Modeling mobility through dynamic topologies

Abstract

The representation of spatially moving entities with models exhibiting a static topology is, in many cases, very difficult to accomplish due to the computational complexity of the communication cost: O(n2), in the general case. Actually, in systems involving sensors, like radars, many messages will also be discarded when entities become out of communication range, burden the simulation with unwanted messages that need to be discarded. To improve model efficiency and expressiveness we develop a solution based on dynamic topologies and peer-to-peer communication (P2PC). In this approach, the topology undergoes a runtime adaptation in order to reflect the current interaction pattern among entities, being peer-to-peer communication links created and destroyed according to the characteristics of emitters and sensors, that include, for example, a minimum threshold power on signal detection. We show that time-varying topologies overcome some of the limitations inherent to the widely used publish/subscribe communication (PSC), employed for example, in the High Level Architecture infrastructure (HLA). We show that the combination of P2PC (enabled by dynamic topologies), and regions of interest (ROIs) (based on the HLA/PSC), defines a flexible simulation architecture for modeling spatially moving entities that improves over solutions based exclusively on PSC/ROIs. Our approach is described with examples of two systems that benefit from topology adaptation: a defense system and a mobile phone network.