| 1. |
- Russo, Alessio
(författare)
-
Efficient Exploration and Robustness in Controlled Dynamical Systems
- 2023
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- In this thesis, we explore two distinct topics. The first part of the thesis delves into efficient exploration in multi-task bandit models and model-free exploration in large Markov decision processes (MDPs). This section is driven by the recent research community's interest in uncovering optimal methods to navigate complex decision-making problems. In the second part, we examine attack vectors against MDPs and dynamical systems, which is motivated by the research community's recent push to enhance the safety of Machine Learning models against malicious attacks. Additionally, we explore how to quantify uncertainty in an off-policy evaluation problem, reflecting our ongoing efforts to deepen understanding in this domain. In the first part of the thesis, we present an investigation into the sample complexity of identifying the optimal arm in multi-task bandit problems. In this setting, an agent can select a task and efficiently learn through cross-task knowledge transfer. We derive instance-specific lower bounds that any probably approximately correct (PAC) algorithm should satisfy, revealing both theoretically and empirically significant improvements in scaling over previous methods. Subsequently, we investigate model-free exploration in Reinforcement Learning (RL) problems. By leveraging an information-theoretical viewpoint, we focus on the instance-specific lower bound for the number of samples needed to identify a nearly-optimal policy. We develop an approximation of this lower bound involving quantities that can be inferred using model-free approaches. This leads to a new model-free exploration strategy applicable to continuous MDPs. In the second part of the thesis, we begin by investigating two types of attacks on MDPs: those that alter the observations and those that manipulate the control signals of the victim. We present strategies for designing optimal attacks to minimize the collected reward of the victim. Our strategies show how uncertainties induced by the attack can be modeled using a partially observable MDP (POMDP) framework. We also illustrate how to devise attacks that achieve lower detectability, approaching the problem from a statistical detection perspective. Next, we explore the problem of an eavesdropper aiming to detect changes in an MDP. Approaching this problem from a statistical detection perspective, we introduce a novel definition of online privacy based on the average amount of information per observation of the underlying stochastic system. We derive privacy upper bounds and calculate policies that attain higher privacy levels, supplementing our analysis with examples and numerical simulations. Finally, we present a new off-policy estimation method based on Conformal Prediction that outputs an interval containing the target policy's true reward, demonstrating how to handle the distribution shift between target and behavior policies, and maintain the certainty level while reducing the interval length.Next, we shift our focus onto linear dynamical systems. We study poisoning attacks on data-driven control methods, focusing on how slight changes in the dataset induced by an adversary can significantly deteriorate control methods and potentially destabilize the system. We propose a novel algorithm for computing effective attacks, providing a theoretical basis for our strategy. We also study the detectability of poisoning attacks, focusing on the impact of data poisoning on least-squares estimates. We establish conditions under which the set of models compatible with the data includes the true model of the system, and we analyze different poisoning strategies for the attacker. On the basis of the arguments presented herein, we propose a stealthy data poisoning attack on the least-squares estimator that can evade classical statistical tests.
|
|
| 2. |
- Ofe, Hosea A., 1984-
(författare)
-
Orchestrating emerging digital ecosystems : investigating the establishment of an open data platform in the Swedish public transport industry
- 2020
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Digital platforms are affecting most contemporary organizations as they mediate an increasing range and number of interactions in their ecosystems. While the discourse on digital platform ecosystems has gained in interest over the years, it often revolves around dominant global firms and how they utilize their control over governance and architecture configurations to exercise power in shaping trajectories. This dissertation seeks to provide insights into the processes through which new digital platforms ecosystems are established by identifying challenges in orchestrating emerging digital ecosystems and approaches through which these can be navigated. To this end, my research focused on the establishment of an open data platform in the public transport industry in Sweden.My theoretical and empirical investigation provides three contributions to our understanding of orchestration of emerging digital ecosystems. The first contribution is the identification of key challenges in orchestrating an emerging ecosystem through a review of extant literature. The review suggests that challenges in orchestrating emerging ecosystems revolve around three goals: (1) attracting and generating network effects; (2) control and coordination; and (3) creating and capturing value. Thus, whether ecosystem establishment is successful or not depends in large part on how providers are able to address these challenges. The identification of challenges and remedies could be helpful for practitioners and scholars when assessing and diagnosing emerging ecosystems. However, I suggest that the different challenges and proposed solutions should not be treated as fixed and isolated guidelines in assessing ecosystems. Instead, providers should consider the challenges holistically in their ecosystem since there are interplays and interactions between their underlying socio-technical aspects. The second contribution is a conceptualization of the nature of orchestration in emerging digital ecosystems. I demonstrate that orchestration in an emerging ecosystem is inherently embroiled in a web of fragile power relationships among actors, unbounded participation, unbounded control, emergent outcomes, and persistent competing concerns. The third contribution of my thesis is the practical implications for how providers can approach orchestration and address challenges in emerging digital ecosystems. The fragile nature of emerging ecosystems suggests that orchestration is not limited to arm´s length measures but also stands to benefit from social interactions and relationship-building among actors with distinctive interests and understanding of their own rights.
|
|
| 3. |
- Ramesh, Chithrupa, 1982-
(författare)
-
State-based Channel Access for a Network of Control Systems
- 2014
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Wireless networked control systems use shared wireless links to communicate between sensors and controllers, and require a channel access policy to arbitrate access to the links. Existing multiple access protocols perform this role in an agnostic manner, by remaining insular to the applications that run over the network. This approach does not give satisfactory control performance guarantees. To enable the use of wireless networks in emerging industrial applications, we must be able to systematically design wireless networked control systems that provide guaranteed performances in resource-constrained networks.In this thesis, we advocate the use of state-based channel access policies. A state-based policy uses the state of the controlled plant to influence access to the network. The state contains information about not only the plant, but also the network, due to the feedback in the system. Thus, by using the state to decide when and how frequently to transmit, a control system can adapt its contribution to the network traffic, and enable the network to adapt access to the plant state. We show that such an approach can provide better performance than existing methods. We examine two different state-based approaches that are distributed and easy to implement on wireless devices: event-based scheduling and adaptive prioritization.Our first approach uses events to reduce the traffic in the network. We use a state-based scheduler in every plant sensor to generate non-coordinated channel access requests by selecting a few critical data packets, or events, for transmission. The network uses a contention resolution mechanism to deal with simultaneous channel access requests. We present three main contributions for this formulation. The first contribution is a structural analysis of stochastic event-based systems, where we identify a dual predictor architecture that results in separation in design of the state-based scheduler, observer and controller. The second contribution is a Markov model that describes the interactions in a network of event-based systems. The third contribution is an analysis of the stability of event-based systems, leading to a stabilizing design of event-based policies.Our second approach uses state-based priorities to determine access to the network. We use a dominance protocol to evaluate priorities in a contention-based setting, and characterize the resulting control performance. An implementation and evaluation of this channel access mechanism on sensor nodes is also presented.The thesis finally examines the general networked control problem of jointly optimizing measurement and control policies, when a nonlinear measurement policy is used to perform quantization, event-triggering or companding. This contribution focuses on some of the fundamental aspects of analyzing and synthesizing control systems with state-based measurement policies in a more generalized setting. We comment on the dual effect, certainty equivalence and separation properties for this problem. In particular, we show that it is optimal to apply separation and certainty equivalence to a design problem that permits a dynamic choice of the measurement and control policies.
|
|
| 4. |
- Alisic, Rijad
(författare)
-
Defense of Cyber-Physical Systems Against Learning-based Attackers
- 2023
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Cyberattacks against critical infrastructures pose a serious threat to society, as they can have devastating consequences on the economy, security, or public health. These infrastructures rely on a large network of cyber components, such as sensors, controllers, computers, and communication devices, to monitor and control their physical processes. An adversary can exploit the vulnerabilities in these cyber components to gain access to the system and manipulate its behavior or functionality.This thesis proposes methods that can be employed as a first line of defense against such attacks for Cyber-Physical Systems. In the first part of the thesis, we consider how uninformed attackers can learn to attack a Cyber-Physical System by eavesdropping through the cyber component. By learning to manipulate the plant, the attacker could figure out how to destroy the physical system before it is too late or completely take it over without raising any alarms. Stopping the attacker at the learning stage would force the attacker to act obliviously, increasing the chances of detecting them.We analyze how homomorphic encryption, a technique that allows computation on encrypted data, hinders an attacker's learning process and reduces its capabilities to attack the system. Specifically, we show that an attacker must solve challenging lattice problems to find attacks that are difficult to detect. Additionally, we show how the detection probability is affected by the attacker's solution to the problems and what parameters of the encryption scheme can be tweaked to increase the detection probability. We also develop a novel method that enables anomaly detection over homomorphically encrypted data without revealing the actual signals to the detector, thereby discouraging attackers from launching attacks on the detector. The detection can be performed using a hypothesis test. However, special care must be taken to ensure that fresh samples are used to detect changes from nominal behavior. We also explore how the adversary can try to evade detection using the same test and how the system can be designed to make detection easier for the defender and more challenging for the attacker.In the second part of the thesis, we study how information leakage about changes in the system depends on the system's dynamics. We use a mathematical tool called the Hammersley-Chapman-Robbins lower bound to measure how much information is leaked and how to minimize it. Specifically, we study how structured input sequences, which we call events, can be obtained through the output of a dynamical system and how this information can be hidden by adding noise or changing the inputs. The system’s speed and sensor locations affect how much information is leaked. We also consider balancing the system’s performance and privacy when using optimal control. Finally, we show how to estimate when the adversary’s knowledge of the event becomes accurate enough to launch an attack and how to change the system before that happens. These results are then used to aid the operator in detecting privacy vulnerabilities when designing a Cyber-Physical System, which increases the overall security when removed.
|
|
| 5. |
- Engström, Annica, 1967-
(författare)
-
Dimmornas bro : En berättelse om konstruktionen och iscensättandet av kliniska adjunkter
- 2012
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This dissertation is a study of a relatively new teacher function in nurse education, a teacher function that can be called clinical teacher, link lecturer, link teacher or lecturer involved in practice. The study departs from a constructivist perspective and the aim is to study the introduction, performance and regulation of this teacher function. The material consists of internationally and nationally published research 1978-2009, reports from the authorities, job ads for clinical teachers, job descriptions and contracts, archive material, evaluation studies, professional journals, texts on the Internet and 15 interviews with clinical teachers in Sweden. Using discourse analysis this material is studied in terms of interpretive repertoires, subject positions and ideological dilemmas. The primary focus of this dissertation concerns what problems the new teacher function is supposed to solve, how the problem should be solved, the effects of the chosen solutions and what´s staged, challenged and reproduced. Secondly, this dissertation highlights what categories of teachers are required and in demand for the new teacher function, how the teachers themselves manage their function and how the teacher function is maintained and challenged. The analysis identifies and highlights current interpretive repertoire, teachers possible subject positions and ideological dilemmas and how they are governed and challenged in the teaching function.
|
|
| 6. |
- Henriksson, Erik, 1982-
(författare)
-
Compensating for Unreliable Communication Links in Networked Control Systems
- 2009
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Control systems utilizing wireless sensor and actuator networks can be severely affectedby the properties of the communication links. Radio fading and interferencemay cause communication losses and outages in situations when the radio environmentis noisy and low transmission power is desirable. This thesis proposes amethod to compensate for such unpredictable losses of data in the feedback controlloop by introducing a predictive outage compensator (POC). The POC is a filter tobe implemented at the receiver sides of networked control systems where it generatesartificial samples when data are lost. If the receiver node does not receive thedata, the POC suggests a command based on the history of past data. It is shownhow to design, tune and implement a POC. Theoretical bounds and simulationresults show that a POC can improve the closed-loop control performance undercommunication losses considerably. We provide a deterministic and a stochasticmethod to synthesize POCs. Worst-case performance bounds are given that relatethe closed-loop performance with the complexity of the compensator. We also showthat it is possible to achieve good performance with a low-order implementationbased on Hankel norm approximation. Tradeoffs between achievable performance,communication loss length, and POC order are discussed. The results are illustratedon a simulated example of a multiple-tank process. The thesis is concludedby an experimental validation of wireless control of a physical lab process. Herethe controller and the physical system are separated geographically and interfacedthrough a wireless medium. For the remote control we use a hybrid model predictivecontroller. The results reflect the difficulties in wireless control as well as theyhighlight the flexibility and possibilities one obtains by using wireless instead of awired communication medium.
|
|
| 7. |
- Paridari, Kaveh, 1987-
(författare)
-
Hierarchical energy management in smart grids : Flexibility prediction, scheduling and resilient control
- 2019
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The electric power industry and society are facing challenges and opportunitiesof transforming the present power grid into a smart grid. Energymanagement systems (EMSs) play an important role in smart grids. A generalhierarchical structure for EMSs is considered here, which is composed ofa lower layer and an upper layer.The first research objective of the thesis is detailed modeling, schedulingand control of flexible loads at the lower layer of EMSs. To do this, a wellstudiedframework has been extended, which focuses on scheduling of staticloads and dynamic loads for home energy management systems (HEMSs).Then, a robust formulation of the framework is proposed, which takes theuser behavior uncertainty into account so that the cost of optimal schedulingof appliances is less sensitive to unpredictable changes in user preferences.Considering that the optimization algorithms in the proposed framework canbe computationally intensive, an efficient plug-and-play policy is proposedand validated through several simulation studies.The second research objective is to predict, plan, and control the aggregatedflexible load at the upper layer. Here, an iterative distributed approachamong aggregator and HEMSs is designed, to maximize the aggregated profitmade out of the shared energy storage system, while technical and operationalconstraints are satisfied. In addition, a strategy is proposed for flexibilityprediction of aggregated heterogeneous thermostatically controlled loads ina single micro-community of households. Then, algorithms are designed forplanning and control of aggregated flexibility in several micro-communities,to be used for bidding in energy and reserve markets.To meet these research objectives, the control systems in the hierarchicalEMSs are connected over IT infrastructures and are in interaction with endusers.While this is done to achieve economical and environmental goals,it also introduces new sources of uncertainty in the control loops. Thus,the third research objective is to design policies to make the EMSs resilientagainst uncertainties and cyber attacks. Here, the user behavior uncertaintyhas been modeled, and a robust formulation is designed so that the optimalsolution for scheduling of appliances is more resilient to the uncertainties. Inaddition, fault-tolerant control techniques have been applied to a hierarchicalEMS to mitigate cyber-physical attacks, with no need for major re-designof the local control loops in already existing EMSs. Moreover, stability andoptimal performance of the proposed attack-resilient control policy have been proven.
|
|
| 8. |
- Ramesh, Chithrupa, 1982-
(författare)
-
Contention-based Multiple Access Architectures for Networked Control Systems
- 2011
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Networked Control Systems (NCSs) use a wireless network for communication between sensors and controllers, and require a Medium Access Controller (MAC) to arbitrate access to the shared medium. Traditionally, a MAC for control systems is chosen primarily based on the delay it introduces in the closed loop. This thesis focuses on the design of a contention-based MAC, in a time-varying, resource-constrained network for closed loop systems. In this thesis, we advocate the use of a state-aware MAC, as opposed to an agnostic MAC, for NCSs. A state-aware MAC uses the state of the plant to influence access to the network. The state-aware policy is realized using two different approaches in the MAC: a regulatory formulation and an adaptive prioritization. Our first approach is a regulatory MAC, which serves to reduce the traffic in the network. We use a local state-based scheduler to select a few critical data packets to send to the MAC. We analyze the impact of such a scheduler on the closed loop system, and show that there is a dual effect for the control signal, which makes determining the optimal controller difficult. We also identify restrictions on the scheduling criterion that result in a separation of the scheduler, observer and controller designs. Our second approach is a prioritized MAC that uses state-based priorities called Attentions, to determine access to the network. We use a dominance protocol called tournaments, to evaluate priorities in a contention-based setting, and analyze the resulting performance of the MAC. We also consider a NCS that uses a wireless multihop mesh network for communication between the controller and actuator. We design an optimal controller, which uses packet delivery predictions from a recursive Bayesian network estimator.
|
|
| 9. |
- Stefansson, Elis
(författare)
-
Complexity-aware Decision-making with Applications to Large-scale and Human-in-the-loop Systems
- 2023
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This thesis considers control systems governed by autonomous decision-makers and humans. We formalise and compute low-complex control policies with applications to large-scale systems, and propose human interaction models for controllers to compute interaction-aware decisions.In the first part of the thesis, we consider complexity-aware decision-making, formalising the complexity of control policies and constructing algorithms that compute low-complexity control policies. More precisely, first, we consider large-scale control systems given by hierarchical finite state machines (HFSMs) and present a planning algorithm for such systems that exploits the hierarchy to compute optimal policies efficiently. The algorithm can also handle changes in the system with ease. We prove these properties and conduct simulations on HFSMs with up to 2 million states, including a robot application, where our algorithm outperforms both Dijkstra's algorithm and Contraction Hierarchies. Second, we present a planning objective for control systems modelled as finite state machines yielding an explicit trade-off between a policy's performance and complexity. We consider Kolmogorov complexity since it captures the ultimate compression of an object on a universal Turing machine. We prove that this trade-off is hard to optimise in the sense that dynamic programming is infeasible. Nonetheless, we present two heuristic algorithms obtaining low-complexity policies and evaluate the algorithms on a simple navigation task for a mobile robot, where we obtain low-complexity policies that concur with intuition. In the second part of the thesis, we consider human-in-the-loop systems and predict human decision-making in such systems. First, we look at how the interaction between a robot and a human in a control system can be predicted using game theory, focusing on an autonomous truck platoon interacting with a human-driven car. The interaction is modelled as a hierarchical dynamic game, where the hierarchical decomposition is temporal with a high-fidelity tactical horizon predicting immediate interactions and a low-fidelity strategic horizon estimating long-term behaviour. The game enables feasible computations validated through simulations yielding situation-aware behaviour with natural and safe interactions. Second, we seek models to explain human decision-making, focusing on driver overtaking scenarios. The overtaking problem is formalised as a decision problem with perceptual uncertainty. We propose and numerically analyse risk-agnostic and risk-aware decision models, judging if an overtaking is desirable. We show how a driver's decision time and confidence level can be characterised through two model parameters, which collectively represent human risk-taking behaviour. We detail an experimental testbed for evaluating the decision-making process in the overtaking scenario and present some preliminary experimental results from two human drivers.
|
|
| 10. |
- Teixeira, André
(författare)
-
Toward Cyber-Secure and Resilient Networked Control Systems
- 2014
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Resilience is the ability to maintain acceptable levels of operation in the presence of abnormal conditions. It is an essential property in industrial control systems, which are the backbone of several critical infrastructures. The trend towards using pervasive information technology systems, such as the Internet, results in control systems becoming increasingly vulnerable to cyber threats. Traditional cyber security does not consider the interdependencies between the physical components and the cyber systems. On the other hand, control-theoretic approaches typically deal with independent disturbances and faults, thus they are not tailored to handle cyber threats. Theory and tools to analyze and build control system resilience are, therefore, lacking and in need to be developed. This thesis contributes towards a framework for analyzing and building resilient control systems.First, a conceptual model for networked control systems with malicious adversaries is introduced. In this model, the adversary aims at disrupting the system behavior while remaining undetected by an anomaly detector The adversary is constrained in terms of the available model knowledge, disclosure resources, and disruption capabilities. These resources may correspond to the anomaly detector’s algorithm, sniffers of private data, and spoofers of control commands, respectively.Second, we address security and resilience under the perspective of risk management, where the notion of risk is defined in terms of a threat’s scenario, impact, and likelihood. Quantitative tools to analyze risk are proposed. They take into account both the likelihood and impact of threats. Attack scenarios with high impact are identified using the proposed tools, e.g., zero-dynamics attacks are analyzed in detail. The problem of revealing attacks is also addressed. Their stealthiness is characterized, and how to detect them by modifying the system’s structure is also described.As our third contribution, we propose distributed fault detection and isolation schemes to detect physical and cyber threats on interconnected second-order linear systems. A distributed scheme based on unknown input observers is designed to jointly detect and isolate threats that may occur on the network edges or nodes. Additionally, we propose a distributed scheme based on local models and measurements that is resilient to changes outside the local subsystem. The complexity of the proposed methods is decreased by reducing the number of monitoring nodes and by characterizing the minimum amount of model information and measurements needed to achieve fault detection and isolation.Finally, we tackle the problem of distributed reconfiguration under sensor and actuator faults. In particular, we consider a control system with redundant sensors and actuators cooperating to recover from the removal of individual nodes. The proposed scheme minimizes a quadratic cost while satisfying a model-matching condition, which maintains the nominal closed-loop behavior after faults. Stability of the closed-loop system under the proposed scheme is analyzed.
|
|
