Annual Reviews in Control最新文献

Risk-sensitive control has received considerable interest since the seminal work of Howard and Matheson (Howard and Matheson, 1971/72) because of its ability to account for fluctuations about the mean, its connection with $H_{\infty }$ control, and its application to financial mathematics. In this article we attempt to put together a comprehensive survey on the research done on ergodic risk-sensitive control over the last four decades.

So-called robot–object Lagrangian systems consist of a class of nonsmooth underactuated complementarity Lagrangian systems, with a specific structure: an “object” and a “robot”. Only the robot is actuated. The object dynamics can thus be controlled only through the action of the contact Lagrange multipliers, which represent the interaction forces between the robot and the object. Juggling, walking, running, hopping machines, robotic systems that manipulate objects, tapping, pushing systems, kinematic chains with joint clearance, crawling, climbing robots, some cable-driven manipulators, and some circuits with set-valued nonsmooth components, belong this class. This article aims at presenting their main features, then many application examples which belong to the robot–object class, then reviewing the main tools and control strategies which have been proposed in the Automatic Control and in the Robotics literature. Some comments and open issues conclude the article.

A longstanding open problem of systems neuroscience is to understand how the brain calibrates thousands of reflexes to achieve near instantaneous disturbance rejection. While reflexes typically act locally at the site of sensory measurements, the adaptation of reflex gains is the result of an ingenious architecture in which knowledge of disturbances is transferred from the cerebellum to the deep cerebellar nuclei or the brainstem. This paper investigates the use of control theory as the mathematical foundation to explain the mechanisms by which such forms of learning, as well as forgetting, manifest themselves in systems neuroscience. Particularly, we use adaptive control and averaging theory to model the computations performed in learning appropriate reflex gains. While forgetting is perceived as counter-productive to learning, we show that if incorporated correctly, it can endow the much needed robustness to train thousands of reflexes without interfering with their adaptation. This is accomplished using the $\mu$-modification which achieves robustness of adaptive schemes through the estimation of exciting subspaces. Our techniques are combined in a comprehensive model, with simulations illustrating their effectiveness.

We propose a method for designing policies for convex stochastic control problems characterized by random linear dynamics and convex stage cost. We consider policies that employ quadratic approximate value functions as a substitute for the true value function. Evaluating the associated control policy involves solving a convex problem, typically a quadratic program, which can be carried out reliably in real-time. Such policies often perform well even when the approximate value function is not a particularly good approximation of the true value function. We propose value-gradient iteration, which fits the gradient of value function, with regularization that can include constraints reflecting known bounds on the true value function. Our value-gradient iteration method can yield a good approximate value function with few samples, and little hyperparameter tuning. We find that the method can find a good policy with computational effort comparable to that required to just evaluate a control policy via simulation.

In this paper, we focus on the optimal operation of a multi-agent system affected by uncertainty. In particular, we consider a cooperative setting where agents jointly optimize a performance index compatibly with individual constraints on their discrete and continuous decision variables and with coupling global constraints. We assume that individual constraints are affected by uncertainty, which is known to each agent via a private set of data that cannot be shared with others. Exploiting tools from statistical learning theory, we provide data-based probabilistic feasibility guarantees for a (possibly sub-optimal) solution of the multi-agent problem that is obtained via a decentralized/distributed scheme that preserves the privacy of the local information. The generalization properties of the data-based solution are shown to depend on the size of each local dataset and on the complexity of the uncertain individual constraint sets. Explicit bounds are derived in the case of linear individual constraints. A comparative analysis with the cases of a common dataset and of local uncertainties that are independent is performed.

In this survey, we describe controlled interacting particle systems (CIPS) to approximate the solution of the optimal filtering and the optimal control problems. Part I of the survey is focussed on the feedback particle filter (FPF) algorithm, its derivation based on optimal transportation theory, and its relationship to the ensemble Kalman filter (EnKF) and the conventional sequential importance sampling–resampling (SIR) particle filters. The central numerical problem of FPF—to approximate the solution of the Poisson equation—is described together with the main solution approaches. An analytical and numerical comparison with the SIR particle filter is given to illustrate the advantages of the CIPS approach. Part II of the survey is focussed on adapting these algorithms for the problem of reinforcement learning. The survey includes several remarks that describe extensions as well as open problems in this subject.

This article focuses on extending, disseminating and interpreting the findings of an IEEE Control Systems Society working group looking at the role of control theory and engineering in solving some of the many current and future societal challenges. The findings are interpreted in a manner designed to give focus and direction to both future education and research work in the general control theory and engineering arena, interpreted in the broadest sense. The paper is intended to promote discussion in the community and also provide a useful starting point for colleagues wishing to re-imagine the design and delivery of control-related topics in our education systems, especially at the tertiary level and beyond.

This article provides a comprehensive and illustrative presentation of the young field of encrypted control. In particular, we survey the evolution of encrypted controllers from their first appearance in 2015 until 2023 and derive a categorization into two generations mainly characterized by the utilized cryptographic methods. We further envision future developments and challenges of encrypted control. Throughout our presentation, we build less on technicalities but rather on intuitive tutorial-style explanations. This way, we intend to build a bridge from control engineering to cryptography and to make the interdisciplinary field of encrypted control more accessible.