Impact of distances estimation errors on the communication reliability in DSRCbased vehicular networks

Transcription

1 Impact of distances estimation errors on the communication reliability in DSRCbased vehicular networks Omar Chakroun, Jean Marchal and Denis Gingras LIV laboratory, Electrical and Computer Engineering Department, Université de Sherbrooke, Canada s: Abstract Vehicular communication technology aims to use communication equipment and infrastructure introduced in vehicles (OBU) and roadside units (RSU) to improve road safety and provide useful services for road users. In the case of safety services relying on communication, the communication range, communication reliability and accurate measurements on the vehicles dynamics may be critical in determining if, in fact, these services will be guaranteed or not. In this work, we study the impact of the distances / localisation estimations errors on the communication reliability based on the deployed communication schemes under DSRC communication technology as a backbone for safety related applications deployment. We examine the impact of such estimations errors on three main metrics; 1) the impact on the network connectivity, 2) the impact on the packet reception ratio, and 3) the impact on the resulting notification delay. We expose the communication scheme performances under two case studies; a) an overestimation of the distances, and b) an underestimation of the distances. All of which is considered mainly in a V2V safety communication architecture based on BSM messages exchange. Its worth noting that the majority of the deployed applications in ITS need accurate positioning information and localisation usually in the order of centimeters and that the latter accuracy is only achieved with dedicated equipment such as RTK. Nevertheless, the majority of the deployed communication scheme assume that the positioning information is 100% accurate in a sub-decimeter scale. Through this paper, the discussion will be illustrated by some theoretical considerations and it will be proven that even a 10m error in distance estimation, which corresponds to the actual GPS accuracy, can have a considerable impact on the previously defined communication performances metrics. KEYWORDS: DSRC, distance/localisation accuracy, communication reliability. I-Introduction The majority of the protocol used to diffuse information in VANETs are based on distance or location information in order to operate. These protocols leverage precise localisation information gathered by the integration of specific equipment such as GPS receivers or triangulation like techniques taking into consideration the received signals strengths from specific emitters with known location and while considering the distance between the emitter and the receiver as the main criteria for signal attenuation. The latter assumption is not accurate since signals strength can be affected by a multitude of other aspects such as reflection and refraction in large building surfaces and is of a high impact especially in urban environments. Messages dissemination solutions within VANETs assume that the location information is accurate in order to operate. Though, the location information can be of a multitude of nature, it will be mainly used to estimate the distance between an emitting node and an eventual receiver or relay node. This estimation will be compared to the actual communication range in order to select the best node within the range that limits the number of hops to relay the information to the maximum distance while ensuring a high probability of reception to avoid messages retransmission [1, 2]. It is worth noting that in VANETs, vehicles exchange Basic Safety Messages (BSM) that contain information on their dynamics (speed, heading, positioning, etc.) over one or multiple hop to construct an awareness of the surrounding environment [3]. Based on the latter exchange, each vehicle will evaluate the distance between it and any vehicle in the communication range in order to favor one or the other to relay its message or in order to estimate the probability of reception.

2 Multitudes of location/position based forwarding techniques in VANETs were designed in order to ensure a highly reliable data exchange and to avoid network resources degradation. Smart Broadcast (SB) [4] uses a distance based forwarding technique electing the farthest node in the communication range of the emitter as a relay based on blackburst. Korkmaz et al. proposed two designs; Urban Multi-hop Broadcast (UMB) [2] that uses a continuous message exchange to calculate distances between communicating nodes and elect the farthest one as a relay. The other design is Ad hoc Multi-hop Broadcast (AMB), which is an improvement of UMB electing the closest node to an intersection as a relay to a particular section of the road. Another scheme, Fast Broadcast (FB) [5] uses an adaptation of the waiting time before rebroadcasting by giving the farthest vehicle in the communication range a higher priority to relay the message. REAR [6] considers the Packet Reception Rate (PRR) as a main metric to guaranty, but does not offer any bound on data forwarding delays. GVGrid [7] rely on categorizing communicating vehicles based on their speed, heading, by constructing a cluster of vehicles depending on their proximity to each other. Naumov et al. introduced Connectivity Aware Routing (CAR) [8], which by pre-establishing the dissemination path guarantees lower delays. It uses HELLO messages exchange from the source to the destination and on the reverse path to construct a routing route similarly to AODV. Position-Based Adaptive Broadcast (PAB) [9] integrates a design to overcome network disconnections by implementing a store-and-forward scheme used in case of links breakage. It uses position and speed information to construct a global routing map in the network. DTSG [10] introduces another approach of geocasting, called time-stable as it acts on the time when messages are geo-casted. DTSG integrates the idea of helping vehicles navigate in the opposite direction. However, geo-casting needs a continuous exchange of control messages containing location and positioning information. Ayaida et al. in [11] presented a highly interesting concept combining routing protocol with location-based services for a hybrid and hierarchical geographic routing protocol. In such schemes, the messages are forwarded using the last receiver updated position. All the aforementioned two techniques rely on an accurate location and distance measurement information exchange. Meanwhile, the actual positioning accuracy is measured in meters and that the maximum achievable positioning accuracy is within 1m and needs heavy equipment such as RTK systems [12]. It is worth noting also that a distance estimation error can be result of a multitude of factors such as an estimation error in a sensor, the used positioning algorithm accuracy, or the impact of environment on the received signals from GPS satellite or fixed reference points. The latter estimation error can have a negative effect on the communication scheme and that effect is proportional to the dependency between the communication scheme and the accuracy of the location information. In this paper, we are discussing the impact of distances estimation errors on the broadcast based communication schemes in VANETs. Our discussion will be only based on theoretical analysis through three main criteria; 1) the network connectivity, 2) the packet reception rate (PRR) and 3) the end-to-end delivery delay (E2E). The remainder of this paper is organized as follow; Section II presents the impact of the distances measurement errors on the network connectivity. Section III introduces the latter errors impact on network performances in term of E2E delay and PRR and presents some theoretical results. Finally, section IV concludes the document. II-Impact of the distance estimation errors on the network connectivity Geocast techniques are the most vulnerable dissemination techniques to the errors induced in the location determination. They rely on such information in order to privilege messages dissemination in a particular direction on broadcast it to a specific geographic region. They use a multitude of relays in order to transport the message directly to the interested receiver or to a close relay that will broadcast the message in the nearby area of the potential receiver. The selection of such relays is based on the construction of tables that contains location information exchanged between nodes and depending on their proximity to the end receiver (inferred distance based on the receiver and emitter location estimation). The inferred distance is then compared to the communication range (CR) of the emitter and then a selection of the possible relay nodes is performed depending on the selected criteria (minimum delay, maximum reliability, etc.). a choice based only on location

3 information make these techniques especially vulnerable to the disconnection problem which is reflected by the lack of reliability of the chosen links. Broadcasting and flooding techniques are particularly robust to the error in distances estimation since they don t rely on a preselection of relay nodes and simply broadcast the message to all the neighborhood. These techniques doesn t rely on a positioning information but they translate it to an information on distance to determine the probability of reception of a given message. Of course, broadcasting based techniques are subject to the same constraints regarding signals attenuation function of the distance between the emitter and the receiver. One of the techniques to make them robust against such estimation errors is the use of the greedy forwarding technique that leverages an exchange of messages called black burst. These particular type of messages will be used to favor the farthest nodes within the communication range to relay the message by reducing the waiting time after reception in these particular nodes. After the successful reception of the message, every node that received the message will decrement a waiting counter that is inversely proportional to its distance from the emitter. The first node, which counter ends, will rebroadcast the message to the next section of the network and every node that hears that retransmission will remove the corresponding message from its waiting queue. Consequently, the duplicated reception can be perceived as an implicit acknowledgment. These techniques are particularly immune since every node that received the message can play the role of a relay but the main disadvantage is that even in the case of inexistence of any relay node, the message is diffused and no control is made on the availability of a possible relay. This can cause network performances degradation, induce extra delays if a retransmission technique is implemented and cause interferences on the distant communications (happening in other communication ranges). Unicast techniques are facing the same challenges since a relay node preselection is always made prior to the transmission phase and that routing tables has to be maintained to ensure that goal. Generally, routing tables are constructed based on the location information exchanged to favor a pre-established path. Unicast techniques evaluate the routing path based on the distance between every couple of nodes and infer to the optimal path depending on the measured distance and the communication range. Thought, an error in distances estimation can lead to a connectivity problem and consequently cause the non-reception of a critical safety message. The figure hereafter, illustrates two cases where the choice of the relay node is made on the connectivity barrier (i.e. maximum communication range). RR3 and ER3 designate the real position and the estimated position of node 3 respectively. Rdr3 and Edr3 designate the real distance between node E and node 3, and the estimated distance between node E and node 3. (a) Case where the distance is overestimateestimated (b) Case where the distance in under- Figure 1 Scenarios for distance under and over estimation In the first case, the distance is over-estimated, the network connectivity is maintained since the real distance between the emitter and the receiver is less that the maximum achievable communication range of the emitting node. This does not affect the message dissemination but we will discuss later some of the network performances that can be affected.

4 In the second case, where the distance is under-estimated, the network connectivity is compromised since the real distance between the emitter and receiver nodes is higher than the maximum achievable communication range. In this case, the node 3 will not receive the message, will not be advised by a potential safety related message, and consequently the information will be not forwarded to the next section of the road. III-Impact of the distance estimation errors on the communication schemes performances In this section, we will discuss the impact of the distances estimation errors on the performances of the communication scheme on two majors network performances metrics; 1) the end to end delay (E2E) and the packet reception rate (PRR). This study will be based on the previously illustrated cases. A-Impact on the End-to-End delay (E2E) Let us assume that a transmission in one hop over a distance Dx corresponding to the maximum communication range while considering a transmission power Px takes a time Tx. Let us assume that all nodes use the same transmission power and that this decision is made collectively. In the first case, where the distance is over-estimated, the connectivity is maintained and the delay will not be affected since only the portion corresponding to the flight time is affected and the latter is considered negligible compared to the queuing and transmission time. Let us assume now that in order to reach the end receiver, we need to make m hops while choosing a relay at a distance Dr, equivalent to the maximum achievable communication range CR at each hop. The fact of choosing a node at a lower distance di than Dr can cause the need of an extra hop in order to reach the end receiver. This will induce and extra delay which has to be taken into consideration. At the end, the communication scheme will necessitate m+β hops with β as shown here after and the overall E2E delay can be expressed as follows (1) 2. In the second case, where the distance is under-estimated, the network connectivity is interrupted and the message will not be successfully delivered. Let us assume that in that case, a retransmission technique is implemented. Let us assume that the latter retransmission is triggered after a time out delay TO. Let us assume that to get a successful retransmission, the system make n unsuccessful attempts and succeed in the n+1 attempt. Thus, the transmission delay over one hop can be expressed as follows.. (2) As in the first case, eventually the communication scheme will need extra-hops in order to reach the end receiver. This will lead to an E2E delay that can be expressed as follows while considering the same β as before. 2.. (3) It is worth noting that the latter delay expression only considers the case where the error on distance estimation happened at a one hop level. Of course, the latter can be extended to the cases where the error is made at every hop or a combination of hops.

5 B-Impact on the packet reception rate (PRR) In this section, we are discussing the impact of distances estimation errors on the packet reception rate. In the first case, where the distance is over estimated, the connectivity is maintained and the real relay node position is closer to the emitting node. Consequently, that particular transmission will experience a better reception and the overall PRR will not be negatively affected. Meanwhile, if the distance is underestimated, the connectivity is lost and the PRR will be affected if the overall retransmission delay exceed the maximum permissible delay (100ms) for a safety message. Let us assume that we use a communication scheme that leverages a broadcasting or unicast like behavior and that pre-establish the relay nodes selection over the path before starting data transmission. Let us assume also that the distance measurement is based on signal attenuation combined with a localisation scheme that use the latter measurement to infer to the position based on trilateration-like algorithm. Thus, we can express the signal strength received in the receiver side using the Friis model [13] with Gi, denoting the emitter and receiver antenna gains, λ the wavelength, Pi the emitting power, L the considered system loss and d the distance between the emitter and the receiver. (4) 4 Let us consider that the transmission power is the same for all nodes and that this decision is made collectively. Let us also consider a message to be successfully received if the PRR exceeds a certain threshold PRRth. Let us denote the communication range ensuring a PRR equal or higher than the PRRth by CRVC (Communication Range Verifying Constraint). The latter CRCV is consequently lower that the CR. Let us now consider the function below that illustrates the link between the PRR, the emitting power, the estimated distance between the emitter and the receiver [15]. x designates the received power, Rx designate the receiver threshold, m the fading parameter used in the Nakagami-m distribution and Ω the mean signal strength at a distance d from the emitter obtained by the Friis formulation exposed above. PR(x>Rx) = 1- Fd(Rx;m,Ω) = / (5) Let us take m=3, the value of the fading parameter that matches the case of a vehicular network in a highway environment. Thus, we can derive that! PR(x>Rx) = 1 3 (6) Normally, a received power level Rx is detected at a distance equal to the maximum achievable communication range CR from the emitter. Let us now consider a quadratic error on the path loss as the one defined using the Friis model. We can derive the received signal strength Rx as follows while considering a transmission power Tp Rx = avec G = (7) Moreover, we can derive the mean signal strength received at a distance d as Ω(d) = (8) By substituting Rx and Ω(d) in expression 6, we obtain the probability of reception at a distance d lower that the crossover distance while considering a maximum communication range corresponding to CR. ht,hr designate respectively the emitter and receiver antenna heights. designates the wavelength.

6 PR(d,CR) = 1 3 (9) In the case where the distance is higher than the cross over distance, the TwoRayGround model path loss is applied and consequently the probability of reception can be expressed as follows where. PR(d,CR, ) = 1 3 (10) We notice the close link between an accurate distance estimation and the probability of reception. This is of a particular interest in the case where the communication scheme implement the route discovery or the path preselection based on the distance information. Figure 2 illustrate the probability of reception for multiple transmitting power while varying the considered distance compared to the emitter. In the graph, the estimated maximum achievable communication range (CR) is based on the deterministic TwoRayGround model [14]. Figure 2 Impact of the variation of the emitting power and the distance on the packet reception rate Table 1 shows the effect of distance estimation error on the measured PRR. We notice that the error has a lower impact when we consider a higher communication range (higher transmitting power). This is trivial since the fact of increasing the transmitting power increases the communication range and consequently increase the probability of reception especially at close range from the emitter. Let us take the case where the communication scheme uses a PRR threshold to determine if a chosen link can be selected to relay the information. An error on the distance estimation will result in an error on the measurement of the PRR and consequently may lead to a bad choice of the relay node. Let us consider CRVC as the maximum usable communication range that verifies the

7 constraint on the PRR (i.e. PRR th) thus an error on the distance measurement can affect the CRVC and may lead to three cases. In order to discuss the impact such effect, let us denote the real distance between the emitter node and the receiver by d and the error on the distance regarding the direction of travel of the information as Δd. If the sum of the distance and the maximum error is within the communication range verifying the constraint on the PRR (CRVC); d + Δd < CRVC. In this case, the connectivity is maintained and the constraint on the PRR is respected. The only effect is on the E2E only. If the sum of the distance and the maximum error is higher that the CRVC but within the maximum achievable communication range CR; CRVC < d + Δd < CR. in this case, the connectivity is maintained but the constraint on the PRR is no more respected. The system will experience a lower PRR. If the sum of the real distance and the maximum error is higher that the maximum communication range; d + Δd > CR. In this case, the connectivity is no more maintained and the system will experience a loss of messages if no retransmission technique is implemented. It can be considered as in the case where the distance is under-estimated as in the previous section with all the related consequences. Table below shows the impact of a deviation in the distance estimation on the PRR while varying the considered communication range and the deviation in it self (values between parentheses). It also presents the case where the deviation is expressed in percentage from the real distance. Table 1 Influence of the distance estimation error on the PRR Maximum communication range (meter) PRR deviation (±1m) 1,13 % 0,97 % 0,38 % 0,28 % 0,23 % 0,19 % PRR deviation (±5m) 6,36 % 3,01 % 1,97 % 1,47 % 1,17 % 0,97 % PRR deviation (±10m) 14,41 % 6,40 % 4,12 % 3,03 % 2,40 % 1,98 % PRR deviation (±10%) 6,36 % 6,40 % 5,86 % 6,02 % 5,77 % 5,90 % We can notice that the impact of a low error of the distance measurement (1-5m) on the PRR is relatively low when considering communication range higher than 150 meters. It does not exceed 2%. Meanwhile, if we consider an error around 10 meter, which corresponds to the actual GPS measurement accuracy, the deviation of the PRR is higher and can reach more than 14% depending on the envisioned communication range. Although low errors have a limited impact on the PRR, they have to be quantified and taken into consideration in order to ensure the effectiveness of the communication scheme and to ensure a high messages delivery rate especially those related to safety messages in a highly dynamic environment such as vehicular networks. The majority of the proposed dissemination solutions in VANETs take the extreme cases where the connectivity is lost into consideration and propose schemes to overcome these situations. The main countermeasures are those related to retransmitting the messages after a Timeout or the use of implicit acknowledgment (hearing the retransmission) by detecting the retransmission of the delivered message. Meanwhile, they do not offer mechanisms for the intermediate situations where the connectivity is maintained but with lower performances. IV-Conclusion With DSRC becoming the leading technology in vehicular communications, several different techniques have been developed to insure the best and most reliable data exchange possible. All of them depend on the location of entities provided by embedded GPS or trilateration like algorithms based on received signal strength. This shows evidences about the fact that the lack of accuracy of the actual positioning techniques will not represent a real problem and furthermore concluding on the fact that the technology will evolve, guaranteeing better results. However, this has been over-repeated for more than a decade and we are still reusing the same localisation technologies that still did not really show better accuracy performance. This lack of performances can, depending on the case, seriously affect the network connectivity or degrade the performances of the communication scheme by introducing higher delay or very low reception probability even at close

8 communications ranges. Using the latter technologies without accounting for their unreliability will affect the deployed safety related applications performances and consequently not fulfill its premises. On the other hand, DSRC is showing some defects that cannot be ignored for a longer period. Otherwise, the already out-of-date technology will represent a true hindrance when its deployment will be at high scale. References 1. O.K. Tonguz and N. Wisitpongphan and B. Fan, DV-CAST: A distributed vehicular broadcast protocol for vehicular ad hoc networks, IEEE Wireless Communications, vol.17, no.2, pp.47-57, April G. Korkmaz, E. Ekici, F. Ozguner and U. Ozguner, Urban multi-hop broadcast protocol for intervehicle communication systems, In Proceedings of the 1st ACM international workshop on Vehicular ad hoc networks (VANET 04). ACM, New York, NY, USA, 76-85, BSM, 4. E.Fasolo,R.Furiato,A.Zanella, Smart Broadcast Algorithm for Inter-vehicular Communications, IWS 2005/WPMC 2005, Aalborg, Denmark, Sep P. Akkhara and Y. Sekiya and Y. Wakahara, Efficient Alarm Messaging by Multi-Channel Cut- Through Rebroadcasting based on Inter-Vehicle Communication, IAENG International Journal of Computer Science, Vol. 36, No. 2. (2009), pp H. Jiang and H. Guo and L. Chen, Reliable and Efficient Alarm Message Routing in VANET, The 28th International Conference on Distributed Computing Systems Workshops, Y.W. Lin and Y.S. Chen and S.L. Lee, Routing Protocols in Vehicular Ad Hoc Networks: A Survey and Future Perspectives, J. Inf. Sci. Eng. 2010, 26, V. Naumov and T.R. Gross, Connectivity-Aware Routing (CAR) in Vehicular Ad-hoc Networks,26th IEEE International Conference on Computer Communications, May Y. Yang and L.D. Chou, Position-Based Adaptive Broadcast for Inter- Vehicle Communications, IEEE International Conference on Communications, ICC Workshops 08, pp , May, Rahbar, H.; Naik, K.; Nayak, A., DTSG: Dynamic time-stable geocast routing in vehicular ad hoc networks, The 9th IFIP Annual Mediterranean Ad Hoc Networking Workshop (Med-Hoc- Net), 2010, vol., no., pp.1,7, June M. Ayaida, M. Barhoumi, H. Fouchal, Y. Ghamri-Doudane, and L. Afilal, Joint routing and location-based service in VANETs, J. Parallel Distrib. Comput. 74(2): (2014) 12. RTK, Friis model, Two ray Ground reflection model, M Killat, F Schmidt-Eisenlohr, H Hartenstein, et al. Enabling efficient and accurate large-scale simulations of VANETs for vehicular traffic management. Proceedings of the 4th ACM International Workshop on Vehicular Ad Hoc Networks (VANET '07), September 2007, Montreal, Canada, 29 38