ABSTRACT. This paper proposes a simple constrained proportional controller with gain scheduling for simultaneous thermal and state-of-charge (SOC) balancing of a multilevel converter based modular battery. The proposed balancing controller is devised by investigating structural properties of constrained linear quadratic (LQ) model predictive controller (MPC) introduced in our earlier study. This investigation reveals a particular factorization of time-varying control gain matrices, which leads to approximation of matrix gains as scalar gains under the assumption of small parametric variations among battery cells. The gains are scheduled in load current for nominal cells. This special structure enables the identification of two dominant operational modes of the balancing controller: SOC balancing mode in low to medium load current range and thermal balancing mode in high current range. This study also proposes a simple algorithm for control projection on constraint polytope. The proposed balancing controller is tested in simulations for a modular battery with four significantly mismatched cells. The performance is comparable to MPC, which uses true battery parameters. The performance and the simplicity of the controller make it attractive for real-time implementation in large battery packs.

ABSTRACT. We would like to present our progress in the project "Integration Of Optimal Energy Management In Vehicle". The purpose of the project is to apply the latest findings in predictive energy management to control the powertrain and the auxiliary vehicle systems. The goal is to the implement in a real truck an energy management algorithm that would take into account both kinetic buffer and vehicles auxiliary system. The algorithm is expected to give 6-8% fuel savings compared to the vehicle without predictive energy management algorithms.

ABSTRACT. To secure safety, reliability and performance of an electri- fied vehicle, it is important to monitor the State of Charge (SoC) of its battery. Today, there are no sensors that can measure SoC directly. Instead, it is usually estimated with an algorithmic filter. Since batteries are nonlinear, all feasible filters are only able to approximate the posterior densities which, in other words, means that their performances will be more or less suboptimal. To be able to evaluate the performance of a filter, it is of great value to know how well a parameter or a state can be estimated. It can then be decided if it is worth spending time on tuning the filter, or implementing a more advanced filter. Furthermore, analyzing the achievable accuracy can be a way to better understand the application. One suitable measure for benchmarking the performance is the Cramér-Rao Lower Bound (CRLB), which is a lower bound on the Mean Square Error (MSE) of any estimator. In this paper we adopt a method to numerically determine the posterior CRLBs with a Monte Carlo-based algorithm. The posterior CRLBs are calculated for combined esti- mation of the states and the parameters of a commonly used equivalent circuit model. It is investigated how the posterior CRLBs depend on the amplitude and the fre- quency of the current. Furthermore, the posterior CRLBs are computed for a commercially available lithium- ion battery using data from laboratory experiments, and the results are compared to the MSEs of an Extended Kalman Filter (EKF). It is shown that the MSEs of the EKF are close to the posterior CRLBs, which means that the EKF seems to be a good observer for this application.

ABSTRACT. This paper presents a rule-based battery discharge strategy for a Plug-in Hybrid Electric Vehicle. The main idea is to exploit results of previous research and synthesize a strategy that mimic optimal battery State of Charge (SoC) discharge behaviour. A module called the Target SoC Optimizer computes a sequence of target SoCs at specied driving distances along a given trip prediction. A second module, called the Power Limit Optimizer, is then responsible for adapting the engine on/o power limits so that the upcoming target SoC is reached. Simulations and vehicle tests in a vehicle dynamometer show that the proposed strategy can decrease fuel consumption with more than 5% compared to the nominal Depletion-Sustain discharge strategy.

ABSTRACT. The object of this research is the extension of kernel-based system identification methods to systems with input uncertainties. We formulate a general framework and cast various classical problems, like blind identification and Hammerstein nonlinear systems, as indentification of models with inpu,t uncertainties.

ABSTRACT. This work analyses the Kalman filter in a {\em minimax} setting. Borrowing from the framework of modern online learning, we phrase the performance of the filter in terms of the {\em regret} of not knowing the true sequence of system states. We derive worst-case bounds that hold uniformly for arbitrary sequences of states. This result is achieved without making any stochastic assumptions on the underlying dynamical system. The bounds are compared to results obtained in an $H_\infty$ framework.

ABSTRACT. In this extended abstract we investigate how the framework for rational covariance extension can be adapted to handle noisy covariance estimates. Covariance estimation errors impede the accuracy of the resulting model and in some cases there does not even exist a positive extension consistent with an estimated covariance sequence. To deal with this, we consider the use of regularization to model such errors. This approach is compared with using exact matching to the biased covariance estimate, which is extendable.

ABSTRACT. The Ensemble Kalman filter (EnKF) is used to estimate the state of extremely high-dimensional models in geoscientific problems. Regardless of its popularity and the many citations in the meteorology and oceanography literature, it has received only little attention by control and signal processing researchers. The ever increasing amount of sensor data to process and the inherent need for scalable algorithms, however, motivates for a closer look at the EnKF. We derive the EnKF as Monte Carlo realization of the Kalman filter and show that its sampling based nature makes it applicable to nonlinear problems without any changes. Its relations to other nonlinear filters such as sigma point Kalman filters and the particle filter are explored. A simulation example shows the EnKF performance in a space debris tracking application.

ABSTRACT. In Svensson and Schön (2016), we suggest an approach to nonlinear system identification by parameterizing the state transition and observation functions in a state space model as basis function expansions, and estimate the coefficients in the basis function expansions from data. With a connection to Gaussian processes we also develop priors on the coefficients for tuning the model flexibility and to prevent overfitting to data. The priors can be seen as a regularization, and helps the model in generalizing the data without sacrificing the richness offered by the basis function expansion. To perform the identification efficiently, we tailor an algorithm using state-of-the-art particle methods. The approach indicates promising results when evaluated on a classical benchmark as well as real data.

ABSTRACT. Developed in this note is a scheduled state-feedback controller synthesis method for discrete-time Linear Parameter Varying (LPV) systems subjected to control input saturation constraints. The static state-feedback gain is scheduled with an exact replica of the parameter matrix. The saturation eect is modeled by introducing time-varying parameters as functions of the control inputs, which are also used to schedule the controller. The synthesis method is then specialized to distributed state-feedback by imposing a particular structure on the feedback gain matrix.

ABSTRACT. A loopshaping method for designing decentralised controllers is presented. The method allows the design of each controller to be posed as an independent loopshaping problem, written in terms of only local network dynamics. Despite being entirely decentralised, the method guarantees stability and levels of performance for the full network, which may be arbitrarily large and heterogeneous.

ABSTRACT. We propose a demand/supply balance perspective to study the information requirement for networked stabilizability. In order to stabilize a networked control system, each control input requires certain communication resource for transmission. As such, the control inputs can be considered as the demand side for communication resource, while the communication channels can be considered as the supply side. The supply capabilities of the communication channels are characterized by their respective capacities. In this regard, the networked stabilization boils down to the demand/supply balance of the communication resource. We propose two approaches for the realization of such a balance. One is to tailor the supply to meet the demand via channel resource allocation. The other is to shape the demand to meet the supply via MIMO communication. Through the two approaches, we establish a fundamental limitation on the information constraints for networked stabilizability.

ABSTRACT. Energy management systems (EMS) are used to control energy usage in buildings and campuses, by employing technologies such as supervisory control and data acquisition (SCADA) and building management systems (BMS), in order to provide reliable energy supply and maximise user comfort while minimising energy usage. Historically, EMS systems were installed when potential security threats were only physical. Nowadays, EMS systems are connected to the building network and as a result directly to the outside world. This extends the attack surface to potential sophisticated cyber-attacks, which adversely impact EMS operation, resulting in service interruption and downstream financial implications. Currently, the security systems that detect attacks operate independently to those which deploy resiliency policies and use very basic methods. We propose a novel EMS cyber-physical-security framework that executes a resilient policy whenever an attack is detected using security analytics. In this framework, both the resilient policy and the security analytics are driven by EMS data, where the physical correlations between the data-points are identified to detect outliers and then the control loop is closed using an estimated value in place of the outlier. The framework has been tested using a reduced order model of a real EMS site.

ABSTRACT. The consensus problem for multi-agent systems with quantized communication or sensing is considered. Centralized and distributed self-triggered rules are proposed to reduce the overall need of communication and system updates. It is proved that these self-triggered rules realize consensus exponentially if the network topologies have a spanning tree and the quantization function is uniform. Numerical simulations are provided to show the effectiveness of the theoretical results.

ABSTRACT. A novel unified approach to design of robust controller, gain-scheduled controller (GSC), and robust switched controller with arbitrarily switching algorithm is presented. The proposed design procedure is based on the Bellman-Lyapunov equation, guaranteed cost, and robust stability conditions using parameter-dependent quadratic stability approach. The obtained feasible design procedures for robust, GSC, and switched controller design are in the form of bilinear matrix inequalities with guaranteed convex stability conditions. The obtained design results and their properties are illustrated in simulation examples.

ABSTRACT. The selection of leaders in leader-follower multi-agent systems can be naturally formulated as a matroid optimization problem. In this paper, we investigate the online and stochastic version of such a problem, where in each iteration or round, we select a set of leaders and then observe a random realization of the corresponding reward, i.e., of the system performance. This problem is referred to as a stochastic matroid bandit, a variant of combinatorial multiarmed bandit problems where the underlying combinatorial structure is a matroid. We consider semi-bandit feedback and Bernoulli rewards, and derive a tight and problemdependent lower bound on the regret of any consistent algorithm. We propose KL-OSM, a computationally efficient algorithm that exploits the matroid structure. We derive a finite-time upper bound of the regret of KL-OSM that improves the performance guarantees of existing algorithms. This upper bound actually matches our lower bound, i.e., KL-OSM is asymptotically optimal. Numerical experiments attest that KL-OSM outperforms state-of-the-art algorithms in practice, and the difference in some cases is significant.

ABSTRACT. This paper proposes an adaptive scheduling and routing algorithm which substantially improves throughput of urban traffic networks by incorporating topological information and network state information into the traffic light control mechanism and routing of vehicles in the network. First, we introduce a first order traffic flow model in order to describe traffic flow dynamics with the help of a discretized Lightwill-Whitham-Richards (LWR) model. Next, we propose scheduling and routing algorithms which control traffic flow based on the parameters of this model. We also present an algorithm for distributed estimation of network state which enables local computation of routing decisions. Finally, we evaluate the performance of the proposed algorithms based on microscopic simulations of a traffic network in Stockholm using the VISSIM simulator.

ABSTRACT. This paper investigates a combinatorial optimization problem motived from a secure power network design application in [D\'{a}n and Sandberg 2010]. Two equivalent graph optimization formulations are derived. One of the formulations is a relaxed version of the connected dominating set problem, and hence the considered problem is referred to as relaxed connected dominating set (RCDS) problem. The RCDS problem is shown to be NP-hard, even for planar graphs. A mixed integer linear programming formulation is presented. In addition, for planar graphs a fixed parameter polynomial time solution methodology based on sphere-cut decomposition and dynamic programming is presented. The computation cost of the sphere-cut decomposition based approach grows linearly with problem instance size, provided that the branchwidth of the underlying graph is fixed and small. A case study with IEEE benchmark power networks verifies that small branchwidth are not uncommon in practice. The case study also indicates that the proposed methods show promise in computation efficiency.

ABSTRACT. The Research Concept Vehicle is a concept road vehicle developed at KTH. It is currently being used for research within the Autonomous Driving field, and it will be participating in the Grand Cooperative Driving Challenge 2016. In this article a brief overview of the control system implemented in the vehicle is given. We further explore a novel way of driving, in which the vehicle is able to move sideways without changing its orientation.

ABSTRACT. We explore the use of Model Predictive Control (MPC) techniques to solve vehicle lateral motion control problem on highway scenario i.e. maintaining lane at high speed. In particular, the problem of autonomously driving a vehicle along a desired path is formulated, where safety and performance constraints must be guaranteed for all possible road curvatures within a compact set. Safety constraints are translated into a maximum lateral deviation and orientation error w.r.t. a desired path, while performance requirements are formulated in terms of bounded lateral acceleration and velocity.

ABSTRACT. An approach to investigate the effect of system uncertainty on the optimal control configuration selection in multivariable systems is proposed. An optimal control configuration, i.e the configuration which best agrees with input-output pairing rules according to certain interaction measure (IM) can be obtained automatically by formulating the control configuration selection as a Transportation Problem (TP). The proposed approach then checks whether this optimal control configuration is valid for given system uncertainties or if a change in the configuration could be expected.

ABSTRACT. The aim of this article is to present the development of a Robotic Operating System (ROS) enabled full-scale Unmanned Ground Vehicle (UGV) for search and rescue missions. The proposed design is characterized by low cost, easiness of reconfiguration and simplicity with respect to similar platforms. As it will be presented, the underlying hardware and software developments aimed to provide multiple basic autonomous behaviors, such as velocity, position, and trajectory tracking. The efficacy of the developed framework has been experimentally evaluated in multiple real-life experiments where the full-scale UGV has been utilized to perform autonomous driving in open areas.

ABSTRACT. Kraftsystem i data- och telekommunikationsutrustning består av en kaskad av reglerade kraftomvandlare. Vid konstruktion av regulatorer i systemet fås då laster av konstant effekttyp. Systemet blir olinjärt och har hittills mestadels analyserats med mycket enkla modeller. I litteraturen har uppmärksammats att dessa modeller i många fall är alltför enkla för ett bra analysresultat. I artikeln föreslås en strukturerad metod att modellera dessa olinjära system. Metoden medger att det elektriska nätet enklare och systematiskt kan beskrivas med högre modellordning.

ABSTRACT. Many synthesis and analysis problems in classical control theory, such as output regularization, computation of invariant subspaces, model reduction, pole placement, etc., can be cast as matrix equations in terms of Sylvester and Lyapunov equations \cite{zhou1998essentials,castelan2005solution}. In this work we study large scale nonlinear systems and formulate the analysis of the system modes as Sylvester type equations. These are in turn solved using iterative methods that are preconditioned by utilizing an extension of the Sherman-Morrison-Woodbury (SMW) formula.

ABSTRACT. In this extended abstract, we summarize our recent developments in dual-arm robotic folding assembly. In particular, we improved the assembly controller and contact point estimation, with respect to previous work.

ABSTRACT. Human-robot interaction has been one of the most popular robotics research direction driven by both industrial and societal needs and can be further divided to cognitive and physical Human-Robot Interaction (pHRI). Current work in pHRI considers simple scenarios with known objects, prespecified grasping points (e.g. through handles) and hence the mapping of the human intention expressed through forces to the robotic force/torque sensor is known. Few researchers have addressed the stability of the overall system and the corresponding condition. In our work, we evaluate the passivity condition for damping controllers for physical human-robot interaction and derive a simple gain condition that can guarantee the passivity.

ABSTRACT. Star convex shapes are used for defining the boundary of objects which generate multiple measurements at a scan in extended target tracking applications. A hypersurface model is constructed from this model to represent the 2-d extent of objects by including a scaling variable. The object boundary can be defined as a function of the angle and can be represented either in frequency or time domain. Regardless of this choice, the resulting model involves non-linearities with respect to object's centroid and orientation. So far, these non-linearities are handled by using extended or unscented Kalman filters. In this study, we explore the potential gain in using Sequential Monte Carlo (SMC)methods for these models. More specifically, due to conditional linear Gaussian structure of relevant parameters, we suggest using a Rao-Blackwellised particle filter (RBPF) for inference.

ABSTRACT. In this paper, a nitrifying activated sludge process with ammonium feedback control is considered, where the aeration is adjusted based on the effluent ammonia concentration. For this control strategy, two complicating factors are present, these being the delayed measurement of the effluent ammonium concentration and the saturation of the control signal. A linear lag compensator is analysed for controlling the plant. As compared to conventional PI-control, this controller has a limited low-frequency gain, a fact that allows global stability to be established with the Popov criterion for a selected range of delays. The performance of the controller is assessed with simulations using a simplified version of the Activated Sludge Model No 1.

ABSTRACT. Proportional, Integral, and Derivative (PID) and discrete-time Proportional, Summation, and Difference (PSD) controllers are most widely used in industry due to their simplicity and performance characteristics. However, with these conventional fixed gain controllers we could have difficulties to handle nonlinear or time-variant characteristics. This problem led to a various gain-scheduled PID/PSD controller design techniques in both state-space and frequency domain. For the gain-scheduled PID/PSD controller design in the state-space, the bilinear/linear matrix inequality-, and the Bellman-Lyapunov function-based approaches, obviously lead to a non-convex performance and stability conditions with respect to scheduled parameter. To overcome the problem of non-convex design procedure, several approaches were developed based on convexifications, which introduced some conservativeness in the gain-scheduled PID/PSD controller design. In this paper, a novel methodology is proposed for robust discrete-time output feedback gain-scheduled PSD controller design for uncertain linear parameter-varying systems with hard input constraints, and with a realistic scenario, when the output is noise corrupted. The proposed design procedure guarantees the robust affine parameter-dependent quadratic stability and the parameter-varying guaranteed cost ($H_2$ performance) for a prescribed rate of change of scheduled parameters. Conditions to guarantee robust stability and performance requirements are translated directly to a convex optimization problem (without convexification) subject to linear matrix inequality constraints. A numerical example shows the benefit of the proposed method.

ABSTRACT. Today it is very common that military vehicles are equipped with multiple types of sensors,e.g., visual and long-wavelength infrared (LWIR) sensors. These can be used to detect, track and classify targets, as either friendly, neutral, or hostile. We show in simulation the effect of autonomous sensor control in a convoy protection scenario with multiple connected platforms. A solution to the sensor management problem is presented with focus on core features. Aspects on how different control policies can affect the mission goals are discussed.

ABSTRACT. This contribution consists of the identification and comparison of different models for a nonlinear system: the Cascaded Tanks system. The identification of this system is challenging due to the combination of soft and hard nonlinearities. Several linear/nonlinear black-/white- box model structures have been estimated and the results show that, while the best model for simulation performance is a nonlinear state-space model based on the physical modeling of the system, the best prediction error is achieved by a linear ARMAX model.

ABSTRACT. Estimation of the joint movements by using surface Electromyography (EMG) signals has great importance as a bio-inspired approach for the control of robotic limbs and prosthetics. However interpreting surface EMG measurements is challenging due to the nonlinearity and user dependency of the muscle dynamics. To solve this challenge we here propose to use an integrated EMG signal to identify the EMG-joint angle relation instead of using common EMG processing techniques. Then we estimate the joint angles for elbow flexion-extension movement by using an auto-regressive integrated moving average with exogenous input (ARIMAX) model. The experiments showed that the suggested approach results in a 21:85% average increase in the estimation performance of the elbow joint angle compared to the standard EMG processing and identification.

ABSTRACT. The interest for dynamic network estimation has increased during the last years with a wide variety of applications. In many cases, it is intractable or undesirable to measure all variables in a network and thus, to estimate the complete network dynamics. In this contribution, we will discuss an approach to estimate a part of the network dynamics when some of the necessary variables are unavailable. The approach relies on additional measurements that are dependent on the unavailable variables and thus indirectly contain information about them.

ABSTRACT. In this paper, the energy cost for asymptotic consensus control of linear systems is discussed and optimized. We consider a family of control design parametrized through Algebraic Riccarti Equation(ARE) and discuss the worst control energy cost for Q = 0 in the ARE and propose an optimization problem in terms of the edge weights. We show that fully-connected graph with equal edge weights is the global optimum among all graphs from the view of optimization. The dual problem also helps to tell which edges would not help to decrease the optimal control cost if we want to add some edges to the graph. In addition, we further discussed the case of given edge weights and optimize the parameter Q in the ARE. We provide a reasonable sub-optimal solution by solving an optimization problem.