Routage g éographique dans les r éseaux de capteurs et actionneurs. (Geographic routing in sensor and actuator networks)

This thesis is about wireless multi-hop networks such as wireless sensors networks or hybrid sensor/actuator networks and actuator networks. Those kinds of networks are composed of independent entities (nodes, i.e. the robots) which have very limited computing and memory capabilities. Moreover, they are battery powered and have to work in an efficient way. They communicate through the radio medium and do not require any static infrastructure. In order to relay messages between actuators up to the base station, we use what is called "routing protocols". In order to be efficient, those protocols have to find a routing path, over multiple robots, in a local and distributed manner, without any global knowledge about the network. My works rely on CoMNet, the first geographic routing protocol which relies on the controlled mobility of the robots while guaranteeing network connectivity despite their movements. CoMNet relocates next hop node on each routing step, according to a predefined relocation pattern. Its aims to adapt the network topology to the routed traffic in order to save energy. Nevertheless, CoMNet does not consider the consequences of those relocations more than in a one-hop way. By making node N relocating node N+1 on the routing path, CoMNet adapts the topology locally. But CoMNet ignores the fact that making this N + 1 node move changes its routing possibilities too. We proposed MobileR (Mobile Recursivity), a new routing protocol which takes this phenomena into account. MobileR, at each routing step, anticipates the routing in a multi-hop manner through computations over its one-hop neighbors. It makes the current forwarding node compute all the possible routing paths toward the destination using relocation patterns recursively. The cost-over-progress of each path is computed and the next hop selected is the first hop on the best computed path. Still, the protocol is completely localized and no information, except the forwarding packet, is shared between nodes. On the one hand, the relocating nodes principle is not without consequences. It brings new issues to solve. For instance, in wireless sensor networks, events are likely to be detected by multiple sensors. Consequently, on events occurrence, multiple sensors transmit message toward the destination. But those source nodes are geographically close, as they report the same event. Hence the routing paths for the data they transmit are very close and even merge close to the destination. This phenomena has to be considered. In current routing protocol for actuators, those common routing path parts provoke useless oscillation and premature node death as relocating scheme compete for the same nodes. As a response to this phenomena, I propose the PAMAL (PAth Merging ALgorithm) routing algorithm. PAMAL detects those routing path crossing and handles them in a purely localized way. It makes nodes oscillation stop and provokes a path merging upstream and uses a packet aggregation downstream. Thanks to this behavior, PAMAL makes the network lifetime increase up to 37% with the simplest possible aggregation. But on the other hand, controlled mobility also makes possible new answers to old routing issues in wireless sensors networks. The Greedy Routing Recovery (GRR) routing protocol takes controlled mobility into account in order to increase delivery rate on topology with holes or obstacles. Indeed, None of the few existing routing protocols for actuator networks proposes any mechanism to bypass obstacles or topology holes where greedy routing is impossible. They all rely on a greedy forwarding routing strategy toward destination node. Hence they all fail when the current holding node has no neighbor closer to the destination than itself. GRR includes a dedicated relocation pattern which will make it bypass routing holes and create a routing path on which greedy forwarding will be possible. The obstacle, or hole, is circumvented by relocating nodes all around. Thanks to this light recovery, next routing are going to be in a fully greedy forwarding way.Simulation results show that the light recovery mechanism we use in GRR has a hit ratio of 72% over network topologies where traditional CoMNet fails.