Several protocols have been proposed to implement query and data dissemination architectures in WSNs [8–10]. Most of the aforementioned studies on data dissemination protocols still assume both a single sink and knowledge of the sink's location, but mobile sink environments are used by many current applications such as mobile phones and PDAs, so many researchers have recently studied the problem of supporting sink mobility [11–24].

EEDD [14] has proposed an Energy-Efficient Data-Dissemination protocol to solve both the target and inquirer mobility problem. It has adopted a virtual grid-based two-level design to schedule the active sensor nodes. GGR [15] is a proposed Geographic Grid Routing protocol to provide a multi-path approach at all stages of communication. It focuses on the reliability aspects of the routing protocol.

Perhaps the most similar protocol to SDLS is the Two-Tier Data Dissemination (TTDD) protocol [16], which is source oriented to provide the location information of the source node. TTDD is a well-known grid-based routing protocol to provide query and data dissemination for multiple mobile sinks. Upon detection of an event, the source node proactively creates a virtual grid structure throughout the sensor field with dissemination nodes located at the grid cross points. TTDD operates in a two-tier hierarchical manner to support the mobile sink’s routing. First, the lower tier is within the cell of the sink’s current location. When the sink wants to obtain the sensed data, it floods its query within the lower tier until it reaches the closest dissemination node, called the immediate dissemination node. Second, the query message is then forwarded through the higher tier, which is made up of either the source node or a dissemination node from the sink’s cell to the source’s cell. The dissemination node requests a data download to the source node along the reverse grid path with the data announcement. Finally, once the source node receives the query message, packets are forwarded from the source to the immediate dissemination node along grid horizontal and vertical axes. Requested data will flow down in the reverse direction to the sink. Thereafter, the requested data from the source node is forwarded by trajectory forwarding [16]. TTDD solves the sink mobility problem using a grid structure. However, if the number of source nodes is increased, data dissemination point management can considerably increase the communication and storage overhead of the system, due to the grid structures of data dissemination points from each source node [23]. Therefore, an efficient protocol with multiple source nodes is necessary [12–15,20–24].

Several location-based routing approaches have been considered to address effectively the location tracking problem. However, most of the existing studies on location service protocols were designed for MANETs [17, 18]. Directly employing them in WSNs is ineffective. Location-based routing has been considered to address the issue of providing locations as an effective routing approach in WSNs [19–24]. If the sensor node's location information is found, most applications can use this information to make routing decisions. Greedy Perimeter Stateless Routing (GPSR) [26] is a geographic routing system for multi-hop wireless networks. If a source node can identify a destination node's location, GPSR can forward a packet to the nearest neighbor node in order to reach the destination node.

A locator [20] has been proposed as a data dissemination architecture to support mobile sinks. Once a sensor node wants to report its sensed data to sinks, the source can obtain the current sinks' locations from the locators. These locators are selected for the sensor fields using a deterministic geographic hash function and are replicated uniformly. Railroad [21] has proposed a data dissemination architecture by proactively exploiting a virtual infrastructure called rail. This rail is placed in the middle of the sensor field where all the metadata from the event data are stored. The rail acts as a meeting area for the events and queries. Once a source node detects an event, the data remain stored locally and only relevant metadata are sent to the rail. Then, the sink can retrieve metadata with the queries circulating around the rail. When corresponding metadata is found, the source node of the data transmits the corresponding data to the sink node that has issued the query. Hierarchical Location Service [22] has been proposed as a scalable hierarchical location service in which the location updating structure for the mobile sink is hierarchically built, which can reduce the updating cost. Further, the overall updating cost in WSNs with multiple sinks is greatly reduced by enforcing Voronoi scoping in which each mobile sink only updates its location to the location servers closest to it.

Perhaps the most similar protocol to SDLS is the Anchor Location Service (ALS) protocol [23]. ALS is a grid-based protocol that provides the location information of the sink. ALS constructs a single global grid structure, assigning the sensors nearest the grid points as grid nodes. The sink selects the nearest grid node as its sink agent to distribute its location information. At the beginning of the process, the sink agent sends out four anchor setup messages in four straight orthogonal directions (North, South, East, and West) and all the grid nodes that lie along the routing path are recruited as anchors. Once an event occurs, the source node will register with the nearest grid node, known as the source agent. The source agent will then send four location query packets to find the location of the sink agent. It then receives the location of the sink agent. The source node finally sends the data packets to the sink agent using GPSR protocol. However, it is difficult to identify the location of a moving sink. The following problem occurs when the sink constructs an anchor system. First, the source node cannot identify the location of the sink if the sink does not construct an anchor system. Second, the sink has to wait until the sensor node detects the event. Third, when the sink has high mobility, the sink agent has to relay the data from the source node to the sink. Thus, the distance between the source node and the sink may increase significantly when a detour occurs. ALS also suffers from concentrated energy consumption because it always uses border nodes when it provides the sink location. Therefore, it is necessary to propose an energy efficient protocol to balance sensor node energy consumption.

The Sink Location Service (SLS) [24] informs the source node about the sink location to discover sink locations with low overhead. Each senor node gets four anchor locations during network initialization phase. A source node and a sink send sink location announcement and query messages along two paths respectively by geographic routing. The sensor node located on the crossing point of the sink location announcement path and the sink location query path informs the source node of the sink location, then the source sends the data packet to the sink by geographic routing. No flooding or global grid structure is needed for sink location announcement or query. Although SLS provides sink locations to minimize the overhead incurred for the provision of source nodes with the location of sinks, it does not solve forwarding data from a source node to a mobile sink. This process decreases the resilience of the protocol. The main reason for this is if one query message fails, there is no other query message to retrieve sink location information. In addition, no sensor node might be located on the crossing point of these two paths. In this case, the source node cannot find the location information of the sink.

When a sink sets on static location, geographic routing can be used successfully, i.e. sending the packets to the closest node to the destination. As for the static sink, it only needs to broadcast once to the whole network at the initial stage of the network operation, which would not result in significant energy dissipation. If a sink moves quickly to be out of last hop node’s transmission range, geographic routing can’t be executed correctly without a mobile sink’s location updates because the sink can’t receive last hop forwarding without any additional reactions. For solving this problem, the mobile sink can periodically flood its location to all sensors but it is not efficient and scalable for large sensor networks and each time the sink changes its location in the network. Due to high packet overhead of flooding, when these nodes use up their energy, no more data can be transmitted back to the sink. Thus, sinks should update limited number of sensors own current location instead of flooding to all sensors. Therefore, the mobile sink only needs to broadcast to a subset of nodes in the network each time when it arrives at a new site, thus saving much energy.

In a general way, a source node that has detected an event of interest tries to send its data to the sink. Therefore, the mobile sink broadcasts its location information over the subset of nodes in the network [see Figure 2(a)]. However, when the sink moves out of a destination node, it has to rebroadcast its location information over the network to a new destination node. We argue that this repeated network-wide broadcasting brings extra high energy loss which could impair the lifetime gains from sink mobility, especially in large-scale WSNs.

To address this problem, we provide source node’s location in the WSNs [see Figure 2(b)]. A significant difference between our proposed scheme and the traditional mobile sink supporting scheme is that the source node provides its location service server. Therefore, it is necessary to propose an energy efficient location service protocol to balance sensor node energy consumption.

In this paper, we propose a data dissemination model called EDA (Eight-Direction Anchor) for geographic. Sinks should be able to acquire current source node’s location with bounded time and overhead. EDAs are location server sensors that track source node’s position. Figure 2 illustrates our motivation for proposing the location service servers. As shown in Figure 2(a), selection of location service servers will be established using six lines of grid nodes, three lines of grid nodes traversing the network vertically and three lines of grid nodes traversing the network horizontally. But from Figure 2(a), EDA can reduce the traversing path from 6 to (2+2 2) compared with ALS. EDA prevents intensive energy consumption at the border sensor nodes; consequently, the energy consumed is well balanced for all the sensor nodes.

After the source node gets the sink’s location, the source node should be able to transfer the sensed data to shortest path. Figure 3 illustrates our motivation for proposing the shortest path relay.

When the sink has mobility, the sink agent has to relay the data from the source node to the sink. Then, this occasions a detour case (see Figure 3(a)). Therefore, it is necessary to propose an energy efficient shortest path relay. In our proposed algorithm, once the sink is aware of the source node’s location, the sink can directly request the data forwarding from the source node [see Figure 3(b)]. It can efficiently forward data from a source node to a sink with minimal delay path. Therefore the sink can quickly get source node’s data.

SDLS constructs a shared global grid structure once all the sensor nodes are deployed. Global grid construction can be enabled when all sensor nodes obtain their locations using a local positioning system and maintain their one-hop neighbors’ geographic locations in their tables. Specifically, the grid is set up in parallel with the X-axis and Y-axis of the positioning system. The positive directions of the X- and Y-axes of the predefined positioning space are pointing to the east and north, respectively. The baseline coordinate (Xbase, Ybase) of a pre-defined positioning system is set in the mission messages or hard coded in each sensor node. The all sensor field is divided into small virtual grids. The basic idea of the grid-based approach is to establish a grid using the intersection points called grid points. The distance to the grid point is the cell size and is determined by the user such that grid points are not within transmission range. The sensor field is divided into square-shaped grids of user-defined size α. Each cell is an α × α square. Based on the global grid, the virtual coordinate (Xp, Yp) is determined to be grid point, the nearest node to the grid point is called the “grid node”, and the grid nodes that take the role of the location server for a specific source node are named “anchors”. The grid points are determined using the baseline coordinate (Xbase, Ybase) as follows: (1) { Xp = X base + i * α , Yp = Y base + j * α } ; { i , j = ± 0 , ± 1 , … }

The first grid point sends a grid node announcement message to each of the four adjacent grid points, using simple geographic forwarding through intermediate nodes. As a result, the node that is geographically closest to each of the grid points becomes the grid node. Finally, several nodes are selected as grid nodes to cover all grid points across the original network. Every grid node will be in charge of a location server and will maintain the location information of the source node in its cache. After selection, the grid nodes will announce their role to the adjacent grid nodes. Then, neighboring grid nodes send notification messages to each other to establish a neighbor grid node table. The main objective of global grid is to deliver location information of the source nodes to the multiple mobile sinks. In the case of multiple mobile sinks, it is difficult to keep tracking the location of mobile sinks, since links between the source node and the sink may be easily broken. In our protocol, once the sink obtains the location of the source node using the global grid structure, the sink can directly send a query to the source node. This will be explained in Section 3.5.

Although some location services have been proposed to address location tracking, they suffer from inefficient energy consumption or their specifications are not detailed in. Especially, ALS [23] suffers from concentrated energy consumption, as it always uses border nodes, when it provides the sink location. Though SLS [24] generates lower overhead than existing protocols such as flooding, TTDD, and ALS, they do not consider both location response time and resilience in case one query message fails. Therefore, we propose the Eight-Direction Anchor (EDA) system to solve these problems. EDA is built using the following procedures.

Once an event occurs, a source node selects the nearest grid node as its source agent (see Figure 4, source node and selected source agent) and distributes the location information of the source node.

These anchor setup messages are generated by a geographical routing protocol. That is, the source agent distributes the location information of the source node to its eight-direction neighboring grid nodes until the message reaches the network border. EDA can prevent intensive energy consumption at the border sensor nodes and considers location response time and success rate to obtain location information. Algorithm 1 explains how to make an Eight-Direction Anchor system at source nodes.

A sink queries the network to collect the location of source nodes. Query and data dissemination are achieved using the following procedures. After creating the anchor through the anchor nodes selection stage, the sink can identify the location of the source node by sending a query when the sink wants to receive events. To do this;

The sink will register with the nearest grid node, known as the sink agent (see Figure 4, sink and selected sink agent).

Furthermore, once the sink agent obtains the source node’s location information, it does not obtain it again. If one query message fails, the other two queries will serve as backups to retrieve the source node’s location information. This also shortens the query response time, because the sink will utilize the first response.

Once the sink agent receives the location information of the source node, the sink agent forwards it to the sink. The sink then finally can communicate with the source node using the GPSR protocol.

Since our protocol provides the location of the source nodes, instead of the location of the mobile sinks, the anchor system, which is the location server, can maintain the location information of the source nodes, although the sinks move fast and continuously. Algorithm 2 explains how to find location of source nodes at sinks.

In this section, we propose a Location-based Shortest Relay (LSR) protocol to provide shortest path from a sink and a source node in mobile sink environments. In SDLS, once the sink obtains the location of the source node from the sink agent, the sink can directly send a query to the source node. Our protocol does not need to use a global grid structure to support a mobile sink. The sink selects the nearest sensor node as its Primary Agent (PA) to do so.

After the source node receives the query from the sink, it sends data to the PA. Thereafter, the source node can continuously send the data to the sink, once the data stream starts flowing. We now introduce the following efficient rules to support sink’s mobility. The algorithm 3 explains how to maintain sink mobility and data forwarding.

First, we compare the average energy consumption for location service. The location service is defined as receiving the location of all source nodes available in the network. Especially in SLS, if there is no node located on the crossing point of the two paths or one query message of the source node fails, the location service will continue until receiving all the location information of sinks. As the overhead to construct the grid (SDLS, ALS) or anchor nodes selection (SLS) is one time only, we do not include the results. Figure 6(a) shows a graph of each protocol’s average energy consumption versus the number of nodes making the anchor system from 1 to 10, incrementing by 1. While TTDD consumes more energy than the other protocols, SLS performs better than ALS and SDLS, since TTDD constructs grid structures (anchor system) for each source node and both ALS and SDLS are quite complex compared to SLS. ALS always uses the border nodes as its anchor system, so energy consumption is concentrated around border nodes. Conversely, SDLS balances across nodes in the entire grid. Therefore, it prevents high energy consumption of specific nodes. When we consider the energy consumption for making an anchor system, SLS provides the smallest energy consumption. However, this method will be disadvantaged in another evaluation.

Figure 6(b) shows the average energy consumption when nodes send a location query to find nodes to make an anchor system. Here, we consider that there are four nodes making the anchor system. TTDD requires more energy than the other protocols to make an anchor system; we can see that TTDD consumes more energy than the other protocols. TTDD also has to find the location of source nodes per each source node. However, both SDLS and ALS find node locations using global grid nodes with low energy consumption. The anchor system of ALS concentrates location information on border nodes. Thus, it frequently reaches the border nodes to obtain location information. Therefore, SDLS can obtain location information faster than ALS. As shown in the figure, we can see that ALS’s energy consumption increases faster than for TTDD. In TTDD, a sink finds four fixed source nodes. If the sink must find more source nodes, the protocol will consume more energy. As in forming the anchor system, SLS has the least energy consumption. However, the energy consumption of SLS increases gradually with the number of nodes querying location if no the sensor node is located on the crossing point or one query message fails.

In order to evaluate the average location response time of multiple sinks, we fixed the α = 200 m and varied the number of sinks from 1 to 10 in steps of 1. In ALS and SLS, from Figure 7(a), it can be seen that the location response time slightly increases with the number of sinks. This is because when multiple sinks are deployed in the network, location query messages always have to travel until they find all sink’s location information. ALS shortens the location response time than SLS, because ALS utilizes the first response among the other responses. However, because SDLS constructs anchor system by source node, the location response time remains fairly constant regardless of the number of sinks in the network. Although TTDD gets also fixed source nodes, it takes much time because TTDD constructs grid structures for each source node.

In order to evaluate the average location response time of multiple sources, we fixed the α = 200 m and varied the number of sources from 1 to 10 in steps of 1. In SDLS, from Figure 7(b), it can be seen that the location time slightly increases with the number of source. This is because when multiple sources are deployed in the network, location query messages always have to travel until they find all source’s location information. However, until source nodes are 10, SDLS is less time than ALS to get location information. TTDD is also same that the location time increases with the number of source nodes.

To compare the impact of moving sinks on the location, we report the data communication overhead of SDLS, TTDD, and ALS except SLS, because SLS does not mention supporting mobile sinks after the location service. Figure 8 shows the remaining energy power of the sensor nodes for data communication. For the experiment, we assume the following parameter settings: number of sinks and source nodes = 30, number of sensor nodes = 400, size of the sensor field = 600 × 600 m², transmission radius = 50 m, initial energy = 0.1 J, sink speed = 10 m/s, and the source node generates data packets 30 times. The data communication overhead is the highest with ALS. TTDD consumes less overhead than ALS, because TTDD can find a new path by switching to a new grid node. However, this is not the shortest path. Therefore, SDLS provides the best performance using the LSR protocol.

Figure 9 depicts the average data success ratio, as the number of sinks and source nodes increases from 0 to 30 in increments of 5. From Figure 9(a), we assume that the sink is static and the source generates data packets only once with an initial energy of 0.1J. As shown in the figure, SLS has the lowest data success ratio. Here we can see that SLS has difficulty obtaining location information regardless of mobile sinks. Except for SLS, as shown in the figure, the average data success ratio decreases when the number of sinks and source nodes is over 15. However, SDLS still has a high success ratio, even when the number of sinks and source nodes is 30. From Figure 9(b), we assume that the sink speed is 10 m/s and the source node is generated 30 times with an initial energy of 1J. ALS has the worst data success ratio, as ALS has to continuously relay data, while SDLS provides the shortest path through the LSR for data transmission.

In Figure 10 (a), we simulated the network lifetime of each scheme according to the number of sinks and sources with an initial energy of 0.1 J. As shown in this figure, as the number of sinks and sources increases, network lifetime of all schemes decreases. Even though SLS has the highest network lifetime of all analyzed schemes, SLS has the lowest data success ratio as shown before. TTDD has the worst network lifetime because TTDD reselects a disseminate node for each source node whenever an event is detected. SDLS still has a high network lifetime even when the number of sinks and source nodes is 50.

Figure 10 (b) depicts the network lifetime according to the simulation time from 0 to 200 seconds. Sink speed is 10 m/s and the number of sinks and source nodes is 15, with an initial energy of 1J. Since SDLS distributes data over the entire network by considering residual energy, the network lifetime could be significantly prolonged compared to the other protocols. The network lifetime for both SDLS (LSR) and SDLS (LSR + EDA) is longer than other protocols.

In order to evaluate the average data communication overhead, we vary the velocity of the sink from 5 m/s to 30 m/s in increments of 5 m/s. Each source node generates a data packet 10 times. In other words, Figure 11 shows that after finishing location service, data packets are sent by multi-hop transmission. We can see that average data communication overhead is the highest with ALS. TTDD consumes less overhead than ALS. This is because TTDD can find a new path by updating to a new grid node. However, this is not the shortest path. Therefore SDLS performance is the best.