20th IEEE Symposium on Computers and Communication (ISCC)
An Intelligent Transportation System for Detection and Control of Congested Roads in Urban Centers Celso A. R. L. Brennand∗¶ , Allan M. de Souza∗ , Guilherme Maia† , Azzedine Boukerche‡ , Heitor Ramos§ , Antonio A.F. Loureiro† and Leandro A. Villas∗ ∗ Institute
of Computing, University of Campinas, Brazil of Computer Science, Federal University of Minas Gerais, Brazil ‡ PARADISE, University of Ottawa, Canada § Institute of Computing, Federal University of Alagoas, Brazil ¶ UAST, Federal Rural University of Pernambuco, Brazil Email: {celso, allan}@lrc.ic.unicamp.br, {jgmm, loureiro}@dcc.ufmg.br,
[email protected],
[email protected] and
[email protected] † Dept.
Abstract—Traffic jams frustrate drivers and cost billions per year in time and fuel consumption. In order to avoid such problems, this paper presents an intelligent transportation system that collects real-time traffic information and is able to detect and manage traffic congestion based on this information. Simulation results show that the proposed protocol can reduce the average travel time, CO2 emission and fuel consumption. In particular, the average travel time was reduced in approximately 23%, the average fuel consumption in 9% and average CO2 emission in 10%.
I.
I NTRODUCTION
Traffic congestion is a recurring problem in major urban centers and directly affects the economy because of the increase in fuel consumption. In addition, it affects the productivity of society as congestion increases the time people need to move from one point to another, and it interferes with the environment, because of the increase in the emission of polluting gases. The study carried out in [1] shows that, in 2012, the monetary cost from congestion in the city of S˜ao Paulo, Brazil, was in the order of USS$ 20 billion. Similar numbers can be found for comparable cities around the world. Of this total, 85% is associated with time wasted in traffic, 13% is related to the increase in fuel consumption and 2% can be assigned to an increase in the emission of polluting gases. In an attempt to prevent congestion and improve the efficiency of transportation systems, Intelligent Transportation Systems (ITS) have been proposed. ITSs use advances in technology in the areas of processing, sensing and communication to monitor the traffic conditions in a particular region, manage and decrease congestion, and reduce the number of accidents. In addition, they can be used to provide information and entertainment services to drivers and passengers in order to make their journey a more pleasant process [2], [3]. A Vehicular Network is an important component in an ITS. In this network, vehicles are equipped with processors, sensors and wireless communication interfaces so that they can communicate with one another and with the elements in the network infrastructure (RSU – Road Side Unit), thus creating an ad hoc network while vehicles move through roads [4], [5], [6]. Some studies in the literature propose architectures for the design of an ITS, which range from centralized solutions [7] to distributed solutions [8]. Generally speaking, these solutions
978-1-4673-7194-0/15/$31.00 ©2015 IEEE
663
use both the information provided by vehicles and the characteristics of the infrastructure of the roads to detect and control congestion. Thus, after detecting a congested area or road, a rerouting mechanism is applied to the vehicles to prevent more vehicles from entering the affected area, which results in a more efficient traffic. Thus, this work aims to implement a distributed ITS for detecting and controlling congestions. To this end, RSUs are distributed across the city to ensure total coverage of the region. In addition, each RSU is responsible for managing congestion only in the area covered by its communication radius. Thus, vehicles can interact with several RSUs along their paths to reduce the harmful effects of congestion, such as increased travel time, fuel consumption and CO2 emission. The remainder of this work is structured as follows. Section II provides an overview of the existing approaches to minimize congestion in urban centers. Section III presents our solution for the congestion management. Section IV presents the performance evaluation of our proposed mechanism along with the used methodology and the obtained results. Finally, Section V presents the conclusions and future work. II.
R ELATED W ORK
This work aims to optimize the flow of vehicles in urban centers to minimize vehicle congestion and, consequently, reduce travel time, fuel consumption and CO2 emission. The problem of congestion has been explored in the literature and some solutions have been proposed to optimize the flow of vehicles [7], [8], [9], [10], [11]. Among them, the ones presented in [7] and [8] are the most similar to our proposal, as we discuss below. Pan et al. [7] propose a system for predicting, detecting and controlling congestion to reduce travel time, fuel consumption and CO2 emission. The system uses a centralized approach to gather real-time information about the traffic (position, speed and direction). The collected information is used to detect and predict congestions. After that, the system has a mechanism to control it, in which the re-routing of vehicles approaching congested areas is performed. For the mechanism to control congestion, three re-routing strategies for the vehicles are presented. The first one is known as Dynamic Shortest Path (DSP), which is the default re-routing strategy that associates
20th IEEE Symposium on Computers and Communication (ISCC)
the vehicle to the shortest path, both in distance and time. The second strategy is the Random k Shortest Paths (RkSP), which randomly selects the shortest k paths. Finally, there is the strategy known as Entropy Balanced k Shortest Paths (EBkSP), which uses the same idea of the RkSP strategy. However, a more intelligent mechanism is applied to select the path, as the impact that each one of k paths has on the future density of the road is considered. Doolan and Muntean [8] present a distributed algorithm for re-routing vehicles based on the Vehicle-to-Vehicle communication pattern. The algorithm, also known as EcoTrec, uses features of the highway network and traffic conditions to enhance traffic efficiency and reduce fuel consumption and CO2 emission. In the proposed algorithm, each vehicle periodically disseminates its path and its fuel consumption. This information is used to compute the best paths for each vehicle. If necessary, new paths are suggested for vehicles to decrease fuel consumption and CO2 emission. The calculation of better paths for each vehicle is made in a distributed way, so that each vehicle calculates its best path using Dijkstra’s algorithm. Those solutions present some disadvantages. In [7], the whole re-routing is done centrally, in which a single RSU is responsible for calculating the alternative paths for all vehicles in the network. This typically causes a greater time for the calculation of alternative paths when the road network is large. On the other hand, in [8], the re-routing is done in a distributed way, in which each vehicle calculates its own alternative path. However, as vehicles do not have full knowledge of the network and calculate their alternative paths using only local information (close neighbors), many vehicles may calculate the same alternative path, thus resulting in new congestions. Given this, our solution presents a distributed approach for the re-routing of vehicles using several RSUs. This way, the time for the calculation of alternative paths becomes shorter when compared to [7], as each RSU is responsible for making the re-routing of vehicles located in the area covered by its communication radius. Moreover, the use of a distributed rerouting approach using several RSUs eliminates the problem that exists in [8], since the solution has a broader view of the traffic conditions.
the amount of RSUs used to cover the map. Algorithm 1 presents the process of distributing the RSUs. The input is the total width of the map (Width), the total length of the map (Length) and the range of communication of the RSUs (Radius). The algorithm verifies the amount of RSUs needed to cover the entire map based on the dimensions of the geographical area and on the communication radius of the RSUs (Line 3). After setting the amount of necessary RSUs, their coordinates (x, y) are assigned accordingly to the map (Lines 4–5). Algorithm 1: Distribution of RSUs in the map Input : Width // Total width of the map 1 Length // Total length of the map 2 Radius // Operating range of RSU Output: Coordinates of the points of each RSU 3 4 5 6
// returns the amount of RSUS to cover the entire rsus ← getNumberOfRSUs(Width, Length, Radius); foreach r ∈ rsus do // returns the coordinates (x, y) for each RSU rsus coordinates.add(getCoordinates(r)); end
map
The communication range of the RSUs influences the final cost to implement the system, i.e., the smaller the communication radius, the greater the number of RSUs to ensure the coverage of the entire map. The results presented in Section IV-C show the impact of the communication radius of the RSUs on the proposed system. B. Collection and transmission of data to the RSUs RSUs collect data to acquire information about the events occurring within their communication range and also to understand the behavior of vehicles in their coverage area. Thus, vehicles periodically send information (position, speed, current path, direction and the time they took to move through each road in their path) to the nearest RSU. This information is sent through a single-hop process, considering that both RSUs and vehicles use long-range communications, such as LTE or 3G.
The system is divided into four stages, namely: (i) distribution of RSUs, (ii) collection and transmission of data to the RSUs, (iii) congestion detection, and (iv) congestion control.
Algorithm 2 describes the process in which each vehicle sends its information to the nearest RSU. The algorithm receives as input the set of vehicles in the simulation and the set of coordinates of each RSU, represented in the algorithm as N and RSU sC oordinates, respectively. Thus, each vehicle collects its information (Line 3) and identifies the nearest RSU in relation to its position (Line 4). In addition, each vehicle creates a message containing the collected information and sends it to the nearest RSU (Lines 5–6). After collecting and transmitting data to the RSU, it is necessary its analysis and interpretation. The next section describes the mechanisms to detect and control congestion, which use the data collected by the RSUs.
A. Distribution of the RSUs
C. Detection and control of congestions
The distribution of the RSUs happens homogeneously based on their communication radius and the dimensions of the map. Thus, the higher the communication radius, the smaller
With the information reported by the vehicles, each RSU identifies the characteristics of the region covered by its communication radius. Thus, with these characteristics, each RSU
III.
P ROPOSED S OLUTION
The proposed system comprises of a set of RSUs distributed through the map in order to provide full coverage. Each RSU is responsible for managing vehicles and detecting congestion only within the coverage area of its communication radius. Moreover, the proposed system includes a congestion control mechanism, which periodically performs the re-routing of all vehicles, so they do not go through congested areas.
664
20th IEEE Symposium on Computers and Communication (ISCC)
Algorithm 2: Sending information to the nearest RSU
Algorithm 3: Detection and control of congestions
Input : N // Set of vehicles on the network 1 RSUs Coordinates // Set with the coordinates
Input : N // RSU
of each RSU 2 3 4
1 2
foreach v ∈ N do // returns information about vehicle v status ← v.getInformations(); // returns the nearest RSU to vehicle v nearestRsu ← RSUs Coordinates.getNearstRSU(v.getPosition()); // creates a message with the information of
3
foreach v ∈ N do // sets of edges
7
that make up the route of
route ← v.getRoute(); // returns the current
4
edge of vehicle v to the last edge of the path of vehicle v contained in graph G
vehicle v and sends it to the nearest RSU
6
Graph created by each RSU Number of alternate paths
vehicles v
message ← CreateMessage(status); message.sendTo(nearestRsu);
5
G // K //
Set of vehicles within the reach of each
source ← v.getPosition(); lastEdge ← G.getLastEdge(route); // calculates the k shortest paths
5 6
end
of the source
point to lastEdge for vehicle v
alternativeRoutes ← G.getKShortestPaths(source, lastEdge, K); // selects a path from the set of alternate
7
creates a graph G = (V, A), so that V is the set of crossings within the reach of the RSU (representing the vertices) and A is the set of roads that connect the crossings V (representing the edges). Each road (edge) of graph G has a weight defined by the average speed with which the vehicles travel it and the maximum speed allowed on it. This weight is inversely proportional to the speed at which the vehicles travel the road. Thus, the closer the speed of the vehicles is in relation to the maximum allowable speed, the lower the weight of the road (edge). On the other hand, the lower the speed of the vehicle is in relation to the maximum allowable speed, the higher the weight. With the creation of the graph, the re-routing of the vehicles happens periodically, so that each RSU is responsible for rerouting the set of vehicles in its graph, i.e., the combination of vehicles within its communication radius. The congestion detection occurs in this stage, in which congested roads are identified based on the weights of the roads that make up the graph. The control of the detected congestions also happens in this stage, so that vehicles are re-routed to avoid the congested areas. The re-routing of the vehicles occurs starting at their current point up to the last road of their path contained within the communication radius of the RSU. In our system, the metric used to perform the re-routing was the shortest path. However, a set of k shortest paths is calculated as alternative paths, in which k can be easily adjusted to better adapt itself to the behavior of the network. From these alternative paths, a path is selected through the Boltzmann probability distribution [12]. The purpose of calculating more than one alternative path is to avoid the issue in which the same path is attributed to many vehicles (disadvantage present in [8]). Algorithm 3 describes the process of re-routing the vehicles. The algorithm receives as input a set of vehicles within the communication radius of each RSU (represented in the algorithm as N ), the graph created with the respective features previously described (represented in the algorithm as G) and a variable K that represents the amount of shortest paths that must be calculated. Thus, the RSU calculates the path of each vehicle and its current point (Lines 4–5). Then, the RSU calculates the last edge of the path of the vehicle v contained in its graph (last road of the path of the vehicle within the range of the RSU). After detecting the last road of the RSU,
665
paths, concatenates the remainder of the old path to the new path and assigns the new path to vehicle v
newRoute ← boltzmann(alternativeRoutes, G); if lastEdge! = route.getDestination() then // returns the remainder of the route of the
8 9
vehicle v from the lastEdge
remaningEdges ← route.split(lastEdge); newRoute.add(remaningEdges); v.setRoute(newRoute);
10 11 12
end else
13 14
v.setRoute(newRoute);
15
end
16 17
end
the k shortest paths are calculated starting at the current point of the vehicle v up to the final road of the RSU (Line 7). From the k shortest paths, a new path is selected based on the Boltzmann probability distribution [12] (Line 8). The choice of the path between the k shortest paths is made so that the load of vehicles on the roads is balanced. Each path has a weight calculated as the sum of the weights of all the paths it contains. As the paths with lower weight are the most requested, our system uses the set Rj to decrease the likelihood of them being always chosen, thus balancing the load among the roads. The decision rules to choose the new path are as follows: J = set of vehicles on the map Rj = set of paths of the vehicle j (j ∈ J) rji = path i of vehicle j (j ∈ J) and (rji ∈ Rj ) wji = weight of path rji N (wji ) = normalized value of wji (wji ∈ [0, 1]) defined by Eq. 1:
N (wji ) =
W (rji ) . max{W (rji ) | ∀rji ∈ Rj }
(1)
20th IEEE Symposium on Computers and Communication (ISCC)
The KTj is the Boltzmann constant of vehicle j for temperature T , according to Eq. 2: i (2) e−(N (wj )/T ) . KTj = i∈Rj
The PTj (rji ) is the probability of choosing path i of vehicle j with the parameter of temperature T , according to Eq. 3: PTj (rji ) =
1 KTj
e−(N (wj )/T ) . i
(3)
includes the complete map of the city and the mobility trace of the vehicles, obtained through a real monitoring of the city traffic. The trace defines the times of entry of vehicles on the network during the period between 6:00am and 8:00am. During this period, several points of congestion are formed in the city. The trace has the duration of approximately 6 hours, since the trace ends only when the last entered vehicle completes its route and leaves the simulation. Table I presents some features of the evaluated scenario.
When T → ∞, all candidate paths have the same probability of being chosen, i.e., the process approaches a uniform random distribution. When T → 0, the path with less weight has a high probability of being chosen.
Characteristics Size Amount of roads Peak of vehicles Total vehicles Duration
The E(Rj ) is the path chosen (E(Rj ∈ Rj )), the choice is made according to Eq. 4: E(Rj ) = max{X × PTj (rji )|∀rji ∈ Rj , X ∼ ∪([0, 1])}.
F EATURES OF THE TAPAS C OLOGNE SCENARIO
(4)
After choosing the path, it is verified whether the last road of the path of vehicle v in the RSU is its final destination (Line 9). If it is not the case, the path of the vehicle is divided from the latest road within the communication radius of the RSU (Line 10). The remaining of the path is added to the new path, and, finally, this path is assigned to vehicle v (Lines 11– 12). However, if the last road of the path of vehicle v within the range of the RSU is its final destination, the new calculated path is usually assigned to it (Line 15). IV.
TABLE I.
Values 400 km2 4500 km 136.000 252.000 6 hours
P ERFORMACE E VALUATION
In this section, we present the methodology used to evaluate the proposed solution and the results obtained with it. A. Methodology For the simulations, we used the network simulator OMNeT++ 4.31 , which is an event-based simulator. For the simulation of traffic and mobility of vehicles, we used the simulator SUMO (Simulation of Urban MObility)2 , which is an open source traffic simulator that allows the modeling and manipulation of objects in the road scenario. For the traffic management provided by SUMO. Moreover, we used the Veins 2.13 to model the network in order to increase the accuracy of the results. In the calculation of the CO2 emission and fuel consumption, we used the model EMIT integrated into SUMO. The EMIT is a simple statistical model for instant CO2 emission and fuel consumption based on the acceleration and speed of vehicles, which is derived from the formula HBEFA4 – Handbook Emission Factors for Road Transport. For the simulation and evaluation of our system, we used the TAPAS Cologne scenario [13]. The TAPAS Cologne scenario is a realistic scenario that contains information about vehicle mobility in the city of Cologne, Germany. The scenario 1 http://www.omnetpp.org 2 http://sumo-sim.org 3 http://veins.car2x.org/ 4 http://www.hbefa.net
666
As described earlier, the distribution of the RSUs happens homogeneously based on the communication radius of the RSUs and on the dimensions of the map, so that the higher the communication radius, the smaller the amount of RSUs used for the full coverage of the map. In our evaluation, different configurations of the communication range were tested in the RSUs, and these ranges were 2, 4 and 6 kilometers to assess the impact of the amount of RSUs in the proposed system. When the simulation reaches a steady state, the re-routing of the vehicles starts minimizing the congestions that already exist in the actual trace scenario. Thus, different re-routing frequencies were tested, being them 180, 300 and 600 seconds. For the re-routing of the vehicles, our algorithm has a parameter k for the calculation of the k shortest paths, in which we evaluated the values 3, 5 and 8. Finally, we use the following metrics to validate our system: •
Travel time: the average travel time from the starting point to the destination of all vehicles;
•
Fuel consumption: average fuel consumption of all vehicles to traverse the entire path;
•
CO2 emission: average CO2 emission of all vehicles from the entire path.
The performance of the proposed system was compared to the traditional approach (real trace) and the results are presented in Section IV-B. Moreover, we evaluated the performance of the proposed system with different configurations in terms of communication radius of RSUs, frequency of rerouting and k shortest paths. The results of the evaluation of the proposed system with different configurations are presented in Section IV-C. B. Proposed system versus traditional approach This section presents the results using the proposed system in relation to the traditional approach, in which the vehicles do not interact with the system that provides the re-routing. Figure 1 shows the obtained results. As expected, the proposed system minimizes the effects caused by congestion, that is,
20th IEEE Symposium on Computers and Communication (ISCC)
increased travel time, fuel consumption and CO2 emission. Figure 1(a) shows that the proposed system reduces the travel time by about 23% when compared to the traditional approach. This shows that the mechanism for the proposed re-routing is effective, and it even prevented vehicles remaining stuck at congestions. Figure 1(b) shows that the proposed system reduces fuel consumption by 9% when compared to the traditional approach. This reduction is due to the fact that the proposed system detects and reduces congestion by using alternative paths to avoid congested areas. The proposed mechanism for the control of congestion contributes to a more efficient and continuous traffic, thus, less fuel is consumed. Moreover, it reduces unnecessary braking and acceleration, which contribute to increased fuel consumption. By consuming less fuel, less CO2 is emitted, as we can observe in Figure 1(c). The proposed system introduces a reduction of up to approximately 10% in CO2 emission when compared to the traditional approach. C. Evaluation of the proposed system with different configurations This section presents the evaluation of the proposed system with different configurations. Figure 2 presents the average travel time of the proposed system by varying the communication radius of the RSUs (2, 4 and 6 km), the k shortest paths (3, 5 and 8) and the frequency of re-routing (180, 360 and 600 seconds). Figure 2(a) shows the results of the k shortest paths with a frequency of re-routing fixed in 180 seconds and varying the communication radius of the RSUs in 2, 4 and 6 km. As we can observe, the higher the communication range of the RSUs, the greater the reduction in travel time. This is because RSUs have a greater knowledge of the area of the map and the calculated paths are better in terms of travel time when computed for a smaller area of the map. In addition, we can observe that the amount of k shortest paths has an impact on the results obtained by the proposed system. For the variations of k, we can note that k = 3 has a better performance when compared to other values for k, thus resulting in a decrease of up to 9% in relation to k = 8 and a reduction of up to 4% in relation to k = 5.
– see Figure 2(a)) and varying the range of communication of the RSUs in 2, 4 and 6 km. As we can observe, the frequency of the re-routing positively contributes to the reduction of the average travel time in the proposed system. The results show that the greater the frequency of re-routing, the greater the reduction in the average travel time for the proposed system. Figure 3 presents the fuel consumption of the proposed system by varying the communication radius of the RSUs (2, 4 and 6 km), the k shortest paths (3, 5 and 8) and the frequency of re-routing (180, 360 and 600 seconds). Figure 3(a) shows the results of the k shortest paths (3, 5 and 8) with the frequency of re-routing fixed in 180 seconds and varying the communication radius of the RSUs in 2, 4 and 6 km. As we can observe, the higher the communication range of the RSUs, the greater the reduction in fuel consumption. This occurs because the RSUs have a greater knowledge of the area of the map and the calculated paths are better when computed for a smaller area of the map. Moreover, we can observe that the amount of the k shortest paths has an impact on the results obtained by the proposed system. The results show a difference of up to 7% on the evaluated configurations. Figure 3(b) shows the results with the frequency of rerouting in 180, 360 and 600 seconds, k shortest paths fixed in 3 (better performance presented in the previous comparison – see Figure 3(a)) and varying the range of communication of the RSUs in 2, 4 and 6 km. As we can observe, the range of the re-routing positively contributes to the reduction of fuel consumption in the proposed system. The results show a difference of up to 4% in relation to the intervals of re-routing from 180 to 600 seconds.
(a) K shortest paths Fig. 3.
(a) K shortest paths Fig. 2.
(b) Frequency of re-routing
Fuel consumption.
The results obtained with the proposed system for CO2 emission are presented in Figure 4, varying the communication radius of the RSUs (2, 4 and 6 km), the k shortest paths (3, 5 and 8) and the frequency of re-routing (180, 360 and 600 seconds). As the CO2 emission is related to fuel consumption, the results shown in Figures 4(a) e 4(b) are a consequence of the fuel consumption shown in Figures 3(a) e 3(b).
(b) Frequency of re-routing
Travel time
Figure 2(b) shows the results with the frequency of rerouting in 180, 360 and 600 seconds and k shortest paths fixed in 3 (better performance presented on the previous comparison
667
The results presented in Figures 2, 3 and 4 show that the higher the communication radius of the RSUs, the greater the reduction in travel time, fuel consumption and CO2 emission. Likewise, the results also show that the smaller the amount of the k shortest paths, the greater the reduction of the effects caused by congestion. Observe that greater values of k can
20th IEEE Symposium on Computers and Communication (ISCC)
(a) Travel time Fig. 1.
(b) Fuel consumption
(c) CO2 emission
Results of the simulation
R EFERENCES
(a) K shortest paths Fig. 4.
M. Cintra, “A crise do transito em sao paulo e seus custos,” GVExecutivo, vol. 12, no. 2, pp. 58–61, Jul/Dez 2013.
[2]
G. Karagiannis, O. Altintas, E. Ekici, G. Heijenk, B. Jarupan, K. Lin, and T. Weil, “Vehicular networking: A survey and tutorial on requirements, architectures, challenges, standards and solutions,” Communications Surveys Tutorials, IEEE, vol. 13, no. 4, pp. 584–616, Fourth 2011.
[3]
L. A. Villas, A. Boukerche, G. Maia, R. W. Pazzi, and A. A. Loureiro, “Drive: An efficient and robust data dissemination protocol for highway and urban vehicular ad hoc networks,” Computer Networks, vol. 75, Part A, no. 0, pp. 381 – 394, 2014.
[4]
L. A. Villas, H. S. Ramos, A. Boukerche, D. L. Guidoni, R. B. Araujo, and A. A. Loureiro, “An efficient and robust data dissemination protocol for vehicular ad hoc networks,” in Proceedings of the 9th ACM Symposium on Performance Evaluation of Wireless Ad Hoc, Sensor, and Ubiquitous Networks. New York, NY, USA: ACM, 2012, pp. 39–46.
[5]
L. Villas, A. Boukerche, R. Araujo, A. Loureiro, and J. Ueyama, “Network partition-aware geographical data dissemination,” in Communications (ICC), 2013 IEEE International Conference on, June 2013, pp. 1439–1443.
[6]
G. Maia, L. Villas, A. Boukerche, A. Viana, A. Aquino, and A. Loureiro, “Data dissemination in urban vehicular ad hoc networks with diverse traffic conditions,” in Computers and Communications (ISCC), 2013 IEEE Symposium on, July 2013, pp. 000 459–000 464.
[7]
J. Pan, M. Khan, I. Sandu Popa, K. Zeitouni, and C. Borcea, “Proactive vehicle re-routing strategies for congestion avoidance,” in Distributed Computing in Sensor Systems (DCOSS), 2012 IEEE 8th International Conference on, May 2012, pp. 265–272.
[8]
R. Doolan and G.-M. Muntean, “Vanet-enabled eco-friendly road characteristics-aware routing for vehicular traffic,” in Vehicular Technology Conference (VTC Spring), 2013 IEEE 77th, June 2013, pp. 1–5.
(b) Frequency of re-routing
CO2 emission.
lead to longer (worse) paths, thus directly affecting travel time, fuel consumption and CO2 emission. This can be observed in Figures 2(a), 3(a) and 4(a). Another important factor is the frequency of the re-routing: the longer the interval, the greater the reduction in travel time, fuel consumption and CO2 emission. In that case, vehicles do not need to make unnecessary path changes, thus causing unnecessary braking and acceleration, since a greater frequency of the re-routing contributes to a more efficient and continuous traffic, as we can see in Figures 2(b), 3(b) and 4(b). V.
[1]
C ONCLUSION
In this work, we proposed an intelligent transportation system for detecting and controlling congested roads in urban centers. The proposed system aims to reduce travel time, fuel consumption and CO2 emission. Simulation results show that the proposed system significantly reduces travel time, fuel consumption and CO2 emission. For travel time, a reduction of approximately 23% was presented, fuel consumption showed a 9% decrease and, finally, CO2 emission was reduced by 10%. As future work, we plan to evaluate the proposed system in different scenarios. Moreover, we will investigate a new mechanism to classify the congestion level and compute the k shortest paths considering that classification scheme.
[9]
M. Fahmy and D. Ranasinghe, “Discovering automobile congestion and volume using vanet,” in ITS Telecommunications, 2008. ITST 2008. 8th International Conference on, Oct 2008, pp. 367–372.
[10]
A. M. d. Souza, A. Boukerche, G. Maia, R. I. Meneguette, A. A. Loureiro, and L. A. Villas, “Decreasing greenhouse emissions through an intelligent traffic information system based on inter-vehicle communication,” in Proceedings of the 12th ACM International Symposium on Mobility Management and Wireless Access.
[11]
M. Younes and A. Boukerche, “Efficient traffic congestion detection protocol for next generation vanets,” in Communications (ICC), 2013 IEEE International Conference on, June 2013, pp. 3764–3768.
ACKNOWLEDGMENT
[12]
S. Kirkpatrick, C. D. Gelatt, M. P. Vecchi et al., “Optimization by simmulated annealing,” science, vol. 220, no. 4598, pp. 671–680, 1983.
The authors would like to thank CAPES, CNPq (processes: 480291/2012-8 and 473493/2013-6), FAEPEX, FAPESP and FAPEMIG for the finantial support.
[13]
S. Uppoor and M. Fiore, “Large-scale urban vehicular mobility for networking research,” in Vehicular Networking Conference (VNC), 2011 IEEE. IEEE, 2011, pp. 62–69.
668