Modelling and Analysis of Agent Mobility Patterns by Discrete-Time Markov Chains

This paper proposes an approach for the modeling and analysis of autonomous agent mobility patterns which is based on discrete-time Markov chains graphicallyenhanced to improve modularity. Mobility patterns are associated to autonomous agents when they have to fulfil a distributed task involving migration to remote agent servers where agent computation and/or interaction can likely take place. Two kinds of analysis can be performed on the obtained models: reliability and time analysis. Reliability analysis relies on statistical/mathematical tools and allows for testing the reliability of agent patterns carried out on non-completely reliable networks of agent servers. Time analysis is based on an appositely defined analytical simulation framework and enables the evaluation of time-related performance indexes. Using this approach, several types of patterns were identified, modeled, and analysed: ping-pong, itinerary and free walk. They can be associated to classes of agent tasks in specific application domains.