9 September 2022 in IEEE Transactions on Automatic Control
For single agent systems, probabilistic machine learning techniques such as Gaussian process regression have been shown to be suitable methods for inferring models of unknown nonlinearities, which can be employed to improve the performance of control laws. While this approach can be extended to the cooperative control of multiagent systems, it leads to a decentralized learning of the unknown nonlinearity, i.e., each agent independently infers a model. However, decentralized learning can potentially lead to poor control performance, since the models of individual agents are often accurate in merely a small region of the state space. In order to overcome this issue, we propose a novel method for the distributed aggregation of Gaussian process models, and extend probabilistic error bounds for Gaussian process regression to the proposed approach. Based on this distributed learning method, we develop a cooperative tracking control law for leader–follower consensus of multiagent systems with partially unknown, higher order, control-affine dynamics, and analyze its stability using the Lyapunov theory. The effectiveness of the proposed methods is demonstrated in numerical evaluations.
14 June 2019 in IEEE Transactions on Automatic Control
This paper is concerned with a suboptimal version of the distributed linear quadratic optimal control problem for multi-agent systems. Given a multi-agent system with identical agent dynamics and an associated global quadratic cost functional, our objective is to design distributed control laws that achieve consensus and whose cost is smaller than an a priori given upper bound, for all initial states of the network that are bounded in norm by a given radius. A centralized design method is provided to compute such suboptimal controllers, involving the solution of a single Riccati inequality of dimension equal to the dimension of the agent dynamics, and the smallest nonzero and the largest eigenvalue of the Laplacian matrix. Furthermore, we relax the requirement of exact knowledge of the smallest nonzero and largest eigenvalue of the Laplacian matrix by using only lower and upper bounds on these eigenvalues. Finally, a simulation example is provided to illustrate our design method.
11 June 2019 in IEEE Control Systems Letters
In this paper we consider the distributed linear quadratic control problem for networks of agents with single integrator dynamics. We first establish a general formulation of the distributed LQ problem and show that the optimal control gain depends on global information on the network. Thus, the optimal protocol can only be computed in a centralized fashion. In order to overcome this drawback, we propose the design of protocols that are computed in a decentralized way. We will write the global cost functional as a sum of local cost functionals, each associated with one of the agents. In order to achieve ‘good’ performance of the controlled network, each agent then computes its own local gain, using sampled information of its neighboring agents. This decentralized computation will only lead to suboptimal global network behavior. However, we will show that the resulting network will reach consensus.
15 June 2022 in IEEE Control Systems Letters
In this letter, we present a trajectory tracking control law for a class of partially unknown nonlinear systems that combines backstepping and event-triggered online learning. We employ Gaussian processes to learn the unknown system model using measurement data collected online, while the proposed control law is active. Our approach uses an efficient event-triggered online learning scheme that exclusively collects informative data to update the estimated model used for control. The resulting control law guarantees that the tracking error is globally uniformly ultimately bounded. The inter-event time is shown to be lower-bounded by a positive constant. Moreover, we also discuss how to obtain a trade-off between the cardinality of the collected training data and the size of the ultimate tracking error bound. In a simulation example, our approach is shown to outperform a state-of-the-art offline learning-based approach both in terms of tracking performance and data efficiency.
18 February 2025 in IEEE Transactions on Automatic Control
This article considers a localized data-driven consensus problem for leader–follower multiagent systems with unknown discrete-time agent dynamics, where each follower computes its local control gain using only their locally collected state and input data. Both noiseless and noisy data-driven protocols are presented to achieve leader–follower consensus, by addressing the challenge of the heterogeneity in control gains caused by the localized data sampling and distinct parameters of agents. The design of these data-driven consensus protocols involves low-dimensional linear matrix inequalities. In addition, the results are extended to the case where only the leader's data are collected and exploited. The effectiveness of the proposed methods is illustrated via simulation examples.
21 June 2022 in IEEE Transactions on Automatic Control
This article investigates the $\mathcal {H}_{2}$ and $\mathcal {H}_\infty$ suboptimal distributed filtering problems for continuous-time linear systems. We consider a linear system monitored by a number of filters, where each of the filters receive only part of the measured output of the system. Each filter can communicate with the other filters according to a priori given strongly connected weighted directed graph. The aim is to design filter gains that guarantee the $\mathcal {H}_{2}$ or $\mathcal {H}_\infty$ norm of the transfer matrix from the disturbance input to the output estimation error to be smaller than a priori given upper bound, while all local filters reconstruct the full system state asymptotically. We provide a centralized design method for obtaining such $\mathcal {H}_{2}$ and $\mathcal {H}_\infty$ suboptimal distributed filters. The proposed design method is illustrated by a simulation example.