Multi-agent flocking with random communication radius
2012 American Control Conference (ACC)
In this paper, we consider a multi-agent system consisting of mobile agents with second-order dynamics. The communication network is determined by a metric rule based on a random interaction range. The goal of this paper is to determine a bound on the probability that the agents asymptotically agree on a common velocity (i.e. a flocking behavior is achieved). This bound should depend on practical conditions (on the initial positions and velocities of agents) only. For this purpose, we exhibit
... i.i.d. process bounding the original system's dynamics. We build upon previous work on multi-agent systems with switching communication networks. Though conservative, our approach provide conditions that can be verified a priori. Our result is illustrated through simulations. I. INTRODUCTION Cooperative behaviors generating complex phenomena are observed in nature ,  . Multi-agent systems also find applications in technical areas such as mobile sensor networks , cooperative robotics  or distributed implementation of algorithms  . A central question arising in the study of multi-agent systems is whether the group will be able to reach a consensus. Intuitively, agents are said to reach a consensus when all individuals agree on a common value (e.g. the heading direction of a flock of birds, the candidate to elect for voters). To carry out formal studies on consensus problems, one usually assumes that the multi-agent system follows some abstract communication protocol and then investigates conditions under which a consensus will be reached. Existing frameworks include discrete and continuous-time systems involving or neglecting time-delays in the communication process. The communication network between agents is usually modeled by a graph. Its topology is either assumed to be fixed or can switch over time. The switching topology of the interactions is sometimes assumed to depend on the state of the agents (e.g. the strength of the communication can be a function of the distance between agents). Consensus can be modeled in a deterministic fashion, however, in many applications it seems that the topology of the network is quite random. Recently there has been a growing interest in studying consensus algorithms in a probabilistic setting , , , , , , , where network changes can be independent, identically distributed (i.i.d. ) over time , ergodic-stationary , or Markovian .