R Outing and S Cheduling for R Eal

Wireless sensor networks have been attracting significant research and commercial interest recently. In many sensor network applications, such as real-time surveillance and monitoring, the sensed data need to be delivered with some real-time constraints, such as an end-to-end deadlines. Since the data dissemination model is different from the tranditional wireless ad hoc networks or any centralized systems, new challenges arise. Different solutions have been proposed, such as real-time sensitive routing protocols, data packet prioritization, and real-time scheduling. This thesis demonstrates that real-time scheduling in sensor networks is influenced by several factors: (1) the routing protocol can influence the paths taken in the network and the interaction between contending flows; and (2) the packet scheduling algorithm can arbitrate the use of the network in a deadline sensitive manner to ensure that most packets make their deadline. Most existing work in real time scheduling has focused on the second factor, with little work addressing the first factor and the interaction of the two. Further, to prioritize and schedule the data traffic so as to meet the end-to-end deadline, the end-to-end delay should be tracked carefully. As observed, the end-to-end delay of a data packet travelling from the source to destination is influenced by several factors: (1) the physical and MAC layer transmission delay; (2) the processing delay at each hop; and (3) the queueing delay at each hop. While most existing work attempts to minimize the end-to-end delay, it often ignores the queueing delay as a major component of the overall delay. In this thesis, we first study the effect of routing protocols on real-time sensor network iii applications. We compare stateful and stateless routing protocols, showing the advantage and disadvantage of both. Two routing protocols that are representative of these classes, shortest path routing and greedy forwarding routing, are studied real-time constraints. We then study existing scheduling algorithms for sensor networks and analyzes the components of the end-to-end delay of a data packet from the source to destination. We conclude that existing scheduling algorithms do not consider all those possible contributing parts and therefore do not lead to effective solutions under moderate to high contention. Finally, the thesis proposes a new scheduling algorithm called Just-in-Time Scheduling (JiTS), which deals with all the contributing components of the end-to-end travelling delay of data packets in sensor network. It compares JiTS with another prioritizing and scheduling algorithm for real-time data dissemination in sensor networks, velocity monotonic scheduling. It simulates both scheduling algorithms based on two typical routing protocols, shortest path routing and greedy forwarding with NS2.