Agreement strategies for multi-robot systems

Stratégies de commande collaborative pour des systèmes multi-robots




  Carlos Canudas de Wit

  Alexandre Seuret




Sandra Hirche    TUM, Munich, Germany
Christopher Edwards    University of Leicester, Leicester, United Kingdom
Antonio M. Santos Pascoal      IST, Lisboa, Portugal
Dimos V. Dimarogonas    KTH, Stockholm, Sweden
Mazen Alamir    GIPSA-Lab/CNRS, Grenoble, France
Alexandre Seuret  

   GIPSA-Lab/CNRS, Grenoble, France



Keywords: Agreement strategies, Cooperative control, Distributed control, Consensus algortihms, Formation control.  


Abstract: The idea of deploying formations of relatively unsophisticated autonomous robots to accomplish complicated tasks has roots in the early works studying the flocking and foraging behaviors among birds. The main question was how one can mimic different behaviors witnessed in populations of birds, animals, insects, etc. among a population of artificial agents. The emerging use of large-scale multi-agent and multi-vehicle systems in various modern applications has recently raised the need for the design of control laws to perform challenging spatially-distributed tasks such as search and recovery operations, exploration, surveillance, environmental monitoring or pollution detection and estimation, among many others.

This dissertation focuses on distributed control strategies for a set of mobile robots, with a particular attention to agreement protocols. A significant part of the manuscript deals with consensus algorithms of arbitrary linear heterogeneous agents, representing, for example, different models or generations of robots. Motivated by the fact that only a few works consider heterogeneous cases of the synchronization problem, a control strategy is proposed based on a consensus algorithm which is decoupled from the original system. The new algorithm offers the major advantage to separate the stability analysis of each agent and the convergence analysis of the distributed consensus algorithm.
On a second set of works, a special attention is paid to consensus algorithm's convergence rate. Focusing in memory based approaches, the stabilizing delay principle is used. More precisely, a correctly weighted  state sampled component is added to the control law allowing us to artificially manipulate the graph's algebraic connectivity.
Finally,  algorithms for the compact deployment of agents are designed and analyzed.  This manuscript proposes a completely distributed algorithm allowing swarm self-organization while improving the network's connectivity properties. For these protocols, the desired formation is entirely specified by the angles formed by agents within the formation.


   Download manuscript