Automatica最新文献

With the advent of integrated sensor technology (smart flow meters and pressure sensors), various new numerical algorithms for leak localization (a core element of water distribution system operation) have been developed. However, there is a lack of theory regarding the limitations of leak localization. In this work, we contribute to the development of such a theory by introducing an example water network structure with parallel pipes that is tractable for analytical treatment. We define the leak localization problem for this structure and show how many sensors and what conditions are needed for the well-posedness of the problem. We present a formula for the leak position as a function of measurements from these sensors. However, we also highlight the risk of finding false but plausible leak positions in the multiple pipes. We try to answer the questions of how and when the leaking pipe can be isolated. In particular, we show that nonlinearities in the pipes’ head loss functions are essential for the well-posedness of the isolation problem. We propose procedures to get around the pitfall of multiple plausible leak positions.

Recent works have demonstrated that the computational demand of Nonlinear Model Predictive Control (NMPC) can be alleviated when quasi-Linear Parameter Varying (qLPV) models are used. Yet, a difficulty arises: the future scheduling parameter values are typically unavailable. In Morato et al. (2022), a recursive extrapolation method, based on local Taylor expansions of the qLPV function, is proposed. In this communique, we investigate the convergence of such scheme by applying the analysis procedure originally proposed by Hespe and Werner (2021), w.r.t. the MPC from Cisneros et al. (2016), using an inexact Newton root determination problem. We generate an upper bound for the rate of contraction over samples, considering the scheduling trajectories estimation error. Based on simulation results, we advocate that the Taylor-based qLPV MPC technique from Morato et al. (2022) is indeed a highly competitive NMPC solution: similar to state-of-the-art solvers, with relieved numerical complexity (one QP per sample).

This paper proposes a dual opinions co-evolution model based on the dual attitudes theory in social psychology, where every individual has dual opinions of an object: implicit and explicit opinions. The implicit opinions are individuals’ inner evaluations affected by the experience, while the explicit opinions are external expressions of the evaluations and are influenced by public opinion. We theoretically analyze the dynamics of coupled dual opinions in an adversarial discussion scenario, in which two opposing groups participate in the discussion. We rigorously give a method to grade the degree of individual bias. When individuals have a slight bias, dual opinions reach a consensus, showing the phenomenon of acceptance. When individuals are moderately or even extremely biased, they will exhibit the phenomenon of outward compliance, that is, their publicly explicit opinions conform to public opinion, while implicit opinions will reach a persistent disagreement and even polarization. Our investigation indicates that dual opinions may not be monotone initially but eventually converge monotonically. Besides, we analyze the influence of parameters on the co-evolutionary process and results. Numerical simulations of the dual opinions formation process produce inward and outward conformity phenomena, namely acceptance and compliance.

In this paper we address decentralized control problems in the context of sustainable management of discrete-time deterministic dynamical systems. We consider $N$ players, each with their own controlled dynamical system, which are linked through the dynamics of a global variable. We consider sustainable constraints as constraints in the state space. From this general problem defined in both viability theory and game theory, we define $N$ viability sub-problems. For each of these, we consider the guaranteed viability kernel, subset of the constraint set where it is possible to maintain the subsystem whatever the strategy of other players. When these kernels are not empty, we propose a general algorithm to compute individual strategies that ensure the sustainability of all. Numerical experiments on a model of agricultural cooperative show the feasibility of our approach and suggest that it can be the basis for the definition of new operating rules when the sustainability constraints cannot be ensured.

This article studies a fundamental problem of security of cyber–physical systems (CPSs): that of detecting, almost surely, the presence of malicious components in the CPS. We assume that some of the actuators may be malicious while all sensors are honest. We introduce a novel idea of separability of state trajectories generated by CPSs in two situations: those under the nominal no-attack situation and those under the influence of an attacker. We establish its connection to the security of CPSs, particularly in detecting the presence of malicious actuators (if any) in them. As primary contributions, we establish necessary and sufficient conditions for the aforementioned detection in CPSs modeled as Markov decision processes (MDPs). Moreover, we focus on the mechanism of perturbing the pre-determined control policies of the honest agents in CPSs modeled as stochastic linear systems, by injecting a certain class of random process called private excitation; sufficient conditions for detectability and non-detectability of the presence of malicious actuators, assuming that the policies are randomized history-dependent and randomized Markovian, are established. Several technical aspects of our results are discussed extensively.

This study investigates different observer architectures for the state estimation of linear aperiodic sampled-data systems and formally shows that a complex structure is not necessarily better than a simpler one with respect to the considered cost functional and to a class of timer-dependent Lyapunov functions. The work also shows how to design the state observer through sum of squares programming, presenting a numerical example.

Hybrid integrator-gain systems (HIGS) are hybrid control elements used to overcome fundamental performance limitations of linear time-invariant feedback control, and have enjoyed recent successes in engineering applications such as high-precision motion systems. However, despite the relevance of digital implementations, the creation of sampled-data versions of HIGS and their formal analysis have not been addressed in the literature so far, and will form the topic of the present paper. Thereto, we present discrete-time HIGS elements, which preserve the main philosophy behind the operation of HIGS in continuous time. Moreover, stability criteria are presented that can be used to certify input-to-state stability of discrete-time and sampled-data HIGS-controlled systems based on both (i) (measured) frequency response data, and (ii) linear matrix inequalities (LMIs). A comparison between these stability criteria is presented as well. A numerical case study is provided to illustrate the application of the main results.

In this paper, we investigate the consensus tracking control for a class of heterogeneous multi-agent systems with multiple unknown time-varying control directions and unknown direction actuator faults. Different from existing work, the directions of the multiple time-varying control coefficients are subject to fault-induced sign-switching. To address this challenge, a series of high-order Lyapunov functions and differentiable functions are introduced to avoid non-integrable terms. Then, a novel contradiction statement and some Nussbaum functions are used to handle the summation of multiple unknown control coefficients with time-varying amplitudes and directions. Meanwhile, a novel distributed observer with quantized communication is introduced to track a reference trajectory with unknown dynamics. It is shown that all closed-loop signals are globally uniformly bounded and the tracking errors converge to residual sets that can be made arbitrarily small. Simulation results illustrate the effectiveness of the proposed scheme.

To fit with a digital environment in networked control systems, sampled-data event-triggered control (ETC) has been proposed where the event-triggering mechanisms sense and process information only at discrete sampling instants (not necessarily periodically). One deficiency in the previous study on sampled-data ETC is the lack of analysis on transmission performance, which yields the potential occurrence of an undesirable phenomenon that the minimum inter-event time is always equal to the minimum inter-sampling interval, no matter how small the latter is. To overcome this drawback, in this paper, we propose a novel sampled-data ETC scheme such that the lower bounds of inter-event times have sampling-independent positive guarantees. A linear networked control system is considered under observer-based controllers, external disturbances, and multiple communication channels. The improved transmission performance is due to the utilization of dynamic event-triggering conditions and some mean-rate error signals, which are the ratios between network-induced errors and some time-increasing functions. After modeling the closed-loop dynamics into a hybrid system, sufficient conditions on the upper bound of inter-sampling intervals and parameters in ETC are provided to ensure input-to-state stability (rather than its practical version) and sampling-independent positive guarantees simultaneously. Furthermore, a new tradeoff relationship between the sampling and transmission performance is revealed: a faster sampling frequency is conducive to improving inter-event times. Finally, a linearized model of an inverted pendulum is simulated to illustrate the efficiency and feasibility of the obtained results.

This paper presents a novel approach to multi-agent reinforcement learning (RL) for linear systems with convex polytopic constraints. Existing work on RL has demonstrated the use of model predictive control (MPC) as a function approximator for the policy and value functions. The current paper is the first work to extend this idea to the multi-agent setting. We propose the use of a distributed MPC scheme as a function approximator, with a structure allowing for distributed learning and deployment. We then show that Q-learning updates can be performed distributively without introducing nonstationarity, by reconstructing a centralized learning update. The effectiveness of the approach is demonstrated on a numerical example.