle, Roof antennas. 2. Following vehicle sending to lead vehicle, Roof antennas. 3. Lead vehicle sending to following vehicle, RVM antennas. 4. Following vehicle sending to lead vehicle, RVM antennas. The following observations can be made. There is a positive correlation between speed of the platoon and SPER. RVM placement of the truck antennas is always superior than roof placement, see also Table 2. Transmitting from the lead vehicle to a following vehicle always has lower SPER (on average) than transmitting from a following vehicle to the lead vehicle, see Table 2. The SPERs in the table are averaged over all three speeds. Table 2. Average SPER for different communication pairs on general road

These conclusions are similar to what was observed on the high speed track although the SPER was generally lower, see Table 3. This can be expected since constant speed was possible all around the circuit and therefore there was much less changes in speed. Also, the curve radius is much greater. Table 3. Average SPER for different communication pairs on high speed circuit.

Figure 5. SPER vs. position behind LV using roof antennas.


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014) The test results of using RVM antennas, Figure 7, are not as easily interpreted. The truck in position two (FV1) has the worst SPER for receiving from the LV, but the best SPER for sending to LV. The first following car in position three (FV2), has the worst SPER for sending to LV. The last car, in position five has the best SPER for receiving from the LV. However, compared to using roof antenna, Figure 5, the SPER is lower in all cases and in both directions. These results are should be investigated further to add understanding.

Consecutive Packet Loss Figure 8 shows Consecutive Packet Loss for transmissions between the LV and each FV, for the two antenna positions. The figure shows the tests done on the general road. The highest number of consecutively lost packets in a row (y-axis) is recorded during one-second periods of the test. The x-axis the amount of time (in one second periods) that is spent with a certain CPL. The tests are a combination of 50, 70 and 85 km/h using either roof or RVM antenna on the trucks. For example. For LV sending to FV1 (left-most bar in each group) with RVM antenna the expectation is that 81% of the time there will be zero packets lost. 19% of the time there will be at most one packet lost in a, i.e. not in a row.

Figure 6. System packet error rate for general road and two antenna placements. LV -> FVx in four sub-figures. The four bars in each speed-group are: Roof antenna LV->FV (light grey) and FV->LV (grey), RVM antenna LV->FV (dark grey) and FV->LV (black)

Figure 7. SPER vs. position behind LV using RVM antennas.

Figure 5 and Figure 7 show the SPER vs. vehicle position in the platoon using the two antenna configurations. Position two is the following truck FV1. Position three, four and five are the following passenger cars FV2-4. The test are done on the general road and are an average of all three tested speeds. The results of the test of using roof antennas, Figure 5, imply monotonically increasing SPER with respect to distance to LV. Using the roof antenna, the position two vehicle has lowest SPER in both direction. This is expected since the vehicle is closest and also has a dual antennas as opposed to the FVcars which have single antennas.

Figure 8. Consecutive loss of packets. Average over all speeds. LV -> FVx in four sub-figures. The four bars in each group are: Roof antenna LV->FVx (light grey), and FVx->LV (grey), RVM antenna LV->FVx (dark grey) and FVx->LV (black).


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014)

CONCLUSIONS AND THOUGHTS The paper describes results of tests done on a V2Vcommunication system that was designed for the SARTRE platooning system. Two examples of cooperative applications are presented: vehicle platooning - as studied in SARTRE, and cooperative UAVs - as studied in KARYON. The performance of the V2V-system is studied according to system packet error rate (SPER) and consecutive packet loss. In both the general road tests and high speed circuit tests, the rear view mirror antenna was clearly superior to the roof antenna on the trucks. As was seen in the results, the SPER increases with both speed and distance between sender and receiver. This will limit the length of the platoon. It therefore seems inappropriate to rely on broadcast communication from the lead vehicle. Two potential solutions can be: 1) Improving the data-architecture, e.g. look ahead three vehicles in front of own vehicle. Not require broadcast contact from LV. 2) Use multi-hop communication to reliably relay messages through the platoon.

Figure 8. (cont.) Consecutive loss of packets. Average over all speeds. LV -> FVx in four sub-figures. The four bars in each group are: Roof antenna LV->FVx (light grey), and FVx->LV (grey), RVM antenna LV->FVx (dark grey) and FVx->LV (black).

In the figures, CPL higher than five is truncated. CPL higher that five accounted for 0,42% in FV3 -> LV, 0,53% in LV->FV4 and 3,07% in FV4 -> LV. The latter is the “weakest communications pair”. These three cases where with the roof antenna. In all other cases for roof and all cases for RVM antenna there was no intervals with CPL higher than five. The significance of the CPL metric is that it relates to the amount of time that a receiver is not updates with fresh data. In the tests the update period is 25 ms. This is how “old” the data is at best, i.e. CPL is zero. A CPL of one implies that the “data-age” is 50ms, i.e. one update is missed. A CPL of two implies that the “data-age” is 75ms, i.e. two update is missed. During all tests, the longest “black-out”-period (consecutive packet loss) was 17 consecutively lost packets, i.e. no data updates from the sending node. This occurred three times and was between FV4 -> LV, i.e. the weakest communication pair. This corresponds to 450 ms without an update. In this case the SPER was 13,3%. In the opposite direction (LV -> FV4) the longest CPL was 12, i.e. 325ms without an update and 3,4% SPER. This case was using the roof antennas. According to the fictive fall-back strategy of the SARTRE platooning system, this temporary communication loss would not disrupt platooning, i.e. force the fall-back scheme to class four. Hence the V2V-system would be sufficient for SARTRE platooning in all the configurations that were tested.

In Table 2 and Table 3 the Sending FV1 to LV seems to have three times worse SPER than LV to FV1. Both pairs have the same system configuration, i.e. dual radio node to dual radio node. We expected the SPERs of the two directions to be more similar and cannot explain this. In a control application loss of packets can imply instability and degraded control performance. To counter these effects a redundancy strategy should be employed. In the control application the data is often sent at regular periods. This is a simple redundancy strategy: data will be sent again even though the latest data packet was lost. However, it requires that the update frequency is adapted to suit the dynamics of the application and not to over-utilize the channel. As is seen in our tests, a V2V communications channel is unreliable and has a high probability of loss. To counter this, a quick fix is to set a high update frequency of data to increase probability of reception. However, this can easily over-utilize the channel with e.g. congestion and dropped packets as a result. Instead, the control application must be designed to handle the dynamic nature of the V2V-communication. Two examples are using an appropriate fall-back strategy and by characterizing the channel and understanding how it can be expected to vary. The latter could be done online to continually monitor performance of e.g. data-age and SPER, and adapt the control-strategy accordingly. Data-age is feasible to monitor online, since packets can be time stamped before sending and that nodes can be clock synchronized. For event-triggered messages that convey e.g. commands other strategies should be used rater than repeating the message. A command can be something that all participants shall obey, and if not the system may become unsafe. Here, consensus protocols, such as membership agreement [16], atomic broadcast and leader election [17] can be used in addition to Acknowledgement of reception and retransmission.


Bergenhem et al / SAE Int. J. Passeng. Cars – Electron. Electr. Syst. / Volume 7, Issue 2 (August 2014) The wireless environment that was investigated implies a high probability of packet loss and consecutive “black-out” periods; even during steady-state operation. This can be put in contrast to the higher reliability of wired networks, such as CAN or FlexRay in the automotive domain. This type of network is used to interconnect components within the cooperative entity; sensors, actuators and processing units. For cooperative applications, the application spans not only the (reliable) networks within the entities, but also includes the (unreliable) wireless link between them. Future work should focus on methods to achieve dependable cooperative systems.

REFERENCES

1. Bergenhem C., Huang Q., Benmimoun A., and Robinson T., “Challenges of platooning on public motorways,” in 17th World Congress on Intelligent Transport Systems, 2010, pp. 1-12. 2. Casimiro A., Kaiser J., Schiller E., Costa P., Parizi J., Johansson R., and Librino R., “The KARYON Project: Predictable and Safe Coordination in Cooperative Vehicular Systems,” in 2nd Workshop on Open Resilient Human-aware Cyber-Physical Systems, WORCS, 2013. 3. Kunisch J. and Pamp J., “Wideband car-to-car radio channel measurements and model at 5.9 GHz,” in Vehicular Technology Conference, 2008. VTC 2008-Fall. IEEE 68th, 2008, pp. 1-5. 4. Cheng L., Henty B. E., Stancil D. D., Bai F., and Mudalige P., “Mobile vehicle-to-vehicle narrow-band channel measurement and characterization of the 5.9 GHz dedicated short range communication (DSRC) frequency band,” Selected Areas in Communications, IEEE Journal on, vol. 25, pp. 1501-1516, 2007. 5. Paier A., Karedal J., Czink N., Dumard C., Zemen T., Tufvesson F., Molisch A. F., and Mecklenbräuker C. F., “Characterization of vehicle-to-vehicle radio channels from measurements at 5.2 GHz,” Wireless personal communications, vol. 50, pp. 19-32, 2009. 6. Karedal J., Tufvesson F., Abbas T., Klemp O., Paier A., Bernadó L., and Molisch A. F., “Radio channel measurements at street intersections for vehicle-to-vehicle safety applications,” in Vehicular Technology Conference (VTC 2010-Spring), 2010 IEEE 71st, 2010, pp. 1-5. 7. Bohm A., Lidstrom K., Jonsson M., and Larsson T., “Evaluating CALM M5-based vehicle-to-vehicle communication in various road settings through field trials,” in Local Computer Networks (LCN), 2010 IEEE 35th Conference on, 2010, pp. 613-620. 8. Molisch A., Tufvesson F., Karedal J., and Mecklenbrauker C., “A survey on vehicle-to-vehicle propagation channels,” Wireless Communications, IEEE, vol. 16, pp. 12-22, 2009. 9. Karlsson K., Bergenhem C., and Hedin E., “Field Measurements of IEEE 802.11 p Communication in NLOS Environments for a Platooning Application,” in Vehicular Technology Conference (VTC Fall), 2012 IEEE, 2012, pp. 1-5.

10. Bergenhem C., Hedin E., and Skarin D., “Vehicle-to-Vehicle Communication for a Platooning System,” Procedia - Social and Behavioral Sciences, vol. 48, pp. 1222-1233, 2012. 11. Bergenhem C., Shladover S., Coelingh E., Englund C., and Tsugawa S., “Overview of platooning systems,” in Proceedings of the 19th ITS World Congress, Oct 22-26, Vienna, Austria (2012), 2012. 12. ALIX 3D3 embedded PC. Available: www.pcengines.ch/alix3d3.htm 13. “IEEE Standard for Information technology, IEEE 802.11-2012 Telecommunications and information exchange between systems Local and metropolitan area networks-Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications”. 14. ETSI, “European Profile Standard for the Physical and Medium Access Control Layer of Intelligent Transport Systems Operating in the 5 GHz Frequency Band, ETSI ES 202 663 (V1.1.0).” in Intelligent Transport Systems (ITS), ed, 2010. 15. Strom E. G., “On medium access and physical layer standards for cooperative intelligent transport systems in europe,” Proceedings of the IEEE, vol. 99, pp. 1183-1188, 2011. 16. Cristian F., “Agreeing on who is present and who is absent in a synchronous distributed system,” in Eighteenth International Symposium on Fault-Tolerant Computing. Digest of Papers. FTCS18 Tokyo, Japan, 1988, pp. 206-11. 17. Fathollahnejad N., Villani E., Pathan R., Barbosa R., and Karlsson J., “On reliability analysis of leader election protocols for virtual traffic lights,” in Dependable Systems and Networks Workshop (DSNW), 2013 43rd Annual IEEE/IFIP Conference on, 2013, pp. 1-12.

ACKNOWLEDGMENTS The research leading to these results was received funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 233683. SARTRE is a three year programme. Project Partners: Applus+ IDIADA, Institut für Kraftfahrzeuge Aachen (ika), Ricardo, SP Technical Research Institute of Sweden, Tecnalia, Volvo Cars, Volvo Technology. At the time of measurements the main author, Carl Bergenhem, was employed at SP - Technical Research Institute of Sweden. Support has also been provided by VINNOVA (Swedish Governmental Agency for Innovation Systems) FFI project RELCOMMH and EU FP7 project KARYON (grant agreement n° 288195).

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE International. Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE International. The author is solely responsible for the content of the paper.