An Autonomous Information Collection and ... - IEEE Xplore

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 17, NO. 4, APRIL 2016

1085

An Autonomous Information Collection and Dissemination Model for Large-Scale Urban Road Networks Qi Zhang, Hao Zheng, Jinhui Lan, Jianwei An, and Hong Peng

Abstract—Autonomous vehicle traffic information systems are an important research direction for next-generation traffic information systems. Existing centralized traffic information systems involve a large initial investment and high operating costs. Furthermore, they suffer from the following problems: the need to communicate large amounts of data, requiring a longer time for road network coverage, unsteady transmission, and the need for an automatic generation and update method for road network congestion information in a large-scale urban road network. To overcome these problems, this paper proposes an intelligent vehicular traffic information system (IVTIS) based on a vehicular ad hoc network (VANET). This system employs a local road network and rapid dissemination model (IVTIS-LNFRN) of congestion information based on link nodes for a large-scale urban road network, and it constructs the corresponding system models. We then use traffic simulation software to evaluate the feasibility of the IVTIS. In particular, we investigate the collection, diffusion, and dissemination of congestion information and the automatic generation and update effect of road network congestion information. Furthermore, we analyze the dissemination effect of different traffic inflow volumes, information packet loss rates, and different rates of IVTIS vehicles. The simulation results show that the proposed system has good autonomy and overall performance in terms of the real-time collection and rapid dissemination of congestion information in a large-scale urban road network. Index Terms—Ad hoc, intelligent vehicle systems & telematics, intelligent transportation system (ITS), traffic information, traffic information systems.

I. I NTRODUCTION

C

ENTRALIZED traffic information systems have emerged as one of the promising solutions to the increasingly serious problem of urban traffic congestion. However, their overall service effect is restricted by the distribution density of Manuscript received August 14, 2014; revised December 25, 2014, April 10, 2015, and June 16, 2015; accepted October 18, 2015. Date of publication March 10, 2016; date of current version March 25, 2016. This work was supported in part by the Beijing Natural Science Foundation under Grant 4122048 and in part by the National Natural Science Foundation of China under Grant 61174181. The Associate Editor for this paper was Z. Liu. Q. Zhang, J. Lan, and H. Peng are with the School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing 100083, China (e-mail: [email protected]; [email protected]; [email protected]). H. Zheng is with the Institute of Software, Chinese Academy of Sciences, Beijing 100080, China (e-mail: [email protected]). J. An is with the School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing 100083, China (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TITS.2015.2497338

relevant facilities such as traffic data collection and information release and the infrastructure investment. Owing to the rapid development of electronic and communication technologies, carto-car communication-based traffic information systems have attracted considerable academic interest in many countries. Vehicular ad hoc network (VANET) is a representative car-to-car communication mode for information dissemination [1]–[3]. Among studies on traffic information dissemination in VANET, many have focused on traffic safety driving information [4]–[6], such as early warning information, assist information, and anti-collision control information for driving vehicles. Others [3], [7]–[9] have focused on the vehicular traffic congestion information service from the viewpoint of the automatic collection, integration, and dissemination of traffic congestion information. However, few studies have focused on the automatic generation and dynamic update of road network congestion information, especially in terms of the automatic collection, diffusion, and dissemination of traffic congestion information in a large-scale urban road network. When studying approaches for providing vehicles with traffic congestion information of the frontal road section through intervehicle communication, Helbing et al. noted that ant groups use a traffic traveling pattern to avoid road congestion when seeking food [10]. Accordingly, some studies focused on applying group intelligence in traffic and explored various approaches for transmitting traffic congestion information of the frontal road section through intervehicle communication [11], [12]. From the transmission pattern of traffic congestion information, drivers can determine the traffic congestion conditions of the frontal road section of interest. However, in this pattern, when a lane is a single one or oppositely facing vehicles do not travel through the frontal road section of interest, traffic congestion information of the frontal road section cannot be obtained from the oppositely facing vehicles. Therefore, this pattern has some limitations in real applications, especially for a road network. To collect and transmit traffic information in a road network, Wischhof et al. proposed a self-organized traffic system (SOTIS) [13], conducted traffic simulations, and demonstrated that information about traffic incidents could be transmitted over a 50-km road network within 6 min. Then, they studied expressways, proposed various information collection and transmission approaches for segment-oriented data abstraction and dissemination (SODAD) [14] based on wide-area wireless communication, and derived conclusions applicable to low penetration rates (2%–3%) through traffic simulations. However,

1524-9050 © 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1086


these approaches are not suitable for general roads. Furthermore, the collection and transmission of information containing much spatial information is not suitable for intervehicle communication over a short range. Based on SOTIS, some other approaches [15], [16] and different system models [17], [18] have been proposed for an urban road network. However, in a large-scale urban road network [19], when automatically transmitting traffic congestion information by vehicles, packet loss [20] often occurs owing to channel competition and congestion or signal interference. These problems are caused by the complex road network structure, existence of surrounding buildings and other obstacles [21], different vehicle flowing directions, vehicle flowing density, intersection and signal waiting times, etc. Broken chains of information dissemination in individual road sections or local road networks can also be caused by limited broadcast distance, uneven distances in the vehicle group, etc. To overcome these problems, some studies have simulated packet loss and developed stable transitive relations for the VANET protocol [19]–[21]. However, urban road networks are characterized by a large number of driving vehicles, complex vehicle running environments, and limited data processing capacity of vehicular information terminals. Therefore, it is challenging to realize real-time and simple traffic congestion information collection, a data fusion method with less communication and information processing, improved traffic information dissemination, and automatic generation and dynamic update of road network congestion information. Furthermore, it is difficult to analyze and evaluate the influence degree of the information packet loss rate and vehicular information terminal carrying rate on traffic information dissemination and diffusion. These major issues have not yet been investigated in detail. The present study aims to resolve some of them. The present study aims to reduce the time required for the collection, fusion, dissemination, and network coverage of traffic information. It focuses on a VANET-based intelligent vehicular traffic information system for an urban road network. It also proposes the IVTIS-LNFRN model that adds some link nodes in a local road network to rapidly spread information over a large-scale city road network.

II. IVTIS-LNFRN M ODEL A. Description of IVTIS This study proposes an IVTIS based on the information dissemination mechanism of VANET, structure mode of modern vehicular GPS navigation systems, and a comparison of the shortcomings of existing SOTISs. IVTIS can automatically collect and integrate traffic congestion information from the road section of interest and rapidly disseminate it with little interaction. Furthermore, it can automatically generate and dynamically update the entire road network’s real-time congestion information by constructing the corresponding system model. The vehicle information terminal includes a GPS unit, communication unit, CPU, displays, and other hardware com-

ponents. The software includes the following six functional modules: 1) Information collection module: This module uses GPS and road GIS to dynamically acquire the original road collection information of the vehicle itself. 2) Information fusion processing module: This module processes partial original information collected by the vehicle itself or others to obtain simplified congestion information of the road section on which the vehicle is driving. 3) Information dissemination module: This module releases or diffuses original information, simplified congestion information, and latest congestion information of the road network to cars in the same or other road sections. 4) Congestion information generation and updating module: The cars use the simplified congestion information collected and fused by themselves and receive information released or broadcasted by other cars to generate their own congestion information of the road network and conduct real-time updating. 5) Receive-and-send parallel communication module: This module receives and releases original information and congestion information of the road network based on the VANET broadcast protocol, thus enabling rapid exchange of traffic information among vehicles. 6) Information display module: This module displays the degree of congestion on in-vehicle GIS maps in different colors to provide a real-time reference for drivers. Fig. 1 shows the VANET-based IVTIS. B. Description of IVTIS-LNFRN The dissemination time required for congestion information collected by the IVTIS model increases with the scale of the road network. In particular, when the number of vehicles is too low, the time taken for congestion information to cover the whole road network will increase greatly, thus affecting the real-time nature of the information. Therefore, based on an overall analysis of the IVTIS-model-based autonomous collection of congestion information and the coverage effect over the road network, this study proposes a node-link-based road network dissemination model for large-scale urban road networks. This LNFRN approach enables the IVTIS-modelbased congestion information to cover the entire urban road network rapidly via several remote link nodes. Fig. 2 shows the dissemination of congestion information under IVTIS-LNFRN. Link nodes can be fixed to central sections or intersections of the local road network, temporarily moving things, or vehicles running in the local road network (e.g., buses, etc.). In addition to being IVTIS vehicle information terminals, they also communicate information to other nodes (such as by SMS or Internet). Therefore, they allow congestion information from any part of the urban road network to be disseminated rapidly across the entire domain. III. IVTIS M ODEL Based on the above IVTIS architecture, we construct the corresponding system model as follows.

ZHANG et al.: AUTONOMOUS INFORMATION COLLECTION AND DISSEMINATION MODEL FOR ROAD NETWORKS

1087

Fig. 1. Schematic illustration of VANET-based IVTIS.

road section to calculate the average vehicle speed at the time of information collection in real-time with accuracy. To combine the advantages of these three methods, this study uses collection method 3) and uses a further treatment in the fusion process.

B. Fusion Processing Model of Traffic Congestion Information

Fig. 2. Dissemination of congestion information based on IVTIS-LNFRN.

A. IVTIS’s Autonomous Collection Method IVTIS’s autonomous collection method mainly determines which road section the car is driving on, distance traveled, and travel time based on GPS and the road GIS. This allows the calculation of the average speed, which is considered the original information in this section. Cars (set of collection vehicles for original information) interact with their original information through VANET broadcasts to form a set of original information owned by the cars on a specific road section. The original information is mainly obtained as follows: 1) determining the instantaneous vehicle speed at a specific time or location in real-time but with low representativeness; 2) calculating the average speed according to the time taken by the vehicle to pass the information collection section, with high accuracy but a certain lag; and 3) allowing moving vehicles to obtain the distance traveled and travel time of vehicles on that

To allow moving vehicles to obtain the distance and travel time of vehicles after entering an information collection section at a certain time interval, because vehicles entering the section have traveled different distances, the corresponding representativeness of their information to the overall traffic information for this section differs. Therefore, in this study, we assigned representative weights to the collected information according to the distance traveled and then conducted the fusion process accordingly. The information collection and fusion processing model is as follows: V k (i) =

lp (i)

p∈Ok (i)

tp (i)

,

i ∈ {1, 2, . . . , N }

p∈Ok (i)

or V k (i) =

p∈Ok (i)

wp (i) · tp (i)

p∈Ok (i)

or V k (i) =

wp (i) · lp (i) =

(1) lp2 (i)

p∈Ok (i)

lp (i) · tp (i)

(2)

p∈Ok (i)

wp (i) · Vp (i)

p∈Ok (i)

p∈Ok (i)

wp (i)

(3)

1088


where lp (i) , p ∈ Ok (i), i ∈ {1, 2, . . . , N } tp (i) tp (i) = tp,c (i) − tp,s (i), p ∈ Ok (i), i ∈ {1, 2, . . . , N } lp (i) wp (i) = , p ∈ Ok (i), i ∈ {1, 2, . . . , N }. L(i)

road section is clear, a value of “−1” is used to indicate that congestion has eased.

Vp (i) =

Here, V k (i) represents the average speed of vehicle k on section i after the fusion processing of the set of original information. lp (i) and tp (i) respectively represent the distance traveled and travel time of vehicle p after entering section i. Ok (i) represents the set of collection vehicles for original information owned by vehicle k on section i. N represents the total number of sections in the road network. wp (i) is the representative weight of the information collected by vehicle p. Vp (i) represents the sectional traffic speed at the collection time when vehicle p is moving on section i. tp,c (i) represents the information collection time of vehicle p on section i. tp,s (i) represents the time for which vehicle p has been moving through section i. L(i) represents the length of section i. To reduce the amount of data processing at the vehicle information terminal and the computational burden of information exchange among vehicles, IVTIS mainly processes information using three simplified methods. The first method acts on the premise of not affecting the fusion precision of the collected information. An upper limit is set for the number of collection vehicles for original information in Ok (i). When the number of vehicles exceeds the upper limit, information from the newest vehicle replaces that from the vehicle that has been on that road section for the greatest length of time. This reduces the amount of information exchanged by communication and the volume of fusion data processing. The second method only releases congestion information (or congestion removal information) of sections in which there are traffic jams. This reduces the amount of information dissemination and transmission. The third method refers to the speed range partition method of the congestion degree by the existing centralized TIS (such as VICS in Japan) to determine the degree of congestion. It publishes the collected information following fusion processing in the form of simplified congestion information as follows: ⎧ ⎪ (Congestion) ⎨1 0 ≤ Vk (i) < Vc (4) Vk (i) = 2 Vc ≤ Vk (i) < Vs (Slow Move) ⎪ ⎩ −1 Vk (i) ≥ Vs (Congestion Removal). Here, Vk (i) represents the simplified congestion information after the fusion of information collected by vehicle k on section i. Vc and Vs represent the upper speed limits for traffic congestion and slow moving, respectively. When V k (i) ≥ Vs , road section i is clear. However, there are two situations: 1) Vehicle k has not received the simplified congestion information before entering section i, and the section is open at that time. If so, it is unnecessary to release any information. 2) Vehicle k has received the simplified congestion information, and this section is jammed. The current vehicle information terminal will maintain this state. Then, when the

C. Traffic Information Dissemination Model The information collection vehicles on certain sections release traffic information to other vehicles through VANET broadcasts. Traffic information is divided into three categories according to different vehicle objectives: 1) to release the original information sent to other vehicles in the same section, including the original traffic information the vehicle collects by itself and that received from other vehicles on the same section through VANET; 2) to release simplified congestion information or congestion removal information after the latest fusion processing to vehicles on other sections; and 3) to release the latest simplified congestion information or congestion removal information received and broadcasted by different vehicles through VANET. Thus, the dissemination model for traffic information released by information collection vehicle k on section i can be described as follows: Bk (i) = Sk (i) ∪ Dk (i) ∪ Rk (i),

i ∈ {1, 2, . . . , N }

(5)

where Sk (i) = {lp (i), tp (i), tp,c (i)|p ∈ Ok (i)} , i ∈ {1, 2, . . . , N } (or) {Vp (i), wp (i), tp,c (i)|p ∈ Ok (i)} , i ∈ {1, 2, . . . , N } (6) (7) Dk (i) = {Vk (i), tk,r (i)} , i ∈ {1, 2, . . . , N } Rk (i) = Vq (j), tq,r (j)|q = k, j = i, j ∈ Gk (i) , i ∈ {1, 2, . . . , N }, Gk (i) ⊆ {1, 2, . . . , N }. (8) Here, Bk (i) represents all traffic information owned and released by vehicle k on section i. Sk (i) represents the original information set owned by information collection vehicle k on section i. Dk (i) represents the simplified congestion information of vehicle k on section i after the information fusion process. Rk (i) represents the latest simplified congestion information about section j broadcast by vehicle k on section i. tk,r (i) and tq,r (j) respectively represent the times when vehicle k on section i and vehicle q on section j initially release the simplified congestion information after information fusion processing. Gk (i) represents the section with the simplified congestion information set owned by vehicle k on section i. D. Generation of Congestion Information for Road Network Through fusion processing of their original information set and the receipt of simplified congestion information released by other vehicles via VANET, vehicles can generate and update congestion information for the road network using their information terminals. In addition, as the interaction with other vehicles increases, the set of congestion information becomes increasingly complete, so as to provide a real-time reference to drivers. The congestion information generation model is as follows: Nk (i) = Dk (i) ∪ Rk (i) = {Vk (i, j), tk (i, j)|j ∈ {1, 2, . . . , N }} (9)


1089

by collection vehicle k of section m, then add the corresponding information to vehicle k.

where when j=i thenVk (i, j) = Vk (i), tk (i, j) = tk,r (i) when j=i

thenVk (t,j) = Vq (j), tk (i,j) = tq,r (j), (q = k, j∈Gk (i))

tk (i, j) ≤ t i, j ∈ {1, 2, . . . , N },

Gk (i) ⊆ {1, 2, . . . , N }. (10)

Here, Nk (i) is the congestion information set of all road networks owned by vehicle k on section i. Vk (i, j) and tk (i, j) respectively represent the simplified congestion information of section i and the time of initial release of the congestion information set of all road networks owned by vehicle k on section i. t is the current time. E. Vehicular Traffic Information Updating Model With the movement of vehicles across the road network and the dynamic alteration of VANET, all vehicles continually receive the original information set of the same collection section and simplified congestion information of the road network and update their own information in real time. According to different situations, the dynamic updating model for the congestion information in the road network will be one of the following: 1) Update Original Information Set: • If vehicles k and k belong to the same information collection section i and the original information set of vehicle k has original information that vehicle k does not have, we update the original information set and corresponding information of vehicle k as follows: When Ok (i) = Ok (i) / Ok (i) i.e., p ∈ Ok (m), p ∈ then Ok (i) = Ok (i) ∪ Ok (i) Sk (i) = Sk (i) ∪ {Sk (i)|p ∈ / Ok (i)} where k = k , i ∈ {1, 2, . . . , N }.

(11)

• When the original information sets of vehicles k and k both have the original information of a vehicle in the same section and the information owned by vehicle k is new, update the information in vehicle k. When p ∈ Ok (i), p ∈ Ok (i), p = p ∈ Ok (i) ∩ Ok (i) and lp (i) < lp (i) or tp (i) < tp (i) or Vp (i) = Vp (i), tp,c (i) < tp ,c (i) then lp (i) = lp (i), tp (i) = tp (i) or Vp (i) = Vp (i), wp (i) = wp (i) and tp,c (i) = tp ,c (i) where i ∈ {1, 2, . . . , N }.

(12)

2) Update Congestion Information Set of Road Network: • If collection vehicle k on section i does not have simplified congestion information of section j, which is owned

/ Gk (i) When j ∈ Gk (m), j ∈ i.e., Vk (i, j) = Φ and Vk (m, j) = Φ then Vk (i, j) = Vk (m, j) tk (i, j) = tk (m, j) Gk (i) = Gk (i) ∪ Gk (m) where k = k, j = i Gk (i) ⊂ {1, 2, . . . , N }, Gk (m) ⊆ {1, 2, . . . , N }, i, j, m ∈ {1, 2, . . . , N } (13) where Vk (i, j) = Φ indicates that vehicle k on section i presently has no information about section j (j = i). • If collection vehicle k on section i and collection vehicle k on section m both have simplified congestion information of section j, but the information of vehicle k is new, update the information in vehicle k as follows: When j ∈ Gk (i) and j ∈ Gk (m) i.e., Vk (i, j) = Φ, Vk (m, j) = Φ and tk (i, j) < tk (m, j) then Vk (i, j) = Vk (m, j) tk (i, j) = tk (m, j) where k = k, j = i Gk (i), Gk (m) ⊆ {1, 2, . . . , N } i, j, m ∈ {1, 2, . . . , N }.

(14)

3) Time for Which Vehicle Traffic Information Should be Preserved: • Preservation time for the original information set: When information collection vehicle k moves out of collection section i and into the next section, the original information set for section i and the collection vehicle set are cleared immediately and replaced by the corresponding information set of the next section. • Time limit for preserving simplified congestion information: When the simplified congestion information of a section owned by vehicle k is not updated within the prescribed time (time limit), the traffic information becomes unclear and has no reference value. In this case, the system eliminates the information automatically as follows: When t − tk (i.j) ≥ T0 then Vk (i, j) = Φ, tk (i.j) = Φ / Gk (i)} Gk (i) = {Gk (i)|j ∈ where j = i, Gk (i) ⊆ {1, 2, . . . , N }, i ∈ {1, 2, . . . , N } (15) where T0 is a pre-set time limit for preserving simplified congestion information of the road network. 4) Pretreatment Update Information: When the IVTIS model updates, if the received data contains errors, it may cause some biased results and degrade the accuracy of the road network congestion information in the original information fusion. Thus, before updating the data, the vehicle information terminal will pre-treat the new information received to eliminate the influence of possible inaccuracies. We now present a relatively simple method.

1090


Fig. 3. Collection, fusion, and broadcast publishing process of IVTIS model.

For the original acquisition information received, according to the traffic laws and general knowledge, the section of travel distance, travel speed, and travel time information that is beyond the normal range is removed. Then, the upper limit is increased properly for the number of collection vehicles in the original information set Ok (i), and the impact of erroneous data on the information fusion results is reduced. For some sections for which road network congestion information is received, there is more than one congested road information publishing source (vehicle). Therefore, the error messages are compared with the correct messages, and the rate is less; therefore, the probability of repeated information being received is less. Therefore, a newly received message is first put in a pretreatment set, and then, the number of messages having the same simplified congestion information in the set is compared. The preset value (≥ 2) is used to decide whether to subsequently update (and clear the set) the congestion information of the section. By this method, an error message can be removed. Even if the error message is used to update information (low probability), it will soon be updated by the correct information received subsequently. F. IVTIS Broadcast Publishing Model The information broadcast by IVTIS vehicles is processed by the receive-and-send parallel communication module of the vehicle information terminals in all vehicles. When the vehicle enters a new section, until it enters the next section (see Fig. 3, T1 ≤ t < T4 ), it will always receive the

traffic information sent by other vehicles, expand and update its own information according to Equations (11)–(14), and simultaneously release this information to others in accordance with a certain broadcast interval (Tr ). During this period, when the vehicle is collecting the information (see the time corresponding to the black dot in Fig. 3), automatically update the original information with the latest collected information and send the simplified congestion information of the section fused with the other traffic information. The collection, fusion, and sending of traffic information is divided into three main stages (shown in Fig. 3): 1) T1 ≤ t < T2 : When a vehicle enters an information collection section, the traffic information collected at this time does not represent the actual situation. IVTIS sets a certain time interval (Ts ) or distance interval (Ls ) and requires vehicles to collect information when they exceed one of these intervals. Prior to this, they do not collect the information. 2) T2 ≤ t ≤ T3 : When a vehicle drives to the time corresponding to Ts or Ls , it begins to collect information and will collect information at a regular interval (Tc ). Then, it will update the latest collection information to its own original information set in time and conduct fusion processing according to Equations (1)–(3) (where Ok (i) = Ok (i) ∪ k). The simplified congestion information by the fusion result of the collect section is released with other traffic information and broadcast repeatedly at a regular interval (Tr ) until the next


collection begins after a certain time interval (Tc ). This process is repeated until the vehicle moves out of the current information collection section. The broadcast publishing model is as follows: When tk,cs (i) + nTc < t tk,cs (i) + (n + 1)Tc and t = (tk,cs (i) + nTc ) + mTr , (m = 1, 2, . . . , M ) then broadcast the following information (with reference to Equations (5)–(8)): Bk (i) = Sk (i) ∪ Dk (i) ∪ Rk (i), i ∈ {1, 2, . . . , N } where n = 0, 1, 2, . . . , Ni − 1; i ∈ {1, 2, . . . , N } (16) M = Tc /Tr (M ≥ 1). Here, Ni is the interval number for collection on section i. M refers to the broadcasting times within a collection interval Tc . Tc and Tr respectively denote the information collection interval and broadcast interval. tk,cs (i) is the time when the information collection starts after the vehicle k drove into the section i. 3) T3 ≤ t ≤ T4 : When a vehicle drives out of the current information collection section and performs the last collection, it repetitively publishes the fused simplified congestion information in accordance with the broadcast interval (Tr ) before entering the next section (intersections and so on during this period). With regard to the listed broadcast interval (Tr ), when the vehicle enters a new section, it arrives at the time corresponding with Ts or Ls , and when it leaves the collection section, the time is reset to the starting point. Fig. 3 shows the collection, fusion, and broadcast publishing process of the IVTIS model. G. IVTIS-LNFRN-Based Dissemination Model In the collection and dissemination of congestion information based on the IVTIS model, the real-time nature of congestion information is affected by the lag time of collection, time required for diffusion processing and VANET receiving and sending, release interval of broadcast information, and dissemination time of congestion information. This is because the information gap caused by chain breaking in information transmission causes propagation time delay. The former factors can be solved using real-time collection, minimum information interaction, an upper limit on the number of vehicles, high-performance vehicle processors, and receiveand-send parallel communication module adopted by the IVTIS model. The most serious factor affecting real-time information dissemination is the scale of the road network. The larger the road network, the longer is the lag time in disseminating road conditions and the further from real-time the system becomes. To apply the IVTIS model to large-scale urban road networks, we propose the long-distance interaction of congestion information among local road networks through link nodes in the urban road network. We utilize simulations to evaluate the coverage effect in the local road network (see Fig. 2) and demonstrate that our proposed approach allows for the

1091

autonomous collection and rapid dissemination of congestion information across large-scale urban road networks. The link nodes should be positioned according to two principles to allow congestion information to be spread rapidly over the entire road network. First, the sum of the effective coverage of all link nodes should match the overall coverage of the urban road network [see (17)]. Second, the dissemination time of the congestion information should satisfy the drivers’ expectations of the real-time nature of congestion information [see (18)]. The dissemination model of IVTIS-LNFRN is as follows: N net

Cnet (l) = Snet

(17)

l=1

tnet ≤ 2Tsub ≤ Tnet.

(18)

Here, Cnet (l) is the effective coverage of the lth link node in the local road network (total length). Nnet is the number of link nodes set in the entire urban road network. Snet is the scale of the entire urban road network (total length). tnet is the time taken for congestion information to cover the entire road network. Tsub is the effective coverage time of the congestion information released by link nodes in the local road network (and similarly when receiving information). Tnet is the maximum value of the dissemination time of congestion information for the entire road network as recognized or accepted by drivers. Thus, the IVTIS-LNFRN-based congestion information dissemination model enables congestion information released by any car in the urban road network to cover the entire network in two times the effective dissemination time required for the local road network (equal to the dissemination time of the diameter of the local road network). This model is attractive for urban road networks in large- and medium-sized cities. IV. T RAFFIC S IMULATION AND E VALUATION To verify and evaluate the rationality and effectiveness of the IVTIS-LNFRN model, we simulated the corresponding functions of the IVTIS models through the secondary development platform (COM interface) in VISSIM; furthermore, IVTIS models could be simulated and evaluated successfully. A. Simulation Environment Settings Fig. 4 shows a schematic diagram of the simulation environment. Table I shows all the parameter settings for the simulation. B. Simulation and Evaluation of IVTIS Models 1) Shortest Time for Disseminating Traffic Congestion Information (i.e., Shortest Time to Disseminate Traffic Congestion Information to Somewhere in the Road Network): Fig. 4 shows the shortest dissemination time for simplified congestion information released by the information collection vehicles on congested sections to all intersections and entrances of the road network when the traffic inflow volume is 1200 veh/h. It is observed that the intersections allow congestion information to be rapidly disseminated through several paths and thus cover the entire road network. For example, only 26 s are required for congestion information released by vehicles on the congested section to

1092


Fig. 4. Diagram of shortest dissemination time for congestion information in road network at inflow rate of 1200 veh/h. TABLE I S IMULATION PARAMETER S ETTINGS

Fig. 5. Shortest dissemination time for congestion information to reach (a) intersections and (b) entrances to the road network for different vehicle inflow volumes.

disseminate to all parts of the road network (longest dissemination path is ∼7.5 km, to Entrance 5). 2) Analysis of Influence of Vehicle Inflow Volume on Dissemination Effect of Congestion Information: Fig. 5 shows the results of the dissemination effect under different vehicle inflow volumes in the road network. With an increase in the vehicle inflow volume into the road network, the vehicle density in the road network increases. As a result, the chain-scission phenomenon caused by uneven vehicle spacing is reduced, and it causes a reduction in the shortest time required for congestion information to disseminate to each intersection and entrance. When the vehicle inflow volume is small (300 veh/h), the lower vehicle volume density and uneven vehicle spacing lead to a significant increase in the dissemination time for congestion information. 3) Analysis of Influence of Number of IVTIS Vehicles on Dissemination of Congestion Information: To analyze the effect of the number of IVTIS vehicles on the dissemination of congestion information, we simulated IVTIS vehicle rates of 100%, 50%, and 25% at a vehicle inflow rate of 1200 veh/h.

Fig. 6. Coverage time of congestion information under different IVTIS vehicle rates.

• Coverage time for intersections and road network: Fig. 6 shows a comparison of the shortest coverage times required for congestion information to reach all intersections and entrances of the road network under different IVTIS vehicle rates. As mentioned previously, many paths can disseminate congestion information to the intersections. Therefore, reducing the IVTIS vehicle rate has little influence on the coverage time for the intersections. For the entire road network, as the dissemination time of congestion information from the edge intersections to entrances in the road network is unstable, when the IVTIS vehicle rate is reduced to 25%, the possibility of the information chain-scission phenomenon increases, and therefore, the coverage time increases to 77 s. • Coverage rate of congestion information in road network: Fig. 7 shows a comparison of the coverage rate of congestion information in the road network over time with different IVTIS vehicle rates. We observe that the


1093

Fig. 7. Coverage rate of congestion information under different IVTIS vehicle rates.

Fig. 9. Diagram of dissemination of congestion information in a large-scale urban road network based on fixed link nodes (solid circles denote link nodes and dotted circles, their coverage radius).

Fig. 8. Road network coverage of congestion information for different packet loss rates.

coverage rate is greater than 95% in all situations some 50 s after the congestion information has been released. The larger the IVTIS vehicle rate, the shorter is the time taken to disseminate traffic congestion information across the road network. According to the above results, the coverage time for the intersections is shorter than that for the road network. Therefore, the IVTIS system model proposed in this study is suitable for urban road networks in which intersections are the main elements. The coverage time for the intersections shown in Fig. 6 suggests that when the IVTIS vehicle rate is 25%, the shortest coverage time for congestion information to reach all intersections is 25 s (the distance to the farthest intersection is 5 km; see Fig. 4). In an ideal state, the congestion information released by any vehicle in a congested section should spread to all intersections of a large-scale urban road network with a diameter of ∼25 km in 2–3 min (such as within five rings of Beijing, see Fig. 9). 4) Analysis of Influence of Packet Loss Rate on Dissemination of Congestion Information: Fig. 8 shows a comparison of the road network coverage rate of traffic congestion information for different packet loss rates with time for vehicle inflow volume of 600 veh/h and IVTIS vehicle mixed rate of 100%. This figure shows that with the increase of packet loss rate, the time required for the coverage of the entire road network by congestion information increases accordingly. This is because

the increase in packet loss rate promotes broken chains in information dissemination. This shows that the packet loss rate greatly influences the dissemination of congestion information. To reduce the packet loss rate at the time of information dissemination, except for research on a more stable VANET routing protocol, we still need to study a more simple and effective information integration and dissemination method involving less but sufficient information dissemination. Furthermore, we aim to improve the performance and power of the vehicular communication terminal. C. Analysis of Dissemination Effect in Large-Scale Urban Road Network Based on IVTIS-LNFRN To increase the dissemination speed and its relative stability of congestion information in a large-scale road network and enhance the real-time nature of information dissemination, we propose that the large-scale road network should be divided into several local road networks. The corresponding fixed IVTIS nodes should be located at the center of the local road networks to allow congestion information to spread among the link nodes in the local road networks through simple remote communication. This will ensure that congestion information autonomously collected and released in a certain local road network is rapidly spread across all local road networks, thus covering the entire urban road network. Fig. 9 shows an example of the link nodes (solid circles indicate link nodes and dotted circles, their coverage radius). According to Figs. 4 and 9, if the shortest dissemination time to the link node from anywhere in the local road network within a radius of 5 km is less than 25 s and the stable effective coverage time is 30 s, then just 20 fixed IVTIS link nodes are needed to ensure that congestion information covers the entire network (as shown in Fig. 9) within 1 min. The dissemination speed and coverage time for the road network are

1094


more stable and rapid than for completely autonomous IVTIS dissemination. Thus, the IVTIS-LNFRN model not only realizes the autonomous collection and dissemination of congestion information under special conditions but also provides real-time and accurate traffic information. In particular, congestion information is disseminated over special sections such as the blind spots between link nodes and the existing centralized TIS. V. C ONCLUSION This study proposes an IVTIS model that can autonomously collect and disseminate congestion information in urban road networks. We analyzed the influence of the vehicle inflow volume, IVTIS vehicle rate, and packet loss rate on the dissemination of congestion information in the road network. Accordingly, we have proposed a link-node-based IVTIS-LNFRN model that is suitable for application to large-scale urban road networks. Based on the simulation results of the IVTIS model, this study further explores the dissemination effect of congestion information in large-scale urban road networks based on the IVTIS-LNFRN model. The results showed that this model is very suitable for the autonomous collection and rapid dissemination of congestion information. Moreover, it can be used as a supplement to existing traffic information systems. It plays an effective role in the collection of information on non-primary roads, especially narrow roads and blind spots. In future research, we will consider the influence of the broadcasting interval, routing selection, special road network structure, and actual traffic distribution on the dissemination of congestion information in detail. We will also aim to avoid information loss. In addition, we will develop a secure vehicle ID system to avoid indirect disclosure of vehicle private information. Furthermore, we need to test the proposed method on an actual road network to determine the feasibility, reliability, and validity of the IVTIS-LNFRN approach and to identify problems and make improvements. N OTATION L IST V k (i)

lp (i) and tp (i) Ok (i) N wp (i) Vp (i) tp,c (i) tp,s (i) L(i)

the average speed of vehicle k on section i after fusion processing of the set of original information; the distance traveled and travel time of vehicle p after entering section i, respectively; the set of collection vehicles for original information owned by vehicle k on section i; the total number of sections in the road network; the representative weight of the information collected by vehicle p; the sectional traffic speed at the collection time when vehicle p is moving on section i; the information collection time of vehicle p on section i; the time for which vehicle p has been moving through section i; the length of section i;

Vk (i)

the simplified congestion information after the fusion of information collected by vehicle k on section i; Vc and Vs the upper speed limits for traffic congestion and slow moving vehicles, respectively; Bk (i) all traffic information owned and released by vehicle k on section i; the original information set owned by inforSk (i) mation collection vehicle k on section i; Dk (i) the simplified congestion information of vehicle k on section i after the information fusion process; Rk (i) the latest simplified congestion information about section j broadcast by vehicle k on section i; tk,r (i) and tq,r (j) the times when vehicle k on section i and vehicle q on section j initially release the simplified congestion information after information fusion processing, respectively; Gk (i) the section with the simplified congestion information set owned by vehicle k on section i; Nk (i) the congestion information set of all road networks owned by vehicle k on section i; Vk (i, j) and tk (i, j) the simplified congestion information of section i and the time of initial release of the congestion information set of all road networks owned by vehicle k on section i, respectively; t the current time; T0 pre-set time limit for preserving simplified congestion information of the road network; Ts the time interval before collection; Ls the distance interval before collection; the time when the information collection tk,cs (i) starts after the vehicle k drove into the section i; Ni the interval number for collection on section i; M the broadcasting times within a collection interval Tc ; Tc and Tr the information collection interval and the broadcast interval, respectively; Cnet (l) the effective coverage of the lth link node in the local road network (total length); the number of link nodes set in the entire Nnet urban road network; Snet the scale of the entire urban road network (total length); tnet the time taken for congestion information to cover the entire road network; Tsub the effective coverage time of the congestion information released by link nodes in the local road network (and, similarly, when receiving information); Tnet the maximum value of the dissemination time of congestion information for the entire road network as recognized or accepted by drivers.


R EFERENCES [1] R. Schmilz, A. Leiggener, A. Festag, L. Eggert, and W. Effelsberg, “Analysis of path characteristics and transport protocol design in vehicular ad hoc networks,” in Proc. IEEE 63rd VTC-Spring, May 2006, pp. 528–532. [2] E. Schoch, F. Kargl, M. Weber, and T. Leinmuller, “Communication patterns in VANETs,” IEEE Commun. Mag., vol. 46, no. 11, pp. 119–125, Nov. 2008. [3] H. Hartenstein and K. P. Laberteaux, VANET: Vehicular Applications and Inter-Networking Technologies. Hoboken, NJ, USA: Wiley, 2010. [4] F. Ahmed-Zaid et al., “Vehicle Safety Communication-Applications (VSC-A) final report: Appendix volume 3 Security,” Nat. Highway Traffic Safety Admin., Washington, DC, USA, Sep. 2011. [5] E. Garcia-Lozano, C. Tripp Barba, M. Aguilar Igartua, and C. Campo, “A distributed, bandwidth-efficient accident prevention system for interurban VANETs,” in Proc. IEEE Int. Conf. SaCoNeT, Jun. 2013, pp. 1–5. [6] Y. B. Liu, C. L. Chen, and X. P. Guan, “A hovering-based warning information dissemination approach in highway entrances,” in Proc. IEEE ICC, Jun. 2014, pp. 2719–2724. [7] T. Nadeem, P. Shankar, and L. Iftode, “A comparative study of data dissemination models for VANETs,” in Proc. Int. Conf. Mobile Ubiquitous Syst., Netw. Serv., Jul. 2006, pp. 1–10. [8] S. Dornbush and A. Joshi, “StreetSmart traffic: Discovering and disseminating automobile congestion using VANET’s,” in Proc. IEEE 65th VTCSpring, 2007, pp. 11–15. [9] S. P. Fekete, C. Schmidt, A. Wegener, H. Hellbruck, and S. Fischer, “Empowered by wireless communication: Distributed methods for selforganizing traffic collectives,” ACM Trans. Auton. Adapt. Syst., vol. 5, no. 3, Sep. 2010, Art ID 11. [10] A. Dussutour, V. Fourcassié, D. Helbing, and J. L. Deneubourg, “Optimal traffic organisation in ants under crowded conditions,” Nature, vol. 428, pp. 70–73, Jan. 2004. [11] A. H. John, A. Schadschneider, D. Chowdhury, and K. Nishinari, “Characteristics of ant-inspired traffic flow: Applying the social insect metaphor to traffic models,” Swarm Intell., vol. 2, no. 1, pp. 25–41, Mar. 2008. [12] W. Narzt, U. Wilflingseder, G. Pomberger, D. Kolb, and H. Hörtner, “Selforganizing congestion evasion strategies using ant-based pheromones,” Intell. Transport Syst., vol. 4, no. 1, pp. 93–102, Mar. 2010. [13] L. Wischhof, A. Ebner, H. Rohling, M. Lott, and R. Halfmann, “SOTIS: A self-organizing traffic information system,” in Proc. IEEE 57th VTCSpring, Apr. 2003, pp. 2442–2446. [14] L. Wischhof, A. Ebner, and H. Rohling, “Information dissemination in self-organizing intervehicle networks,” IEEE Trans. Intell. Transport Syst., vol. 6, no. 1, pp. 90–101, Mar. 2005. [15] E. David et al., “Simulative performance evaluation of the sim self organizing traffic information system,” in Proc. 10th IFIP Med-Hoc-Net, 2011, pp. 79–86. [16] S. Panichpapiboon and W. Pattara-Atikom, “A review of information dissemination protocols for vehicular ad hoc networks,” IEEE Commun. Surveys Tuts., vol. 14, no. 3, pp. 784–798, 3rd. Quart. 2012. [17] X. Yang and W. Recker, “Evaluation of information applications of a self-organizing distributed traffic information system for a large-scale real-world traffic network,” Comput-Aided Civil Infrastruct. Eng., vol. 23, no. 8, pp. 575–595, Nov. 2008. [18] Q. Zhang and J. H. Zhao, “A model for automatic collection and dynamic transmission of traffic information based on VANET,” in Proc. 15th IEEE ITSC, Sep. 2012, pp. 373–378. [19] H. Cheng, X. Fei, A. Boukerche, A. Mammeri, and M. Almulla, “A geometzry-based coverage strategy over urban VANETs,” in Proc. 10th ACM Symp. Perform. Eval. Wireless Ad Hoc, Sensor, Ubiquitous Netw., Nov. 2013, pp. 121–128. [20] M. N. Majeed, S. P. Chattha, A. Akram, and M. Zafrullah, “Vehicular ad hoc networks: History and future development arenas,” Int. J. Inf. Techno. Elect. Eng., vol. 2, no. 2, pp. 25–29, Apr. 2013. [21] A. M. Mezher et al., “Realistic environment for VANET simulations to detect the presence of obstacles in vehicular ad hoc networks,” in Proc. 11th ACM Symp. Perform. Eval. Wireless Ad Hoc, Sensor, Ubiquitous Netw., Sep. 2014, pp. 77–84.

1095

Qi Zhang received the M.S. degree in system engineering from Tianjin University, Tianjin, China, in 1985 and the Ph.D. degree in civil engineering from Osaka City University, Osaka, Japan, in 1998. He is currently an Associate Professor with the School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing, China. His research interests include intelligent transportation systems, vehicular traffic information systems, and systems engineering.

Hao Zheng received the M.S. degree in electronic engineering from the University of Science and Technology Beijing, China, in 2013. He is currently a Big Data Development Engineer with the Institute of Software, Chinese Academy of Sciences, Beijing, China. His research interests include intelligent transportation systems, traffic simulation, and software engineering.

Jinhui Lan received the Ph.D. degree from Beijing Institute of Technology, Beijing, China, in 1998. In 1998, she joined Tsinghua University, Beijing, as a Lecturer. From August 2002 to February 2004, she was a Visiting Academic with Deakin University, Geelong, Australia, working on multisensor systems. Since 2008, she has been a Professor with the School of Automation and Electrical Engineering, University of Science and Technology Beijing, China. From February 2013 to August 2013, she was a Visiting Scholar with the University of Wisconsin–Madison, Madison, WI, USA. In cooperation with other researchers, she has successfully completed more than 30 projects and solved some key technical problems. She is the author or coauthor of over 80 refereed papers in image processing, ITS, measurement and instrument, and multisensor data fusion.

Jianwei An received the Ph.D. degree from Beijing University of Technology, Beijing, China, in 2006. She is currently an Associate Professor with the School of Computer and Communication Engineering, University of Science and Technology Beijing, China. Her research interests include ad hoc networks, cognitive networks, and wireless communication systems.

Hong Peng is currently working toward the M.S. degree with the School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing, China. His research interests include intelligent transportation systems and intervehicle communication.