Convergence Rate for Discrete-Time Multiagent Systems With Time-Varying Delays and General Coupling Coefficients

Consensus of Stochastic Dynamical Multiagent Systems in Directed Networks via PI Protocols

With the rapid development of swarm intelligence, the consensus of multiagent systems (MASs) has attracted substantial attention due to its broad range of applications in the practical world. Inspired by the considerable gap between control theory and engineering practices, this article is aimed at addressing the mean square consensus problems for stochastic dynamical nonlinear MASs in directed networks by designing proportional-integral (PI) protocols. In light of the general algebraic connectivity, consensus underlying PI protocols for a directed strongly connected network is investigated, and due to the M -matrix approaches, consensus with PI protocols for a directed network containing a spanning tree is studied. By constructing appropriate Lyapunov functions, combining with the stochastic analysis technique and LaSalle's invariant principles, some sufficient conditions are derived under which the stochastic dynamical MASs realize consensus in mean square. Numerical simulations are finally presented to illustrate the validity of the main results.

Learning From Architectural Redundancy: Enhanced Deep Supervision in Deep Multipath Encoder-Decoder Networks

Deep encoder-decoders are the model of choice for pixel-level estimation due to their redundant deep architectures. Yet they still suffer from the vanishing supervision information issue that affects convergence because of their overly deep architectures. In this work, we propose and theoretically derive an enhanced deep supervision (EDS) method which improves on conventional deep supervision (DS) by incorporating variance minimization into the optimization. A new structure variance loss is introduced to build a bridge between deep encoder-decoders and variance minimization, and provides a new way to minimize the variance by forcing different intermediate decoding outputs (paths) to reach an agreement. We also design a focal weighting strategy to effectively combine multiple losses in a scale-balanced way, so that the supervision information is sufficiently enforced throughout the encoder-decoders. To evaluate the proposed method on the pixel-level estimation task, a novel multipath residual encoder is proposed and extensive experiments are conducted on four challenging density estimation and crowd counting benchmarks. The experimental results demonstrate the superiority of our EDS over other paradigms, and improved estimation performance is reported using our deeply supervised encoder-decoder.

Products of Generalized Stochastic Matrices With Applications to Consensus Analysis in Networks of Multiagents With Delays

Product theory of stochastic matrices provides a powerful tool in the consensus analysis of discrete-time multiagent systems. However, the classic theory cannot deal with networks with general coupling coefficients involving negative ones, which have been discussed only in very few papers due to the technicalities involved. Motivated by these works, here we developed some new results for the products of matrices which generalize that of the classical stochastic matrices by admitting negative entries. Particularly, we obtained a generalized version of the classic Hajnal inequality on this generalized matrix class. Based on these results, we proved some convergence results for a class of discrete-time consensus algorithms with time-varying delays and general coupling coefficients. At last, these results were applied to the analysis of a class of continuous-time consensus algorithms with discrete-time controller updates in the existence of communication/actuation delays.

Design of Distributed Observers in the Presence of Arbitrarily Large Communication Delays

This paper focuses on the construction of distributed observers in the presence of arbitrarily large communication time delays. In contrast with the traditional centralized observer with the ability to acquire full output of the plant, we design a set of distributed observers, each having access to partial output of the plant through a distributed sensor network. More specifically, each observer obtains partial plant output and communicates with its neighboring observers through consensus protocols. The communication among the network is subject to arbitrarily large time delays. We consider three representative network topologies and for each topology establish conditions to guarantee the observation error systems be exponentially stable. We also consider the design of a pinning synchronization problem as a dual problem of the design of distributed observers. Numerical simulation is carried out to verify the effectiveness of our theoretical analysis.

Avoiding Congestion in Cluster Consensus of the Second-Order Nonlinear Multiagent Systems

In order to avoid congestion in the second-order nonlinear leader-following multiagent systems over capacity-limited paths, an approach called cluster lag consensus is proposed, which means that the agents in different clusters will pass through the same positions with the same velocities but lag behind the leader at different times. Lyapunov functionals and matrix theory are applied to analyze such cluster lag consensus. It is shown that when the graphic roots of clusters are influenced by the leader and the intracoupling of cluster agents is larger than a threshold, the cluster lag consensus can be achieved. Furthermore, the cluster lag consensus with a time-varying communication topology is investigated. Finally, an illustrative example is presented to demonstrate the effectiveness of the theoretical results. In particular, when the physical sizes of the agents are taken into consideration, it is shown that with a rearrangement and a position transformation, the multiagent system will reach cluster lag consensus in the new coordinate system. This means that all agents in the same cluster will reach consensus on the velocity, but their positions may be different and yet their relative positions converge to a constant asymptotically.

Velocity synchronization of multi-agent systems with mismatched parameters via sampled position data

Power systems are special multi-agent systems with nonlinear coupling function and symmetric structures. This paper extends these systems to a class of multi-agent systems with mismatched parameters, linear coupling function, and asymmetric structures and investigates their velocity synchronization via sampled position data. The dynamics of the agents is adopted as that of generators with mismatched parameters, while the system structures are supposed to be complex. Two distributed linear consensus protocols are designed, respectively, for multi-agent systems without or with communication delay. Necessary and sufficient conditions based on the sampling period, the mismatched parameters, the delay, and the nonzero eigenvalues of the Laplacian matrix are established. It is shown that velocity synchronization of multi-agent systems with mismatched parameters can be achieved if the sampled period is chosen appropriately. Simulations are given to illustrate the effectiveness of the theoretical results.