The conventional routing protocols applied to the wireless ad-hoc network always assume that nodes are mostly connected with each other. But in some special environments, ex. Terrestrial mobile networks, as Figure 1, buses can be used to store and forward messages, but the mobility of buses makes the network become intermittent connected; Exotic media networks, as in Figure 2, contain satellite networks, long-distance wireless network, or underwater networks, hence we can predict this will result in intermittent connection, high delay, and broken cause of the environment; Military Ad-Hoc Networks, routing in this kind of network may be in hostile environments where mobility, environmental factors, or intentional jamming may cause for disconnection; Sensor network, these networks are frequently characterized by extremely limited end-node power, memory, and CPU capability. In addition, they are envisioned to exist at tremendous scale, with possibly thousands or millions of nodes per network. Communication within these networks is often scheduled to conserve power, and sets of nodes are frequently named only in aggregate. The challenges of routing in the network are high delivery delay, intermittent connection, large queuing delay, and buffer limitations at intermediate nodes. This kind of network is the so-called delay tolerance network (DTN) [1][2].
Messages need to be stored and forwarded at another time.
Figure 1: Terrestrial mobile networks
Figure 2: Exotic media network
Although messages can be sent to the destination by storage and forwarding of relayed nodes, we cannot promise that messages will have the minimal delivery delay.
For example, Figure 3, messages from node A to node B are routed through node A at time=0, node C at time=0, node C at time=1, node C at time=2, node C at time=3, finally node B at time=3. But the second path through node A to node B at time=2 has smaller delivery delay.
Figure 3: Can not promise that messages will have the minimal delivery delay
Hence, we use space and time information to generate graph with links in space and time domain. And the links in the time domain are connections between the same nodes in different time intervals. For example, we transfer Figure 3 to Figure 4. In Figure 4, the lines between nodes mean that the nodes are in the same transmission range in the same time interval. The dotted lines connect the same nodes in different time intervals. The shortest path from node A to node B is A0 -> A1 -> A2 -> B2, so we will adopt the second path in Figure 3 to send the message from node A to node B.
Routing in space and time can be applied to GPS (global positioning system); the information of dynamic position, time, speed, and schedule of the path to destination can be known by GPS. The navigation center displays the trajectory of the vehicles’
mobility, and vehicles can also get the information needed from the navigation center, for example, who data needed from other vehicles has to be relayed from whom. [3]
proposes to find the path with minimal delivery delay with space and time
information. [4] also uses a knowledge oracle to obtain the connection between nodes.
Figure 4: Finding the path with minimal delivery delay in space and time
The complexity of applying the Dijkstra algorithm to find the routing path using space and time information is where N is the number of nodes, T is total time, and is the graph size. Hence, the complexity is proportional with the number of nodes and the total time considered of finding the routing path. In DTN, the high delay between nodes will lead to larger total time intervals needed by finding the routing path, so the complexity becomes large. For example,
)2
(N×T )
(N×T
Figure 5 shows that if the number of nodes is 50 and the total time intervals are 20, the complexity will be . Our purpose is to decrease the complexity of routing with space and time information in DTN, we focus on reducing the topology by using cluster topology to get the routing path. We propose the cluster-based routing with space and time
)2
20 50 ( ×
information to group the nodes into clusters, thus the topology is reduced.
Figure 5: High delay lead to higher complexity
The topology of finding the routing path in cluster-based routing is cluster topology. Every node in cluster topology is a cluster, and the weight of edges is the time cost by sending messages between two clusters, and the links between clusters are formed by deploying the members of each cluster as the gateway node, like Figure 6, the links between clusters are A0 -> B0 -> B1 -> C1, and A0 -> B0 ->BB1 -> B2 -> C2. B0B , B1 and B2 are gateway nodes. If we ignore the delay in the same time interval, we can obtain that A0C1 =1, and A0C2 =2 in the cluster topology. Compared to the original topology, which uses all the nodes to find the routing path, cluster topology is less complex. However, using cluster-based routing to decrease the complexity will lead to two problems. The first one is the additional space delay because messages will be sent to cluster head first, but the space delay is much smaller compared to the total delivery delay in DTN, so it can be ignored. The second one is the heavier load
of cluster head, and we use load balance clustering by choosing nodes with lower loads to be cluster heads and gateway nodes to avoid congestion.
Figure 6: Change the node topology to cluster topology