In this paper, we propose a new architecture to enhance the privacy and security of networked control systems against malicious adversaries. We consider an adversary which first learns the system using system identification techniques (privacy), and then performs a data injection attack (security). In particular, we consider an adversary conducting zero-dynamics attacks (ZDA) which maximizes the performance cost of the system whilst staying undetected. Using the proposed architecture, we show that it is possible to (i) introduce significant bias in the system estimates obtained by the adversary: thus providing privacy, and (ii) efficiently detect attacks when the adversary performs a ZDA using the identified system: thus providing security. Through numerical simulations, we illustrate the efficacy of the proposed architecture
In this contribution, we consider the classical problem of estimating an Output Error model given a set of input-output measurements. First, we develop a regularization method based on the re-weighted nuclear norm heuristic. We show that the re-weighting improves the estimate in terms of better fit. Second, we suggest an implementation method that helps in eliminating the regularization parameters from the problem by introducing a constant based on a validation criterion. Finally, we develop a method for considering the effect of the transient when the initial conditions are unknown. A simple numerical example is used to demonstrate the proposed method in comparison to classical and another recent method based on the nuclear norm heuristic.
An algorithm for continuous time-delay estimation from sampled output data and a known input of finite energy is presented. The continuous time-delay modeling allows for the estimation of subsample delays. The proposed estimation algorithm consists of two steps. First, the continuous Laguerre spectrum of the output (delayed) signal is estimated from discretetime (sampled) noisy measurements. Second, an estimate of the delay value is obtained via a Laguerre domain model using a continuous-time description of the input. The second step of the algorithm is shown to be intrinsically biased, the bias sources are established, and the bias itself is modeled. The proposed delay estimation approach is compared in a Monte-Carlo simulation with state-of-the-art methods implemented in time, frequency, and Laguerre domain demonstrating comparable or higher accuracy in the considered scenario.
Identifcation of dynamic networks has attracted considerable interest recently. So far the main focus has been on linear time-invariant networks. Meanwhile, most real-life systems exhibit nonlinear behaviors; consider, for example, two stochastic linear time-invariant systems connected in series, each of which has a nonlinearity at its output. The estimation problem in this case is recognized to be challenging, due to the analytical intractability of both the likelihood function and the optimal one-step ahead predictors of the measured nodes. In this contribution, we introduce a relatively simple prediction error method that may be used for the estimation of nonlinear dynamical networks. The estimator is defined using a deterministic predictor that is nonlinear in the known signals. The estimation problem can be defined using closed-form analytical expressions in several non-trivial cases, and Monte Carlo approximations are not necessarily required. We show, that this is the case for some block-oriented networks with no feedback loops and where all the nonlinear modules are polynomials. Consequently, the proposed method can be applied in situations considered challenging by current approaches. The performance of the estimation method is illustrated on a numerical simulation example.
The estimation problem of stochastic Wiener-Hammerstein models is recognized to be challenging, mainly due to the analytical intractability of the likelihood function. In this contribution, we apply a computationally attractive prediction error method estimator to a real-data stochastic Wiener-Hammerstein benchmark problem. The estimator is defined using a deterministic predictor that is nonlinear in the input. The prediction error method results in tractable expressions, and Monte Carlo approximations are not necessary. This allows us to tackle several issues considered challenging from the perspective of the current mainstream approach. Under mild conditions, the estimator can be shown to be consistent and asymptotically normal. The results of the method applied to the benchmark data are presented and discussed.
Nonlinear stochastic parametric models are widely used in various fields. However, for these models, the problem of maximum likelihood identification is very challenging due to the intractability of the likelihood function. Recently, several methods have been developed to approximate the analytically intractable likelihood function and compute either the maximum likelihood or a Bayesian estimator. These methods, albeit asymptotically optimal, are computationally expensive. In this contribution, we present a simulation-based pseudo likelihood estimator for nonlinear stochastic models. It relies only on the first two moments of the model, which are easy to approximate using Monte-Carlo simulations on the model. The resulting estimator is consistent and asymptotically normal. We show that the pseudo maximum likelihood estimator, based on a multivariate normal family, solves a prediction error minimization problem using a parameterized norm and an implicit linear predictor. In the light of this interpretation, we compare with the predictor defined by an ensemble Kalman filter. Although not identical, simulations indicate a close relationship. The performance of the simulated pseudo maximum likelihood method is illustrated in three examples. They include a challenging state-space model of dimension 100 with one output and 2 unknown parameters, as well as an application-motivated model with 5 states, 2 outputs and 5 unknown parameters.
Event-based methods carefully select when to transmit information to enable high-performance control and estimation over resource-constrained communication networks. However, they come at a cost. For instance, event-based communication induces a higher computational load and increases the complexity of the scheduling problem. Thus, in some cases, allocating available slots to agents periodically in circular order may be superior. In this article, we discuss, for a specific example, when the additional complexity of event-based methods is beneficial. We evaluate our analysis in a synthetical example and on 20 simulated cart-pole systems.
This paper considers the remote state estimation in a cyber-physical system (CPS) using multiple sensors. The measurements of each sensor are transmitted to a remote estimator over a shared channel, where simultaneous transmissions from other sensors are regarded as interference signals. In such a competitive environment, each sensor needs to choose its transmission power for sending data packets taking into account of other sensors’ behavior. To model this interactive decision-making process among the sensors, we introduce a multi-player non-cooperative game framework. To overcome the inefficiency arising from the Nash equilibrium (NE) solution, we propose a correlation policy, along with the notion of correlation equilibrium (CE). An analytical comparison of the game value between the NE and the CE is provided, with/without the power expenditure constraints for each sensor. Also, numerical simulations demonstrate the comparison results.
This paper addresses the detection and isolation of replay attacks on sensor measurements. As opposed to previously proposed additive watermarking, we propose a multiplicative watermarking scheme, where each sensor’s output is separately watermarked by being fed to a SISO watermark generator. Additionally, a set of equalizing filters is placed at the controller’s side, which reconstructs the original output signals from the received watermarked data. We show that the proposed scheme has several advantages over existing approaches: it has no detrimental effects on the closed-loop performance in the absence of attacks; it can be designed in a modular fashion, independently of the design of the controller and anomaly detector; it facilitates the detection of replay attacks and the isolation of the time at which the replayed data was recorded. These properties are discussed in detail and the results are illustrated through a numerical example.
This paper addresses the problem of computing fixed interval smoothed state estimates of a linear time varying Gaussian stochastic system. There already exist many algorithms that perform this computation, but all of them impose certain restrictions on system matrices in order for them to be applicable. This paper develops a new forwards–backwards pass algorithm that is applicable under the mildest restrictions possible - namely that the smoothed state distribtions exists in forms that can be characterised by means and covariances, for which this paper also develops a new necessary and sufficient condition.
This paper studies a wireless sensor system, where packeted information is transmitted to a receiver over a fading channel. The sensor's transmission energy can be chosen within a given interval. To obtain the current channel gain information, the sensor may listen to a pilot signal from the receiver and estimate the channel, which requires energy. When the sensor does not listen to the pilot signal, the transmission energy has to be estimated solely based on other information like the channel gain distribution and past channel information. The sensor is equipped with an energy harvester and a rechargeable battery to temporarily store the harvested energy. The paper derives the optimal energy allocation policy, shows structural properties of the optimal solution and derives a heuristic suboptimal energy allocation policy. The performance of the optimal policy is compared with several other simple policies through numerical examples.
This paper investigates methods for quantitatively examining the connectivity and knowledge flow in a university program considering courses and concepts included in the program. The proposed method is expected to be useful to aid program design and inventory, and for communicating what concepts a course may rely on at a given point in the program. As a first step, we represent the university program as a directed graph with courses and concepts as nodes and connections between courses and concepts as directed edges. Then, we investigate the connectivity and the flow through the graph in order to gain insights into the structure of the program. We thus perform two investigations based on data collected from an engineering program at a Swedish university: a) how to represent (parts of) the university program as a graph (here called Directed Courses-Concepts Graph (DCCG)), and b) how to use graph theory tools to analyse the coherence and structure of the program.
Low-voltage distribution grids experience a rising penetration of inverter-based, distributed generation. In order to not only contribute to but also solve voltage problems, these inverters are increasingly asked to participate in intelligent grid controls. Communicating inverters implement distributed voltage droop controls. The impact of cyber-attacks to the stability of such distributed grid controls is poorly researched and therefore addressed in this article. We characterize the potential impact of several attack scenarios by employing the positivity and diagonal dominance properties. In particular, we discuss measurement falsification scenarios where the attacker corrupts voltage measurement data received by the voltage droop controllers. Analytical, control-theoretic methods for assessing the impact on system stability and voltage magnitude are presented and validated via simulation.
The uncertainty in the prediction calculated using the delta method for an over-parameterized (parametric) black-box model is shown to be larger or equal to the uncertainty in the prediction of a canonical (minimal) model. Equality holds if the additional parameters of the overparameterized model do not add flexibility to the model. As a conclusion, for an overparameterized black-box model, the calculated uncertainty in the prediction by the delta method is not underestimated. The results are shown analytically and are validated in a simulation experiment where the relationship between the normalized traction force and the wheel slip of a car is modelled using e.g., a neural network.
This paper presents a systematic approach to design a hybrid oscillator that admits an orbitally stable periodic solution of a certain type with pre-defined parameters. The parsimonious structure of the Impulsive Goodwin's oscillator (IGO) is selected for the implementation due to its well-researched rich nonlinear dynamics. The IGO is a feedback interconnection of a positive third-order continuous-time LTI system and a nonlinear frequency and amplitude impulsive modulator. A design algorithm based on solving a bilinear matrix inequality is proposed yielding the slope values of the modulation functions that guarantee stability of the fixed point defining the designed periodic solution. Further, assuming Hill function parameterizaton of the pulse-modulated feedback, the parameters of those rendering the desired stationary properties are calculated. The character of perturbed solutions in vicinity of the fixed point is controlled through localization of the multipliers. The proposed design approach is illustrated by a numerical example. Bifurcation analysis of the resulting oscillator is performed to explore the nonlinear phenomena in vicinity of the designed dynamics.
This paper proposes a game-theoretic approach to address the problem of optimal sensor placement for detecting cyber-attacks in networked control systems. The problem is formulated as a zero-sum game with two players, namely a malicious adversary and a detector. Given a protected target vertex, the detector places a sensor at a single vertex to monitor the system and detect the presence of the adversary. On the other hand, the adversary selects a single vertex through which to conduct a cyber-attack that maximally disrupts the target vertex while remaining undetected by the detector. As our first contribution, for a given pair of attack and monitor vertices and a known target vertex, the game payoff function is defined as the output-to-output gain of the respective system. Then, the paper characterizes the set of feasible actions by the detector that ensures bounded values of the game payoff. Finally, an algebraic sufficient condition is proposed to examine whether a given vertex belongs to the set of feasible monitor vertices. The optimal sensor placement is then determined by computing the mixed-strategy Nash equilibrium of the zero-sum game through linear programming. The approach is illustrated via a numerical example of a 10-vertex networked control system with a given target vertex.
Efficiently planning trajectories for nonholonomic mobile robots in formation tracking is a fundamental yet challenging problem. Nonholonomic constraints, complexity in collision avoidance, and limited computing resources prevent the robots from being practically deployed in realistic applications. This paper addresses these difficulties by modeling each mobile platform as a nonholonomic motion and formulating trajectory planning as an optimization problem using model predictive control (MPC). That is, the optimization problem is subject to both nonholonomic motions and collision avoidance. To reduce computing costs in real time, the nonholonomic constraints are convexified by finding the closest nominal points to the nonholonomic motion, which are then incorporated into a convex optimization problem. Additionally, the predicted values from the previous MPC step are utilized to form linear avoidance conditions for the next step, preventing collisions among robots. The formulated optimization problem is solved by the alternating direction method of multiplier (ADMM) in a distributed manner, which makes the proposed trajectory planning method scalable. More importantly, the convergence of the proposed planning algorithm is theoretically proved while its effectiveness is validated in a synthetic environment.
Distributed fault diagnosis has been proposed as an effective technique for monitoring large scale, nonlinear and uncertain systems. It is based on the decomposition of the large scale system into a number of interconnected subsystems, each one monitored by a dedicated Local Fault Detector (LFD). Neighboring LFDs, in order to successfully account for subsystems interconnection, are thus required to communicate with each other some of the measurements from their subsystems. Anyway, such communication may expose private information of a given subsystem, such as its local input. To avoid this problem, we propose here to use differential privacy to pre-process data before transmission.
Proper monitoring of large complex spatially critical infrastructures often requires a sensor network capable of inferring the state of the system. Such networks enable the design of distributed estimators considering only local (partial) measurements, local communication capabilities with nearby sensors, as well as the system model. Solutions often assume perfect knowledge of the system model, and white process and measurement noise, which are limiting in engineering settings. In this paper, we consider the minimum energy setting where the model uncertainty and process and measurement noises are bounded but unknown. We provide the first distributed minimum energy estimator for discrete-time linear time-invariant systems, and we show that the error dynamics is input-to-state stable. Lastly, we illustrate the performance in some pedagogical examples, and compare the performance with respect to the centralized implementation of the minimum energy estimator.
We propose models for the decision-making process of human drivers in an overtaking scenario. First, we mathematically formalize the overtaking problem as a decision problem with perceptual uncertainty. Then, we propose and numerically analyze risk-agnostic and risk-aware decision models, which are able to judge whether an overtaking is desirable or not. We show how a driver's decision-making time and confidence level can be primarily characterized through two model parameters, which collectively represent human risk-taking behavior. We detail an experimental testbed for evaluating the decision-making process in the overtaking scenario. Finally, we present some preliminary experimental results from two human drivers.
Standard identification methods give biased parameter estimates when the recorded signals are corrupted by noise on both input and output sides. When the system is close to be non-identifiable, the bias can be large. The paper discusses the possibilities and potential benefits when using either a reduced model structure or a full errors-in-variables model. The case of using an instrumental variable estimator is also treated.
Genital pain / penetration disorders affect a significant portion of the female population and diminish significantly the quality of life of the subjects. Treatments, that often consist in stretching opportunely the vaginal duct by means of opportune vaginal dilators, are known to be invasive, lengthy and uncomfortable. Designing better treatments (e.g., more efficient locations and levels of pressures) nonetheless requires understanding better how the pressure developed in the vaginal channel affects the patient and leads to subjective pain. Here we take a control-oriented approach to the problem, and aim at describing the dynamics of the pressure vs. pain mechanisms by means of opportune state space representations. In particular, we first collect and discuss the medical literature, that describes how the variables that are involved in the treatment of genital pain / penetration disorders with vaginal dilators, are logically related. After this we translate (and complete) this set of logical relations into a qualitative model that allows control oriented analyses of the dynamics. The obtained state space model is then proved to both mimic correctly what is expected from logical perspectives and reproduce behaviors measured in clinical settings.
The paper discusses attacks on networked control loops by increased delay, and shows how existing round trip jitter may disguise such attacks. The attackers objective need not be de-stabilization, the paper argues that making settling time requirements fail can be sufficient. To defend against such attacks, the paper proposes the use of joint recursive prediction error identification of the round trip delay and the networked closed loop dynamics. The proposed identification algorithm allows general defense, since it is designed for delayed nonlinear dynamics in state space form. Simulations show that the method is able to detect a delay attack on a printed circuit board component mounting servo loop, long before the attack reaches full effect.
This paper deals with continuous plants subject to intrinsic pulse-modulated feedback, thus exhibiting hybrid closed-loop dynamics. The system structure implements a hybrid oscillator and arises in living organisms, e.g. when episodically firing neurons control the production of hormones in endocrine glands. Hybrid observers reconstructing both the continuous and discrete states of the hybrid plant from only continuous measured outputs are considered. They excel over the existing solutions through the introduction of two co-ordinated feedbacks of the output estimation error: one correcting the continuous state estimates and another adjusting the discrete ones. Different types of the feedback operator to the discrete estimates are analyzed. The observer design problem is reduced to synchronization of the observer solution with that of the plant. The synchronous mode of the observer is rendered locally stable by the selection of the feedback gains. Numerical illustration of the design procedure and observer performance with respect to a pulse-modulated model of testosterone regulation is provided.
While reinforcement learning has made great improvements, state-of-the-art algorithms can still struggle with seemingly simple set-point feedback control problems. One reason for this is that the learned controller may not be able to excite the system dynamics well enough initially, and therefore it can take a long time to get data that is informative enough to learn for good control. The paper contributes by augmentation of reinforcement learning with a simple guiding feedback controller, for example, a proportional controller. The key advantage in set point control is a much improved excitation that improves the convergence properties of the reinforcement learning controller significantly This can be very important in real-world control where quick and accurate convergence is needed. The proposed method is evaluated with simulation and on a real-world double tank process with promising results.
The paper proposes the use of structured neural networks for reinforcement learning based nonlinear adaptive control. The focus is on partially observable systems, with separate neural networks for the state and feedforward observer and the state feedback and feedforward controller. The observer dynamics are modelled by recurrent neural networks while a standard network is used for the controller. As discussed in the paper, this leads to a separation of the observer dynamics to the recurrent neural network part, and the state feedback to the feedback and feedforward network. The structured approach reduces the computational complexity and gives the reinforcement learning based controller an understandable structure as compared to when one single neural network is used. As shown by simulation the proposed structure has the additional and main advantage that the training becomes significantly faster. Two ways to include feedforward structure are presented, one related to state feedback control and one related to classical feedforward control. The latter method introduces further structure with a separate recurrent neural network that processes only the measured disturbance. When evaluated with simulation on a nonlinear cascaded double tank process, the method with most structure performs the best, with excellent feedforward disturbance rejection gains.