Location-based publish/subscribe - LPS for short - is an important building block forcontext-aware applications in mobile ad hoc networks (MANETs). In LPS, published messages are routed based on their content as well as on the location of publishers and subscribers.
Existing LPS algorithms can be coarsely classified as follows: (1) message-centric approaches consist in broadcasting published messages, (2) query-centric approaches broadcast subscriber queries for subsequently routing messages, and (3) hybrid approaches broadcast queries and messages each within restricted scopes. Each approach is clearly superior to others for particular communication patterns, e.g., for certain ratios between the number of queries and the number of messages in the network. This paper presents an adaptive location-based publish/subscribe (ALPS) algorithm for settings with multiple, unknown, or varying communication patterns. ALPS can be viewed as a parameterized hybrid LPS algorithm that can seamlessly move between message- and query-centricity based on estimations of the current communication pattern.
We evaluate ALPS on two benchmark applications namely in the context of mobile social networking and robot swarms. Our results indicate that ALPS reduces the message complexity by up to a factor 3x compared to the best respective alternative, while performing comparably to the respective optimal solutions with static communication patterns, making ALPS appealing as a one-size-fits-all solution.