Message ferrying has been shown to be an effective approach to support routing in sparse ad hoc or sensor networks. Considering a generic network model where each node in the network wishes to send data to some (or possibly all) other nodes with known (and possibly different) rates, we propose three schemes enabling multiple ferries to coordinate in collecting and delivering the data. We analyze the performance of each scheme and establish bounds on the average and worst-case delay. The latter bounds are useful in offering performance guarantees. We establish that under one of our schemes, constant per-node throughput is achievable within constant maximum (worst-case) delay as the network size grows. Using simulation, we compare our proposed schemes with an alternative, the Ferry Relaying algorithm proposed earlier in the literature. The results show that our schemes perform better and provide guidance on which scheme to use given performance preferences and the number of available ferries.

Using mobile elements as mechanical carriers of data has been shown to be an effective way of prolonging sensor network lifetime and of relaying data in partitioned networks. The existing literature has mostly focused on designing delay minimizing routes for the mobile elements by leveraging variants of the Traveling Salesman Problem (TSP). We show that TSP-based routes can in fact result in data delivery delay arbitrarily worse than that of the optimal solution. The main insight is that as the data generation rates of sensors may vary, some sensors need to be visited more frequently than others. To that end, we consider a network with a single sink and develop a Path Splitter algorithm that “splits” a TSP-based route into several loops intersecting at the sink. Numerical results show that our algorithm can improve average delay by more than 40% in some instances while requiring a modest computational effort to modify the TSP-based route.

The authors report on a successful deployment of an inexpensive mobile wireless sensor network in a commercial warehouse served by a fleet of forklifts. The aim is to improve forklift dispatching and reduce the costs associated with the delays of loading/unloading delivery trucks. To that end, an integrated system including both hardware and software is constructed. First, the forklifts are instrumented with sensor nodes that collect an array of information, including the forklifts' physical location, usage time, bumping/collision history, and battery status. The hardware's capability is enhanced with a theoretically sound hypothesis testing technique to capture the rather elusive location information, and the collection of the data is done in an efficient event-driven manner. The information acquired combined with inventory information is fed into a sophisticated actor-critic type stochastic learning method to generate dispatching recommendations. Because noise is inevitable in such wireless sensor networks, the performance of the algorithm is investigated under different noise levels. In combining wireless sensing with state-of-the-art decision theory, this work extends beyond the standard use of wireless sensor networks as monitoring devices.