Workshop for the action on "Decentralised control of multi-agent systems under constraints" on 7th December 2023 at CNAM, Paris. Registration is free but compulsory. The action also asks colleagues to identify research work on this theme.
For the colleagues that cannot join us in Paris, we will retransmit the presentations via Zoom. The Zoom link can be obtained by writing an email to Michael Defoort (email@example.com)
Programme préliminaire du Workshop du Jeudi 7 Décembre, 10h00 - 16h00
Lieu : CNAM, 292 Rue Saint Martin, 75013 Paris, Salle 11A1-39 (Accès 11, Premier étage)
09:50-10:00: Presentation of the MACs action (decentralized control of multi-agent systems under constraints)
10:00-10:30: Distributed differentiation based on homogeneous high-order sliding modes: Application on dynamics average consensus
Authors: Rodrigo Aldana Lopez (University of Zaragoza), Rosario Aragüés, Carlos Sagüés
Abstract: The homogeneity approach to High-Order Sliding Mode (HOSM) design has proven to be highly valuable in several applications, including the study of Robust Exact Differentiators (RED) and controllers for chains of integrators. Here, we point out a novel application of HOSM in the context of Dynamic Average Consensus (DAC) algorithms.
The DAC problem involves computing the time-varying average of a set of input signals distributed across a network in a distributed fashion. DAC algorithms serve as building blocks for more complex distributed methods across many disciplines, such as sensor networks and robotics. Traditional solutions rely on linear algorithms, which can only achieve bounded error under persistently varying inputs. However, recent attempts were made to use first-order sliding modes to achieve exact convergence to the true average signal in finite time. Unfortunately, these algorithms suffer from chattering, which is particularly important in multi-agent systems where real implementations involve discrete-time implementation, time-varying delays, and a lack of synchronous clocks between agents. This motivates using HOSM and homogeneity to mitigate the effects of chattering.
In this talk, we draw parallels between current linear DAC techniques and High Gain Observers (HGO), motivating the existence of an equivalent of the standard RED in the context of DAC. This insight lays the foundation for developing a novel distributed differentiator derived using the homogeneous HOSM design philosophy. The proposal generalizes the standard RED, effectively differentiating the average of signals distributed across the network.
We will cover the main challenges arising from the proposed algorithm, specially when providing formal convergence guarantees, addressing gain design, and ensuring robustness against network connectivity issues.
10:30-11h15: Generalized Homogenization of Linear Control Protocols for Multi-Agent Systems: Theory and Experiment
Authors: M. Li (INRIA Lille), A. Polyakov, G. Zheng
Abstract: Homogeneity is a dilation symmetry of a dynamical system, which can be utilised for acceleration of converge and improvement of robustness. This talk is devoted to homogeneity-based design of control protocols for multi agent systems. A technique for design of generalized homogeneous control protocols is discussed. Consensus and formation control protocols for MIMO agents are considered. Theoretical developments are supported with numerical and practical experiments on mobile robots.
11:15-12:15: Multi-robot coordinated navigation using guiding vector fields
Author: Ming Cao (University of Groningen)
Abstract: Navigation and path-following are fundamental functions for mobile robots carrying out environmental monitoring and sampling tasks. While guiding vector fields have proven to be an effective tool in such applications, new challenges arise when control algorithms have to be designed for a team of robots with limited communication capacity, on sophisticated paths and in occluded environments. In this talk, I show how to design guiding vector fields to enable motion coordination among robots that follow a given path that may self-intersect; I also show how to construct composite guiding vector fields to avoid colliding with obstacles of arbitrary shapes. Both theoretical guarantees and experimental validations are discussed for practical scenarios.
Biograph: Ming Cao has since 2016 been a professor of networks and robotics with the Engineering and Technology Institute (ENTEG) at the University of Groningen, the Netherlands, where he started as an assistant professor in 2008. Since 2022 he is the director of the Jantina Tammes School of Digital Society, Technology and AI at the same university. He received the Bachelor degree in 1999 and the Master degree in 2002 from Tsinghua University, China, and the Ph.D. degree in 2007 from Yale University, USA. From 2007 to 2008, he was a Research Associate at Princeton University, USA. He worked as a research intern in 2006 at the IBM T. J. Watson Research Center, USA. He is the 2017 and inaugural recipient of the Manfred Thoma medal from the International Federation of Automatic Control (IFAC) and the 2016 recipient of the European Control Award sponsored by the European Control Association (EUCA). He is an IEEE fellow. He is a Senior Editor for Systems and Control Letters, an Associate Editor for IEEE Transactions on Automatic Control, IEEE Transaction of Control of Network Systems and IEEE Robotics & Automation Magazine, and was an associate editor for IEEE Transactions on Circuits and Systems and IEEE Circuits and Systems Magazine. He is a member of the IFAC Council. His research interests include autonomous robots and multi-agent systems, complex networks and decision-making processes.
13:30-14:15: Spectra of graphon operators and dynamical properties of systems on large graphs
Authors: Federica Garin Paolo Frasca, Renato Vizuete (GIPSA-Lab)
Abstract: Graphons are a notion of continuous limit of sequences of large graphs, as well as a model for a class of random graphs. We consider network systems (network-SIS epidemic model and linear consensus algorithm, i.e., Laplacian diffusion), where the network graph is very large. The graph is not perfectly known, but randomly sampled from a known graphon. We show how spectral properties of the graphon can be used to closely approximate some properties of the large-but-finite dynamical system, such as the stability condition and an index of sensitivity to noise for SIS epidemics, and the speed of convergence and the Kirchoff index for the Laplacian diffusion.
14:15-15:00: Consensus dynamics over a graph with static and bounded-confidence type interactions
Authors: Vineeth Satheeskumar Varma (CRAN)
Abstract: In this talk, we study consensus dynamics for multi-agent systems that interact over a static graph as well as a dynamic graph where links are formed only when the agent states are sufficiently close. When applied to social networks, our intention is to account for hard and fast ties present due to physical or social proximity as well as interactions over online social networks that only occur between agents that have similar opinions. When applied to physical systems (like drones), the static graph could represent communication over the internet and the dynamic graph represents communication at short range using WiFi. Albeit simple, the model remains difficult to analyse due to its separation of time scales, discontinuous dynamics of the graph, and the slow variables that evolve in time. We give some theoretical properties of the system, and numerical simulations highlighting the emergent behavior.
15:00-15:45: Bipartite Formation over undirected coopetition networks with inter-agent collision avoidance and connectivity maintenance
Authors: Pelin Sekercioglu (ONERA, L2S), Ioannis Sarras, Antonio Loria, Elena Panteley and Julien Marzat
Abstract: In the study of coordination of multiagent networks, a large number of consensus problems have been studied, with agents reaching global agreement through collaboration. However, there are scenarios in which the system contains agents that compete with each other and their relationship is represented by a competitive interaction. Networks containing both cooperative and competitive interactions are called coopetition networks. For these networks, the general achievable goal is multi-partite consensus, in which there is more than one equilibrium point. More precisely, for structurally balanced networks, the attainable goal is bipartite consensus, in which all agents converge to the same state in modulus, but with opposite sign. In addition to the convergence towards a desired destination, the secondary objective of autonomous vehicles is to guarantee collision avoidance between agents while maintaining the information exchange between cooperative agents. Thus, we address the bipartite formation control problem for networks having both cooperative and competitive interactions. We present a barrier-Lyapunov-function-based control law that guarantees compliance with these inter-agent constraints. We establish asymptotic stability of leaderless and leader-follower bipartite formation-consensus manifolds.
15:45-16:00: Conclusion and perspectives