Date of Award

5-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Automotive Engineering

Committee Member

Dr. Pierluigi Pisu, Committee Chair

Committee Member

Dr. Richard Brooks, Co-chair

Committee Member

Dr. Beshah Ayalew

Committee Member

Dr. Yongqiang Wang

Abstract

The objective of this dissertation is to develop a resilient control approach to secure Cyber Physical Systems (CPS) against cyber-attacks, network failures and potential physical faults. Despite being potentially beneficial in several aspects, the connectivity in CPSs poses a set of specific challenges from safety and reliability standpoint. The first challenge arises from unreliable communication network which affects the control/management of overall system. Second, faulty sensors and actuators can degrade the performance of CPS and send wrong information to the controller or other subsystems of the CPS. Finally, CPSs are vulnerable to cyber-attacks which can potentially lead to dangerous scenarios by affecting the information transmitted among various components of CPSs. Hence, a resilient control approach is proposed to address these challenges. The control approach consists of three main parts:(1) Physical fault diagnostics: This part makes sure the CPS works normally while there is no cyber-attacks/ network failure in the communication network; (2) Cyber-attack/failure resilient strategy: This part consists of a resilient strategy for specific cyber-attacks to compensate for their malicious effects ; (3) Decision making algorithm: The decision making block identifies the specific existing cyber-attacks/ network failure in the system and deploys corresponding control strategy to minimize the effect of abnormality in the system performance. In this dissertation, we consider a platoon of connected vehicle system under Co-operative Adaptive Cruise Control (CACC) strategy as a CPS and develop a resilient control approach to address the aforementioned challenges. The first part of this dissertation investigates fault diagnostics of connected vehicles assuming ideal communication network. Very few works address the real-time diagnostics problem in connected vehicles. This study models the effect of different faults in sensors and actuators, and also develops fault diagnosis scheme for detectable and identifiable faults. The proposed diagnostics scheme is based on sliding model observers to detect, isolate and estimate faults in the sensors and actuators. One of the main advantages of sliding model approach lies in applicability to nonlinear systems. Therefore, the proposed method can be extended for other nonlinear cyber physical systems as well. The second part of the proposed research deals with developing strategies to maintain performance of cyber-physical systems close to the normal, in the presence of common cyber-attacks and network failures. Specifically, the behavior of Dedicated Short-Range Communication (DSRC) network is analyzed under cyber-attacks and failures including packet dropping, Denial of Service (DOS) attack and false data injection attack. To start with, packet dropping in network communication is modeled by Bernoulli random variable. Then an observer based modifying algorithm is proposed to modify the existing CACC strategy against the effect of packet dropping phenomena. In contrast to the existing works on state estimation over imperfect communication network in CPS which mainly use either holding previous received data or Kalman filter with intermittent observation, a combination of these two approaches is used to construct the missing data over packet dropping phenomena. Furthermore, an observer based fault diagnostics based on sliding mode approach is proposed to detect, isolate and estimate sensor faults in connected vehicles platoon. Next, Denial of Service (DoS) attack is considered on the communication network. The effect of DoS attack is modeled as an unknown stochastic delay in data delivery in the communication network. Then an observer based approach is proposed to estimate the real data from the delayed measured data over the network. A novel approach based on LMI theory is presented to design observer and estimate the states of the system via delayed measurements. Next, we explore and alternative approach by modeling DoS with unknown constant time delay and propose an adaptive observer to estimate the delay. Furthermore, we study the effects of system uncertainties on the DoS algorithm. In the third algorithm, we considered a general CPS with a saturated DoS attack modeled with constant unknown delay. In this part, we modeled the DoS via a PDE and developed a PDE based observer to estimate the delay as well as states of the system while the only available measurements are delayed. Furthermore, as the last cyber-attack of the second part of the dissertation, we consider false data injection attack as the fake vehicle identity in the platoon of vehicles. In this part, we develop a novel PDE-based modeling strategy for the platoon of vehicles equipped with CACC. Moreover, we propose a PDE based observer to detect and isolate the location of the false data injection attack injected into the platoon as fake identity. Finally, the third part of the dissertation deals with the ongoing works on an optimum decision making strategy formulated via Model Predictive Control (MPC). The decision making block is developed to choose the optimum strategy among available strategies designed in the second part of the dissertation.

Share

COinS