Cooperative V2V Clustering (2015).pdf

Cooperative V2V Clustering Algorithm for Improving Road Traffic Safety Information Marius Minea Abstract – The paper presents solutions for improving traffic time for computing safety conditions, or delay introduced by safety by providing on-board information regarding traffic protocols for distribution of useful information to all involved behaviour of al local cluster of neighbouring vehicles. The
of 4
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Cooperative V2V Clustering Algorithm for Improving Road Traffic Safety Information Marius Minea  Abstract – The paper presents solutions for improving traffic safety by providing on-board information regarding traffic behaviour of al local cluster of neighbouring vehicles. The proposed solution also may contribute to reducing accidents on major motorways by issuing early warning regarding speed. The solution is based on a local clustering algorithm to ensure information provision between neighbouring groups of vehicles. Advantages and weaknesses of the solution are also analysed.  Keywords – Vehicle clustering algorithm, V2X communications (Infrastructure to Vehicle Communications - I2V, also referred to as V2I, vehicle to vehicle communications V2V), CALM (Communications Access for Land Mobiles). I.   I  NTRODUCTION   Actual European policies for road transportation focus on traffic safety. Among the objectives proposed in [1], Objective #7 states that a significant increase of the infrastructure use for transport is necessary over ”a deployment of equivalent land ... transport management systems (ITS - Intelligent Transport Systems), ..., and deployment of the European global navigation satellite system (Galileo)”. Objective #9 also states that 2050 should be the target year for reducing road traffic fatalities close to zero”. Modern ITS vehicle to infrastructure communications (V2I) recommend CALM for improving connectivity and inter-operability of different infrastructure equipment in a cooperative safety system for road traffic. ISO 13183:2012 specifies the architectural communications framework of ITS for the family of CALM related International Standards. The graphical representations of the architecture partly follow the ISO open systems interconnection (OSI) principles. In [2], the authors investigate the necessity of deploying “soft safety” mobility and convenience applications using a diverse variety of wireless communications technologies along with  proprietary systems and applications. Considering this, the crash-avoidance safety applications require that all  participating vehicles and roadside units are able to sense and understand each other and they are also able to communicate via a harmonised interface. Short range communications such as V2V (vehicle to vehicle) or V2I (vehicle to infrastructure) are used when discussing about traffic safety, especially for Cooperative Collision Avoidance  (CCA) systems. However, the communication link is only one of the processes that may induce delay in traffic management information delivery. Other processes might be: acquisition of dynamic parameters from vehicles (position, speed, heading, headroom etc.), Marius Minea is with Transports Faculty, University Politehnica of Bucharest, 313, Splaiul Independentei, Bucharest, Romania, E-mail: time for computing safety conditions, or delay introduced by  protocols for distribution of useful information to all involved vehicles’ drivers. In this case, the communication link  becomes more crucial, as quality of service (QoS) and privacy of information must be set at very high levels. In critical situations there remains little room for messages delaying due to packet loss retransmissions, jitter or other communication issues. On the other hand, new achievements in standardisation offers CALM (ISO TC204 WG16), which ensures safer V2V and V2I communications (standards IEEE802.11p and P1609, ETSI ERM TG-37 – 2G and 3G standards, IETF – network mobility – NEMO etc.) that are to  be taken into consideration when needing more speed,  prioritisation or bandwidth in dense urban environmental areas communications. In this paper a clustering algorithm for cooperative collision avoidance systems is investigated and case investigations are provided for determining the amount of delaying introduced by communications protocols and other information processing used for CCA systems. CALM M5 is recommendable for critical safety applications such as CCA, as it incorporates: global (European) 5 GHz spectrum, directivity and electromagnetic compatible (EMC) control along with GPRS/UMTS network interconnectivity. II.P RESENT A DVANCES IN V2XC OMMUNICATIONS R  ELATED TO T RAFFIC S AFETY   Different studies in scientific literature present a variety of results concerning afore mentioned aspects. In [3], Chou et. Al. measure the latency, throughput and packet loss in WiMAX and WiFi settings for a traffic communication system in an urban environment with disabled WEP encryption and without external antenna. These studies show that one might be able to get more throughput and a shorter latency from WiFi at short distance (e.g. less than 100 m), assuming there are too many interfering sources in the area that also use the 2.4 GHz band. It is also shown that frame duration can have a critical effect on the performance of WiMAX. Bohm and Jonsson [4] demonstrate that for critical safety transport systems, a communication link over the IEEE 802.11p standard is capable of supporting speed and data flow information for an up to four lanes highway, depending on actual vehicle density and average vehicles velocity. According to the authors of [5], the main objectives of a Traffic Control Centre (TCC) involved in computations for in-vehicle signing may include (considering all vehicles are linked informally to the TCC via wireless communication): -   Incident warning/critical traffic condition warning; -   Road works information (lane closures etc.); 978-1-4673-751 6 - 0 /15/$31.00 ©2015 European Union 369    -   In-vehicle variable speed limit information; -   Traffic congestion warning; -   Route navigation related information etc. For the different extreme case studies, the limiting factors are determined: -   High traffic density case (urban environment): limiting factor – the performance of the communication network at high traffic density; -   High cruising speed (inter-urban environment): the handover delay at high speed In [6] Biswas et. all demonstrate efficiency of a CCA system employing prioritised V2V communications to improve early warning effects on reducing traffic accidents occurrence. Using dedicated messages and prioritised channels, vehicles arriving on a highway to a recent crash site are instantly  prompted to reduce speed, avoiding chain collisions. Examples of such applications include en-route driver information propagation, collision warning and avoidance systems, and adaptive cruise control systems. For critical safety applications that require fast message broadcasting such as CCA, in [7] the authors declare that ad-hoc algorithms are not suggested and instead methodologies that provide formal safety guarantees, such as those found in the control theory and computer science literature are recommended. III.   C ONTEXT A WARENESS A RCHITECTURE AND A LGORITHM FOR D ELIVERING S AFETY I  NFORMATION A recent evolution in automotive industry and V2X communications is the one toward context awareness , involving vehicles aware of their neighbourhood (presence, location and safety in relationship with other vehicles). Modern vehicles may be equipped with the following: -   GPS/DGPS or A-GPS receiver with improved accuracy and in near-future, a Galileo SS receiver; -   e-CALL system for rapid event notification to relevant authorities; possible front-end/back-end radars for obstacle detection and collision avoidance and Event Data Recorder – similar to airplanes’ black-boxes etc. We can then assume such vehicles are ”smart”. The ”smart” feature, however, must also extend to the infrastructure, which may be seen both as a support for safety information collection and delivery. Fig. 1 shows this context. MrvMh d1, rs1d2, rs2d3, rs3 Msr    Fig. 1. Clustering of vehicles The dark red vehicle in Fig. 1 is the central (critical) node of the cluster, beneficiary of safety information. The proposed model supports clustering algorithm with one hop in communication and it is more rapid for CCA systems to have a single hop implementation. The one hop method simplifies the overall communication, clustering strategy and reduces the over exceeding of cluster dimensions. The neighbouring vehicles then fit in one of the following situations: -   Vehicles located in communications range (assuming the DSRC / WiFi is the communication technology, these are included in some tens to hundreds of meters in front and back of the main node):    ∈   ,   ={   ,  =1,…,  }  (1) where    represents the cluster of all neighbouring received vehicle identities and    is the identity of central node vehicle. -   Received vehicles that share the same direction of movement (yellow and blue vehicles in Fig. 1). The information might be collected from a GPS device or electronic compass onboard the vehicles in cluster   :    ∈   ,     ={   ,  =1,…, ℎ }  (2)    ⊂    where    is the cluster of received vehicles that share the same direction of movement with the node vehicle. -   Received vehicles located in safety interest area (blue vehicles in Fig. 1). This cluster needs to be formed only by those received vehicles that travel in front of main node   .     ( ∈   ,   ={   ,   =1,…  }  (3)    ⊂    ⊂    where F is the total number of received vehicles that travel in front of the main node and    the cluster of these vehicles. The purpose is to select from the    cluster only those vehicles travelling in front of the central node vehicle, in order to compute relative speed and safety distance between the central node and these vehicles. -   Received vehicles with a certain threshold of  persistence in the cluster. This condition is set to eliminate those vehicles travelling in the same direction with significant higher or lower speeds than the main node (vehicles on other lanes of the highway  – light blue vehicles in Fig. 1).    ∈   ,   =     ∑    >    (4) In relationship (4) above,    is the cluster of vehicles    with the property that the total number of per time unit received ID messages ∑  =1  exceeds a certain threshold,  , that is, these vehicles maintain a relative close position to the central node vehicle (they do not pass in short time due to their superior/inferior speed). The algorithm tries to evaluate each received vehicle’s affiliation to a specific cluster and then retrieves relevant dynamic information only from selected vehicles. For each vehicle belonging to such interest cluster    a set of information is collected: -   Relative speed:   =|      |,  =1,…,   (5) -   Relative acceleration / deceleration:   =      (6) -   Cluster average speed:   =  ∑     (7) -   Cluster average relative acceleration:   = ∑     . (8) The relative speed and relative acceleration/deceleration represent safety information used to build a weighted alarm function and to deliver early warning to the driver of the 370  central node vehicle in case of an event or emergency braking  produced by a vehicle situated in front. This information might also be displayed by the on-board computer of the main node vehicle. The weighted alarm function could be in the form presented in equation (9): Ψ  =   1  ∙ (    ∈   )+  2  ∙ (   >    )+  3  ∙ (   >    )  (9) In equation (9), Ψ   is the weighted alarm function,,  N   number of vehicles that fit the condition in quotes,  1 …  3  represent weighting factors,   and   coefficients of importance in relationship with the criteria chosen for alarming. However, there is not always a need for the local safety computed algorithm to determine if a critical braking condition has been issued: an accelerometer or brake sensor on each vehicle could also determine early warning message broadcasting. But in case of an incident, the brake is not manoeuvred or late manoeuvred by the driver. So it is recommendable to consider such an algorithm as secondary/backup safety information. IV.   B UILDING THE A LGORITHM   The above presented algorithm is figured below, in a simplified version:  AQUIRING INFORMATIONVehicle ID, heading, position, speed, message no.Determine Received Vehicles Cluster MrvDecision: Same travelling direction?Determine Vehicles travelling in the same direction cluster, M H YES Reject vehicles IDs NO Decision: Does vehicle belong to safety interest area?Determine Vehicles in the interest area cluster, M f  YES Decision: Is message received from vehicle persistant? YES Reject vehicles IDs NO Compute safety parametersDecision: Is condition critical? NOWARNBRAKEYES   Fig. 2. Algorithm for context awareness safety information  processing When estimating the delaying produced from the safety information collection until its actual usage, it can be considered a specific relationship. The onboard units of vehicles in cluster    (Fig. 1) – each mobile device use its own sensors (satellite navigation sensing, odometer and accelerometer) to detect dynamic information. Let’s denote this delay for local processing   τ lp1 . This includes (1): position detection delay   θ d , dynamic parameters processing delay θ p   τ lp1 =  θ d + θ p  (10) The success of the proposed solution relies also on the speed of the on-board information processing - therefore there is no need for infrastructure communications; however, it may require prioritisation of emergency messages in some cases. The next stage where time shifting occurs is the mobile V2X communication network. Delays in information transmission, in this case, are dependent on the density of users and availability of modern generation installations. The necessary time for the information to pass over the communication network is in this case:  2 =    (   +  )+   (   +    )  (11) where τ V2X  denotes the time shifting introduced by the ad-hoc local clustering, θ   and θ a  are delays in the network on the  broadcast link (broadcast of information) and back-link (acknowledge) paths, ξ   and ξ d  represent random delay factors on the broadcast and respectively on the acknowledge transmission paths, depending on the cluster coverage and accessibility in difficult reception areas (such as tunnels). Usually, ξ   and ξ d  induce considerable delays when the signal is lost and message repeating becomes necessary. The influence of the number of local users is represented in equation (9) by the random network congestion factor ν . For estimating the factor ν , in previous works [9] a model was represented by an adaptation of the Pollaczek-Khinchin formula, where the queuing delay becomes δ = θ� s +  λ  ∙  2 ( 1−λ⋅  )  . (12) In equation (10 ), δ stands for data packet delay at queuing, θ� s  represents the average service time at first transmission, θ� s2  the average service time at second transmission (retransmission) and λ the arrival rate of messages at the queue. Considering a coefficient for the congestion of the ad-hoc network due to multiple simultaneous transmission requests, we can further develop the model such as: ν =  =   +  ∙ ( ⋅ )    , (13) where  η ∈ (0.8÷0.95)  is a random coefficient directly dependant on the number of instant cluster users and τ� s  is a geometrically distributed random variable. The service time is  proportional to the number of the transmission attempts and inversely with the cluster (data packet) size. The coefficient  η  is containing this information. Finally, considering also the local algorithm processing time  2  (computing of received information, computing local safety conditions and decision regarding whether issuing or not an alarm), the total time until alarm information is available becomes:   =   1 +  2 +  2 .  (14) The solution has some advantages and disadvantages, as following: ã   Advantages: -   Single hopping / no need for communications infrastructure; -   It obtains permanent data on average cluster speed and relative speed/acceleration; -   Capability to detect situations where a driver did not  brake and a traffic incident has been triggered; -   Early warning in case of incident detection, helping reducing the casualties, injuries or vehicle damage; -   Possibility to extract information regarding vehicles gapping for ramp metering applications. ã   Disadvantages: -   All vehicles must be equipped with onboard units and sensors; -   Computing safety conditions requires additional timing; 371  -   The solution may need a specific communication  protocol in order to avoid messages collisions (similar to Self-Organising Time Division Multiple Access used by ships’ AIS 1 ), but prioritisation may  be introduced for critical safety messages via dedicated time slots; -   Prioritisation policies for relevant safety messages might be a good advance for the implementation of the proposed solution. V.   C ASE S TUDY   The model analyses a case with vehicles platoon travelling at on a highway with an average speed of 130 km/h and interspaced with 35 m each other, considering the parameters: TABLE   1 Vehicle’s dynamics related parameters Speed over ground 130 km/h 36.1 m/s Car length (automobile) 4 m Driver’s reaction min 0.75 s max 1.5 s Deceleration Normal 4.9 m/s Emergency 8 m/s DGPS accuracies (when available) Absolute position: 0.45 m Abs. heading: 1.5 o  Absolute time: 0.1 s Speed sensors accuracy (Odometer and/or GPS) ±10 km/h ±2.78 m/s TABLE 2 Communication/processing/driver reaction related parameters Background ITS traffic messages min 100 kb/s/veh max 1 Mb/s/veh Local dynamic  processing time min 0.5s max 1 s GPS rate 1s - Odometer rate 0.1 s - Driver reaction min 1.5 s max 2.5 s It is important to set the condition that the alarm delivery delay to central node vehicle is short enough for allowing safe manoeuvring the vehicle before collision (Fig. 3).    T   r   a   f   f   i   c   i   n   c   i   d   e   n   t Speed013010050[km/h]Incident locationSpace [m] Tdr T  A Vehicle 0Vehicle 1Vehicle iVehicle in central nodeDnDiD1 E ven t de tec t ion  tr iggers a lgor i t hm a larm ing DaSharp reducing in vehicle 0 cruising speed due to traffic incidentRapid growth of relative speed is detectedRelative speed to vehicle 0   Fig. 3. Diagram for incident early warning and speed control As it can be observed from the above figure, additional delaying is introduced before the vehicle in central node starts 1  AIS – Automatic Identification System emergency braking: if the system is only provided with early warning (without self-braking), then a driver reaction time is also to be considered:   = ∑   ∑     +    (15) with    - time to collision,    – distance between vehicles,    – individual length of vehicles and    driver reaction time. Then the condition to receive in due time the collision warning message will become:   <    . (16) Assuming parameters in Tables 1 and 2, it results that a safety distance from the first ahead traveling vehicle (in order to avoid collision) should be no less than 36.1 ∙  . VI.   C ONCLUSION   In this paper an algorithm for clustering vehicles in order to determine safety traffic conditions on a highway was  presented, along with the communications links analysis and  processes timings. Cooperative collision avoidance systems are presently in development for the ITS on-board structures. Compared to present solutions, the proposed algorithm may also be used as synoptic imaging of relationship between a central/critical node vehicle and the neighbouring vehicles, in order to assess the safety driving conditions. Future research in this field will allow experimental results and improvements of the proposed solution. R  EFERENCES   [1]   EU Commission, ”White paper on Transport – Roadmap to a Single European Transport Area – Towards a Competitive and Resource-Efficient Transport System”,  Directorate General for  Mobility and Transport  , Luxembourg, 2011, ISBN 978-92-79-18270-9. [2]   M. Emmelmann, B. Bochow, K. Kellum, Vehicular  Networking: Automotive Applications and Beyond  , John Wiley and Sons, 2010. [3]   C. M. Chou, C. Y. Li, W. M. Chien, K. C. Lan, “A Feasibility Study on Vehicle to Infrastructure Communication: WiFi vs. WiMAX”,  MDM’09 , Conference Proceedings , pp. 397-398, Taipei, Taiwan, 2009. [4]   A. Bohm, M. Jonsson, “Supporting Real-Time Data Traffic in Safety Critical Vehicle-to-Infrastructure Communication”,  LCN 2008 , Conference Proceedings , pp. 614-621, Montreal, Canada, 2008. [5]   S. Biswas, R. Tatchikou, F. Dion, ”Vehicle-to-Vehicle Wireless Communication Protocols for Enhancing Highway Traffic Safety”,  IEEE Communications Magazine , vol. 44, no. 1, pp. 74-82, 2006. [6]   M. R. Hafner, D. Cunningham, L. Caminiti, D. Del Vecchio, ”Cooperative Collision Avoidance at Intersections: Algorithms and Experiments”,  IEEE Transactions on ITS  , vol. 14, no. 3,  pp. 1162-1175, 2013. [7]   P. Fan, J. Haran, P. C. Nelson, ”Traffic Model for Clustering Algorithms in Vehicular Ad-Hoc Networks”, CCNC 2006  , Conference Proceedings , pp. 168-172, Las Vegas, NV, USA, 2006. [8]   M. Minea, I. Bădescu,  S. Dumitrescu, “Efficiency of Multimodal Real-time Traffic and Travel Information Services Employing Mobile Communications”, TELSIKS’11 , Conference Proceedings , pp. 765-768, Nis, Serbia, 2011.  372
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks