5 Dimensionamiento del tablero
5.3 Acciones sobre el tablero según IAP-11
5.3.1 Acciones permanentes de valor constante (G)
Depending on the application requirements and the WSN deployment area char- acteristics (area, terrain, size, etc), a mobile sink can follow different types of mo- bility patterns, such as unpredicted mobility [82–85], predictable mobility [86–89] and controllable mobility [90–93]. In the class with unpredictable sink mobility, the sink follows a random path in the sensor field and routing protocol can only rely on the current state of topology. There is no guarantee if the sink reaches all the sensors or how much time it takes to do so. Hence, it may result in incomplete data collection. In the class with predicted mobility, by exploiting the predictable nature of sink’s movement, an appropriate strategy is designed to determine the routing paths for data packets, to guarantee the coverage of the sensor field and to optimize data delivery performance. The controlled sink mobility denotes a property of routing protocol rather than the property of the sink’s motion. Based on a parameter of interest, such as residual energy of the nodes, lifetime of packets queued in nodes, or a predefined objective function, or a predefined observable events (alleviate congestions), the sink is guided towards a specific route by applied routing schemes. The aim is to address a particular problem according to the network’s needs. Not only the type of sink’s mobility but also the sink’s speed affects the operation and the network performance. Based on the network structure the protocol applies, the routing protocols can be classified into hierarchical and non-hierarchical approaches. Hierarchical ap- proaches establish a virtual hierarchy of sensor nodes, where the sensor nodes
are composed of different dynamic roles and virtually form a hierarchical struc- ture. Thus, the advertisement overhead of sink positions can be limited to a certain number of nodes and decreased. The constructed hierarchy is normally composed of two or more tiers. The nodes in the overlay structure (high-tier) are responsible to communicate with sink directly obtaining sink’s position and uploading data. The nodes remained (low-tier) acquire sink’s information from the overlay nodes or direct their data packets to the overlay nodes. Accord- ing to the specific virtual structure, the hierarchical approaches can be further classified into: cluster, tree, grid, backbone or area-based. On the other hand, all the sensor nodes have the same role in non-hierarchical approaches. Such a structure alleviates hotspots problems caused by the concentration of data traffic towards the sinks, and eliminates the constructing overhead for a virtual struc- ture. However, this approach requires all the nodes acquiring sink’s information individually which increases overhead dramatically especially in large-scale net- works. The non-hierarchical approaches can further be classified: flooding-based (e.g. Two-Tier Data Dissemination (TTDD) [94]), overhearing based (e.g. Hy- brid MANET Routing Protocol (HMRP) [95]) and geographic information based (e.g. Multicast-Query-based Data Dissemination (MQDD) [96]) approaches. With respect to the QoS constraints specified during the network functional- ity, the approaches can be classified: energy-efficiency based, delay based and throughput based approaches. Because of the low-cost sensor devices, energy- efficiency is the basic requirement for network operation. Moreover, for some real-time applications such as image and video sensor, QoS metrics such as delay and reliability should be guaranteed during the network operations. However, satisfying these metrics, especially in mobile sink scenarios, may cause conflicts with energy-efficiency.
According to the number of MS applied in network, the algorithms can be catego- rized into single MS based, multiple MSs based and dual-agent (combined mobile
2.3. Routing protocols with mobile sinks for WSN data collection 29
Unpredictable
Mobile sink(s) based rou1ng solu1ons
Sink mobility pa3ern
Rou6ng approach
structure constraints QoS
Controllable
Predictable efficiency Energy delay Non-‐
hierarchical Hierarchical
Flooding Overhearing Geographic informa1on
throughput
Cluster Tree Grid Backbone Area-‐based
Number of MSs Dual-‐ agents Mul6ple MSs Single MS
Figure 2.3: Category illustration of mobile sink based multi-hop routing solutions.
and static sink) based [97] classes. In the single MS scenario, depending on the objective and algorithm design, the MS can visit all sensor nodes or only a subset of them. Fig. 2.3 shows the taxonomy of routing protocols.
The use of MSs for aggregation requires the definition of sensors visiting order, i.e., an itinerary has to be scheduled jointly for all the sinks, and this is itinerary planing problem. The chosen itinerary largely affects the network performance, such as energy consumption, network lifetime, data aggregation latency and so on. Some multiple MSs based algorithms [77, 98, 99] are proposed. The managed network is partitioned into several logical/physical domains and each domain is assigned to a separate MS. The key point in multiple MSs scenario is that an efficient method is needed to propose optimal clustering of sensors as well as optimal itinerary design of individual MSs.
Konstantopoulos et al. [77] propose a heuristic algorithm - Tree Based Itinerary Design (TBID) that employs multiple MSs. TBID determines the proper num- ber of MSs for minimizing the total aggregation cost and constructs low-cost itineraries for each of them. Aiming at addressing the multi-agent itinerary plan- ing problem, Chen et al. [13] propose a Minimum Spanning Tree (MST) based
algorithm. The solution is divided into two parts: source node grouping and source node visiting sequence of each MS. The estimated hop count is the main network cost parameter in the algorithm with the precondition that geographical information of all source nodes are stored in the sink. A balance factor which is applied in the network weight calculation function shows a flexible trade-off control between energy cost and task duration. Wu et al. [97] take advantage of dual-agent (combined mobile and static agent). Instead of spreading the global information through repeated broadcasting across the network, the mobile sink only needs to broadcast its location to a subset of nodes in the network each time when it stops. For those nodes that do not know where the mobile sink is, they send their data to the static sink.
The main objective in these multiple MSs based algorithms is to minimize the volume of network traffic exchanged between distributed systems while main- taining relatively low task execution time, especially for time-critical tasks [77]. However, both the cost and efficiency should be evaluated, since the application of multiple MSs can be of high cost. Using a single MS may actually lead to response time, network overhead, and energy consumption better than that the multiple MSs algorithms in some data aggregation applications. Besides, some single MS algorithms can be potentially extended to multiple MSs [100][101]. In the following, we focus on routing protocols that are based on a single MS.