Paso 2: Asignación del presupuesto: Lo que continua luego de reconocida la cartera de proyectos de inversión prevalecidos asumiendo en cuenta los costos totales y el
2.2.2.5. Consideraciones generales
In energy aware routing, paths from sources to sinks are formed such that the nodes with the most remaining energy are used where possible. The selection of modules used in energy aware routing protocols is shown in Figure 4.5.
Link/Node Costs Energy Path Costs Max element Costing
Singh’s Routing Heuristics
Singh [105] presents several energy aware routing heuristics. These approaches are:
• minimise energy expenditure, which will be discussed in Section 4.6,
• minimise max node cost, which is also discussed as part of minimum energy routing in Section 4.6,
• maximise time until network partition, which aims to maximise the time until there is no path from a source to a sink; Singh indicates this problem is NP-complete and therefore impractical,
• minimise variance in node power levels, which Singh also indicates is NP-complete and therefore impractical,
• minimise cost/packet, where link/node cost is the inverse of a node’s energy reserves and the path cost is shortest path.
This final remaining approach (minimise cost/packet) is used as the basis for most energy aware load balancing routing protocols. Chang and Tassiulas [20][21] have experimented with routing strategies to extend a WSNs lifetime, considered to be the time until the first node expires, and have studied this approach in more detail. In their experiments, the cost of a link was based on two factors: the remaining energy at the transmitting node and the energy required to transmit. By trying different weightings of these factors, Chang and Tassiulus determined that the dominant factor in extending lifetime is the remaining energy of a node. Consequently, paths should be selected to avoid the use of nodes with little remaining energy. One problem with the experiments of Chang and Tassiulas is their assumption that nodes only expend energy in transmitting. As discussed in Section 2.5.1, a node may expend more energy in receiving than transmitting and so the experiments of Chang and Tassiulas may not be entirely accurate.
There are also problems with the minimise cost/packet heuristic. Firstly, if cost is based on a node’s energy reserves then a node’s cost may be constantly changing. Determining the remaining energy at a node requires energy expenditure. It therefore becomes difficult or expensive to accurately assess which route is cheapest in the network. In order to provide an accurate approximation to the amount of energy stored in a node, Lin [76] proposes
67
representing the remaining energy of nodes using discreteenergy levels. Using this tech- nique, a four-level logarithmic scale is proposed with level 0 corresponding to full energy, level 1 between half and full energy, level 2 between quarter and half energy, and level 3 between one eighth and zero energy. Whilst this reduces the overhead associated with load balancing, routes are required to be recalculated when the energy level of any node changes. The second problem is that minimising cost/packet only involves the distribution of routing workload and does not consistently use paths of lower cost. Thus, the capacity of the network may be lessened by the use of sub-optimal paths even if more time elapses before the first node expires. The remainder of routing protocols in this section aim to use the minimise cost/packet heuristic while reducing the cost of each routed message.
max-min zPmin
In max-min zPmin [72], the transmission power of nodes is reduced such that the network
remains connected. By reducing transmission power, the energy consumption of the net- work is reduced. In order to carry out load balancing, the routing protocol then selects the (source, sink) path whose node of least energy is the highest, i.e. the maximum minimum remaining energy. Paths containing nodes of little energy are therefore avoided. However, in reducing the transmission power of nodes, the number of connections between nodes will drop. As the number of connections drops, the amount of load balancing that can take place will also drop. For example, if reducing the transmission power causes more bottle- neck nodes to form then it may not be possible to carry out load balancing, since all data must travel through the bottlenecks.
Chang’s Routing Heuristic
Chang [22] has proposed a new link/node cost measurement in which the cost of transmis- sion becomes more important as the node’s remaining energy drops. Therefore, at first, it is acceptable to use suboptimal paths to achieve load balancing. However, as the energy of the network declines, it becomes more important to route such that energy is preserved. There are two difficulties with Chang’s approach. Firstly specific implementation details are absent, making it difficult to comment on the approach. Secondly, the solution involves the use of sub-optimal paths in order to carry out load balancing. The use of these paths
may reduce the capacity of the network. A better solution would be to carry out load balancing on the optimal paths, thus not wasting energy and carrying out load balancing simultaneously.
Capacity Maximisation Routing Protocol (CMAX)
CMAX [58] operates by firstly disregarding all links where the cost of transmission is higher than some fraction of the remaining energy of a node. Thus, links will not be used if they result in a node’s expiration. Link costs are based on both the remaining energy of a transmitting node and the cost of transmitting from that node to the receiver such that a link’s cost is lower if the node has more energy or if it the transmission cost is low. Path cost is calculated based on shortest path routing with the additional requirement that the path cost must be below some constant. The author of CMAX notes that some of the routing protocol’s behaviour is unacceptable in a WSN. For example, if the use of a particular path is deemed too high, or if it involves the use of nodes that are close to expiring, the routing protocol will refuse to route the message. This behaviour will lead to a network where all nodes have a small amount of remaining energy and cannot engage in routing, which is of little practical use.
HEAP
HEAP [127] is a modification of AODV in which each message contains a transmission power p and remaining energyr. The protocol operates similarly to AODV except that on receiving a message, a node may only react if it has more than r remaining energy. Furthermore, any message transmitted by the node must be sent using a transmission power of p. The routing protocol carries out route requests using increasing values of p and decreasing values ofruntil a path is found. A disadvantage with this routing protocol is the extra energy expenditure that occurs in carrying out multiple route requests. Furthermore, more route requests must be carried out if there are only poor paths available, i.e. those in whichris at a minimum andpis maximum.
The majority of these routing protocols are problematic in that they involve additional energy expenditure in order to search for a low cost path that may be load balanced. There is no guarantee that any path will be found, thus causing additional energy expenditure for
69
no benefit. Another limitation of energy aware routing is that it extends the time until any node expires. However, the loss of some nodes may have no impact on the connectivity of the network. On the contrary, it may even be beneficial to overuse certain nodes to prevent certain critical nodes from expiring and partitioning the network.