An Introduction to a GIS-Based In-Road Information Network

An Introduction to a GIS-Based In-Road Information Network


Karim Mohammadi
Iran University of Science and Technology
[email protected]

S. Shahab Sahhafi
Iran University of Science and Technology
[email protected]
This paper proposes a new in-road information network to significantly improve Location Service and Routing Protocol based on GIS (Geographic Information System) road map, speed and location of vehicles. Recent surveys have shown that current in-road information networks are inefficient to find a destination node in long-distant routes (Location Service), and unable to transfer information between source and destination easily for high-speed vehicles (Routing Protocol). In order to overcome these difficulties the new in-road information network suggests using communication map layer and considering each vehicle speed and location, to divide in-road network into two hierarchical levels and makes it similar to a chain of cells. As a matter of fact, these are the cells which virtually communicate with each other, and make Location Service and Routing Protocol become easy and possible. The paper just observes the network layer, and physical layers and interfaces are ignored.

This paper presents a new method based on Geographical Information system (GIS) maps in making an efficient location service and routing protocols for in-road information networks. Creating an in-road information network as an infrastructure for the inter-vehicle communication has been considered recently. The main goal of this network is to transfer traffic data and improve the safety of the road transportation. However, there are some problems which may reduce the reliability and affect the performance of an in-road information network: it is difficult to find a specific vehicle (node) in this large network. On the other hand, it is difficult to track a destination vehicle in the network and forward a data packet to it, even when the current position of the destination node is known.

Mobile ad hoc networks which are made by distributed mechanisms and cooperation among all the nodes are bases of in-road information networks [1] [2].

Problems with physical layer, in which high speed vehicles decrease the capability of current spread spectrum standards, and network layer, in which location service and routing protocol are not reliable, affect the in-road performance of mobile ad hoc networks.

Roads characteristics with fast and unpredictable movements of vehicles drastically increase rate of the change in an in-road information network topology, so location services and routing algorithms in current ad hoc networks have many limitations in road communication. This is because relative positions of vehicles are frequently changed and make creating, control and maintenance of routes very difficult. Also, it makes an increase in the amounts of control messages and beacons which are using network band width improperly.

Overcoming this problem, vehicle position data is needed. So that, each vehicle should be equipped with a GPS receiver to get its own instant geographical coordination. On the other hand, in a position-based network a source needs to know a destination position before sending data, so it must flood a request message in all directions of the network. Although this mechanism can determine the destination position, it is a consuming time and high-cost procedure. However, by using maps, storing and processing spatial and non-spatial data in Geographical Information System (GIS), time and cost will be saved.

The objective of this paper is to develop a new method to design network hierarchy structure based on GIS information and roads maps.

It is hypothesized that all of network nodes (vehicles) are equipped with GPS receivers, maps, GIS and required communication equipment inside.

Position is an important data in networks using position-based routings. Large networks use hybrid routing protocols []. So, the whole network should be divided into many zones with unique IDs (Zone ID). In fact, Zone IDs are very important because they are a part of the address field of each packet. Furthermore, each node has a unique ID, too. Making a packet address, a node has to integrate its ID and its relative zone ID.

In many mobile ad-hoc networks, zones depend on node existence. It means that, some nodes together make a zone, so if these nodes separate from each other, that zone may be divided into smaller zones. Also, it is possible a zone to be eliminated or merged with other ones.

According to importance of packet address field, it must be guaranteed that there aren’t any similar IDs for different zones. This makes sure that the sending data packet can be received properly in the destination. To achieve this aim, making vehicle-independent zones for in-road information network is proposed. Nowadays, some cars have GIS maps in their navigation systems. It is expected that in future, most of the cars will have this system with them, so they can identify in which road they are moving along.

Using maps and spatial and non-spatial data, makes zone identification so easy and fast. In this way, Geographical Information System (GIS) as a strong tool to collect, store, process and display the result of spatial information helps to identify zones and their IDs. Also, it helps the vehicles to know their relative position against other vehicles. Moreover, if a communication layer map is added to GIS road maps it can show, in each part of the road, how much wireless radio range is. This information is used to make unique zones that are the same for all vehicles.

Figure 1 shows a hypothetical map including zone configuration in a road.

Fig. 1. Road topography effect on the length of zones.

As radio signals can not be broadcasted in large distances in mountain regions, zones are smaller than smooth areas.

Having road communication GIS map, locations of zones and their geographical regions are determined. By loading them in all vehicles navigation system, in advance, all of the cars use this information (zones geographical attribute and their unique IDs) and identify the road in which they are moving.

It is possible one thinks that if there isn’t such a map in a particular road, what will happen for the network. The answer is, the network is just used for roads with specific characteristics such as high density of vehicles (crowded roads), and so it is not necessary to have this kind of map for all roads. Since there aren’t enough cars in some roads, making such a network is not always possible, because cars can not often connect together by using short range wireless communication. Also, cars can get road division information in any ways (e.g. cooperation among cars); the important fact is geographical based division.

After all, as in position-based addressing methods, each node is equipped with a GPS receiver; each node will know its instant geographical position. Therefore, if a node wants to obtain position of another node, it must search whole network i.e. it floods request messages. Since there aren’t any predetermined information stations in the network, the only node which can reply properly is destination node. However, it is a time consuming and high cost procedure. To overcome these limitations, a virtual access point (VAP) in each zone will be introduced in the next part.

After zones determination, a leader must be identified for each zone, similar to an access point in hierarchical ad-hoc networks [1]. In fact the leader (access point) is one of the cars existing in the zone. Since this access point is varied by time, it is called a Virtual Access Point (VAP). VAP controls the zone during a specific time and when it leaves the zone, gives its duty to another car.

Suppose all vehicles are aware about zones in a specific road. When a vehicle enters a new zone or starts its activity, it sends a beacon and waits to receive a response from the environment. In case, there is no response, it starts sending a new beacon once again. This process may be repeated so many times (depending on the network design).

After all, when the vehicle receives no response at all, the car will conclude that the only car existing in the zone is it. So this new car will be the zone VAP. On the other hand, if another car exists in the zone when a new car sends beacons, the zone VAP will reply it and after a little time the VAP will take charge of the new car.

Fig. 2. A typical road zone scheme.

Figure 2 shows a typical road zone. Here VAPs take charge of their zones and vehicles. Since the network have two levels, like other hierarchical ad-hoc networks, if a typical node (vehicle) wants to transfer a data packet to another node outside of the zone, it must be sent through VAPs. It is clear that to have a reliable and unconnected link, the length (L) of a zone must be up to half of the maximum radio range (R) in that part of the road.

When a new vehicle took charge of a zone, it starts to send beacons and control messages. According to these messages, all nodes configure themselves and consequently the network. Messages contents and their sending rate affect the performance of the network significantly. High sending rate occupy bandwidth and reduce efficiency. In GIS-based network, since zones are independent and not time varying, control messages and beacons will be reduced. For example, each car needs to send its information twice during its presence in a zone; when it enters and exits, but VAP have to send its messages periodically because it must announce its presence to other nodes all the time. For example, if VAP is damaged, other nodes can understand it, because they won’t receive any messages (beacons) from their VAP.

Each VAP has a table including some information such as unique ID, speed, entrance and estimated exit time of vehicles which are inside the zone. By looking up this table, a VAP can identify which nodes are inside its zone, quickly. Also VAP uses this information to determine its substitute VAP (SVAP). It is important to keep this selection up to date periodically. If for a reason, VAP is damaged, after a little time the SVAP will take charge of the zone so other nodes will send their information to it. This work is done to save time and prevent any disorders in network. In case there is no SVAP while VAP is broken, nodes have to cooperate and determine a new VAP. This is a time consuming process.

Neighbor VAPs exchange their network information and cooperate with each other. When a node leaves a zone, its VAP announces neighbor VAP and gives it the control and information of the leaving node. This procedure is a little complicated for a VAP when it wants to leave its zone. As it is mentioned, a VAP knows its neighbor zones, so when it is going to leave the old zone, it knows whether there is a VAP in neighbor or not. If there is no VAP in the next zone, it will be the next zone new VAP, but if a VAP exists; it will change to an ordinary node and accept the next zone control. On the other hand when a VAP leaves its zone, it gives its table information to SVAP.

The rate with which network topology changes, is an important factor in network operation. The network with lower rate of topology changing has less control messages than a one with high rate. VAPs affect this factor greatly because they control other nodes and periodically broadcast beacons, so in VAP determination process, it is important to select a one which affect the topology as less as possible. As it is said, VAP has a table with some information. VAP knows how much it takes a vehicle to leave the zone, so it knows which car stays in the zone more than others. This may be a good choice for substitute, but if we want to implement our network by current wireless equipment, it makes an important limitation.

Current spread spectrum communication technologies are sensitive to speeds of nodes. It means for nodes with high speed, the efficiency of communication is reduced [3]. This factor must be considered in substitute determination, so it had better choose a SVAP among nodes with speeds near to average zone or lawful speed. This is not a defect for GIS-based network i.e. designers are free to consider any optional factors.

In the GIS based in-road information network, data communication has two classes: (1) data communication within a zone and (2) inter-zone data communication. Nodes (vehicles) in the same zone can communicate together directly. Before they start data transmission, notify the VAP and if there aren’t any problems, the VAP will give them permission. This is done because if VAP receives a message request from another zone for one of these connected nodes, it can interrupt them and get the message to the requested node.

Now suppose a typical node wants to send data packets to another one. At first it sends a request to send message (RTS) to its VAP and notifies it. The VAP looks up the information table. If destination is inside the zone, it will forward RTS to the destination node. Receiving RTS, the destination node sends a reply back to the source. Now the source and the destination can start data transmission.

If a source and its requested destination are not in the same zone, VAP must find the requested node in the network, so it sends a RTS to its direct neighbor VAPs. Since all of the nodes information is saved in VAPs, the source VAP must just search among VAPs. Here, there is a “some for all” location service, because information of all nodes is saved in some nodes. It is important to notice, positions of these stations (information data bases) are fixed and known i.e. each zone has its VAP as a base station, and thus all of nodes can be connected to them without any problems.

Fig. 3. Hierarchical concept in GIS-based in- road network.

There is an interesting fact concern to in-road information networks that makes the design and implementation easy. As a matter of fact, since vehicles have to drive just in roads not outside them, locations of nodes of network are nearly predetermined. If a specific car goes out of a road, it will be eliminated from the network and it’s not important, because according to our aim, we want to make a network just in roads.

According to this fact, it is necessary for a VAP to send RTS messages toward two directions: up-road and down-road. Then unlike other location services and routing protocols in mobile ad-hoc networks, finding an unknown destination is limited to two directions [2]. Then there is a kind of directional flooding mechanism in routing phase in road networks.

By using VAPs, the amounts of RTS and control messages are reduced again. In the routing phase, the only responsible node in each zone is VAP. This means that the maximum number of hops in a two-level GIS based network is (3+ Nz), where Nz is the number of zones between a source and destination. If both of a Source and destination are VAPs then the number of hops will be (Nz+2). As we see, the maximum number of hops in a specific road is predetermined and known, so the amount of delay originated from the number of hops can not be increased from a threshold. An important source for delay is networking traffic that makes data packets wait in queues in VAPs. This is caused from bandwidth limitation and discussed in physical layer so here is not considered.

Fig. 4 Route maintenance of In-road network.

After finding the destination and making a route between the source and destination, the maintenance phase is started. In this phase it is tried to maintain the route during data transmission. Like base stations in a common cellular communication system, VAPs have to control and survey movements of nodes and forward data packets to destinations in new zones. This is depicted in figure 3. As it observed, the source and destination zones are changed during transfer of data. When a connected node (source or destination) wants to leave its zone, the VAP must pass it to the next VAP and then inform the source VAP. Consequently this can be done for the source node. Because in an in-road network VAPs locate along road such as chains, informing source node may be not necessary. That is, when a message arrives to the pervious destination VAP, this VAP correct the address and forward it toward the destination where is in new zone now. It had better be done when a source and destination move in opposite directions.

As a summery we can say our network is a hierarchical (two level) network in which zones is divided based on geographical map that has a communication layer.

Since movements of vehicles in roads are random and unpredictable, statistical software ARENA has been used to simulate and model the network. As mentioned, the rate that the network topology is changed with is an important factor that predicts the amounts of control messages and beacons. In high rate, it is expected that there are plenty of control messages in the network and vice versa.

To analyze this factor in our in-road network, first we must consider a probability function that presents the rate of vehicle entrance in a special zone. An exponential function can be a good choice. In this function an average of time when a car enters a zone (μ = λ ) and the variance ( α) are given. It has the form like the below function:

In simulation the value of has been selected as 8, 30, 70, 110 and 160 second.
Determination of probability distribution function is another issue. The maximum speed which cars must not pass it, is given in each road, so we can estimate the average speed of cars in a special zone as a value near to maximum allowed speed. We chose a triangle probability distribution function for speeds of vehicles. Since the numbers of cars, which have higher speeds than allowed velocity, are less than vehicles, which have speeds less than allowed one, the ramp in higher speed is sharper then the ramp in lower speed.

Fig. 5. The vehicle velocity probability distribution function

The average speed of vehicles in this simulation has been chosen equal to 115 km/h. In addition, it is considered that vehicles have fixed velocity along the zone just for simplicity. Another factor in our simulation was the length of a zone. It supposed equal to 3 km.

Simulation shows an interesting result. For example for λ as figure 6 indicates, the average rate of topology change is equal to average cars speed divided by the length of the zone.

This is because this rate is proportional to VAP change rate. As mentioned before, VAP broadcast control messages more than other nodes, so its movements have the most effect on topology.

Fig.6. Simulation result for .

After all, in our in-road information network the rate of topology change is a predetermined value and can differ by cars average speed and the length of zone (if physical layer has no limitation), so maximum delay and queuing problems can be estimated well. As another note, there is an interesting point relative to topology change rate: zones length and car velocity in many roads are inverse, in other words in mounting regions cars derive slower than flat roads and on the other hand in flat places the length of zones are shorter than one in even regions (because of the range of radio communication). Therefore we expected that topology change rate does not differ greatly from flat to even roads.

Comparison between our method with other identical researches and methods such as GLS, SOTIS [4], DSR and ZPR indicates that all methods suggest using position data in the address field of data packets [1] [2]. However, previous works have not used road characteristics and especially their maps [4] [5]. Generally GIS can be used in in-road networks, simplify models and reduce the amounts of computations. In addition, it guarantee unique zone IDs for network. This is very important and efficiently affects reliability of data delivery.

Table1. Comparison between mobile ad hoc networks in road applications.

Table 1 compares our GIS-based in-road information network with other mobile ad hoc networks qualitatively. The main result of this comparison is VAP attribute in network performance. By using this content in addition to road map it is expected that the complexity of in-road network be reduced.


  1. George Aggelou, “Mobile Ad Hoc Network “, Mc Graw Hill, 2005.
  2. Martin Mauve and Jorg Widmer, University of Mannheim, Hannes Hartenstein, NEC Europe, Heidelberg, “A Survey on Position-Based Routing in Mobile Ad Hoc Network”, pp. 30-39, IEEE Network, November/December 2001.
  3. Keith Biesecker, “Broadband Wireless, Integrated Servieces and Their Application to Intelligent Transportation Systems”, Center for Telecommunication and Advanced Technology, McLean, Virginia, June 2000.
  4. Lars Wischhof, André Ebner, Hermann Rohling, ” Self-Organizing Traffic Information System based on Car-to-Car Communication: Prototype Implementation”, WIT 2004- 1st International Workshop on Intelligent Transportation, Hamburg.
  5. Walter Franz, Christian Wagner, Christian Maihofer, Hannes Hartenstein, DaimlerChrysler AG, Research & Technology 3, 89081 ULM, Germany, “FleetNet: Platform for Inter-Vehicle Communications”, 1st International Workshop on Intelligent Transportation 2004.