1 Multiratecast in Wireless Fault Tolerant Sensor and Actuator Networks ⋆ Xuehong Liu 1 , Arnaud Casteigts 1 , Nishith Goel 2 , Amiya Nayak 1 , and Ivan Stojmenovi´ c 1 1 School of Information Technology and Engineering, University of Ottawa, Ottawa, ON, Canada 2 Cistel Technology Inc., Ottawa, ON, Canada Abstract—We study the multicast problem in wireless sensor networks, where the source can send data to a fixed number of destinations (actuators) at a different rate (multiratecast). A typical motivation of such communication scheme is to enable fault tolerant monitoring applications where data is reported to more than one actuators using different rates that decrease with the sensors distance, so that if the closest actuator fails, others can take over from it. We propose two multiratecast routing protocols: Maximum Rate Multicast (MRM) and Optimal Rate Cost Multicast (ORCM), which are the first localized position- based protocols specifically designed for this problem. The first, MRM, selects the next forwarding neighbor(s) in order to favor destinations requiring the highest rates, while the second, ORCM, evaluates several possible choices and select the best according to a cost over progress ratio criterion. The two protocols are compared by simulation, using a new metric that takes the rate into account when computing a multicast cost. Results show that ORCM provides a better routing performance in case of a small number of destinations, while MRM performs better for large numbers of destinations and has a lower computational cost. MRM also behaves better than ORCM when the variance among the rates becomes important. I. I NTRODUCTION A wireless sensor network (WSN) is a network consisting of spatially distributed sensing devices, whose purpose is to mon- itor a given object, surface, or volume in a cooperative fashion, using wireless communication capabilities. Their application domains are various, from military battlefield to civil area with problems such as pollution monitoring, intrusion detection and tracking, traffic control, etc. In a WSN, each sensor device composing the network, called node, can directly communicate with the devices that are within its radio range, or neighbors (assuming these nodes have the same range). Non-neighbor nodes can also communicate indirectly by using the nodes between them as relays. In this case the communication is said multi-hop and it involves the use of a routing protocol to decide what relay nodes should be used. When such communication happens between a source and a single destination, we talk about unicast. If several destinations are considered, then we talk about multicast, and if all nodes are destination, broadcast. Typical use of WSN include sensor nodes monitoring a given area and reporting the sensed data to a sink or actuator that are capable of applying subsequent actions (e.g. alarm triggering). When the monitoring area is large and critical, such as with intrusion detection or object tracking in a battle region, then several actuators might be deployed. In this case, the region can be subdivided into smaller areas, with each actuator in charge of collecting data from overlapping subsets of areas. Typically, the rate could be inversely proportional ⋆ Supported by NSERC Strategic Grant STPGP 336406-07 and NSERC Discovery Grants. to the distance, so that if the closest actuator of an area is damaged, another one can take over without loosing all the historical information of the place. We focus in this paper on the design of rate-based multicast, or multiratecast, routing protocols to support such applica- tions. We first introduce a new metric to calculate the cost of rate-aware multicast paths. Indeed, the usual hop count and retransmission number metrics do not reflect the real efficiency of a path in this context. This idea is illustrated on Figure 1, where two different paths are proposed to serve a given set of destinations (with given rate requirements). Here the ’shorter’ path in terms of both metrics is actually the most expensive if we consider the number of messages to be effectively sent, that is the cumulative rate of retransmissions. 30 10 10 80 30 80 5 5 5 d 1 (80) d 2 (5) d 3 (30) d 4 (10) s n 1 n 2 (a) Path A - 9 retransmissions, 11 hops, total rate cost: 255 30 80 80 10 80 10 5 d 1 (80) d 2 (5) d 3 (30) d 4 (10) s n 1 n 2 (b) Path B - 7 retransmissions, 9 hops, total rate cost: 295 Fig. 1. Impact of the rate on a multicast path cost. Four destinations (d 1 , d 2 , d 3 , and d 4 ) are considered, each one having a different required rate (between parenthesis). Numbers in circles indicate the retransmission rate. The design of WSN multicast protocols is challenging. Sensor devices have small processing power, limited buffer size, radio bandwidth, and especially limited battery capacity. Regarding the energy consumption, the networking activity is far more critical than the computational or sensing activities. Routing protocols thus need to minimize this aspect in priority. If the network is potentially large or subject to reconfigurations (due to node failure or mobility for example), then protocols relying on a global overlay structure maintenance turn out very costly in overhead, and protocols that use only local information such as position-based (or geographic) routing protocols are preferred.