arXiv subjects

Rong Su

Publications and source records attributed to Rong Su.

At least 55 records · Page 3Linked to original sources

arXiv2022

Distributed Eco-Driving Algorithm of Vehicle Platoon Using Traffic Light and Road Slope Information

This paper investigates the problem of ecological driving (eco-driving) of vehicle platoons. To reduce the probability of the platoon avoiding red lights and increase fuel efficiency, a two-layer control architecture is proposed. The first layer is in charge of optimizing the leader's long-term motion profile using the traffic light and road slope information. The long-term planning model is defined based on the reachability analysis of the platoon to the green light windows. An event-triggered mechanism is proposed to operate the long-term planning model. The second layer is the short-term adaptation, in which the leader attempts to follow the planning motion profile in real time, while the follower keeps track of the nearest preceding vehicle and the leader, to preserve the desired inter-vehicular distances. A Newton's method-based algorithm is implemented to effectively solve both the long-term planning and short-term adaptation problems. The effectiveness of the proposed formulation and algorithm is illustrated by the simulations.

eess.SY↗

arXiv2022

Identification of System Vulnerability under a Smart Sensor Attack via Attack Model Reduction

In this work, we investigate how to make use of model reduction techniques to identify the vulnerability of a closed-loop system, consisting of a plant and a supervisor, that might invite attacks. Here, the system vulnerability refers to the existence of key observation sequences that could be exploited by a specific smart sensor attack to cause damage infliction. We consider a nondeterministic smart attack, i.e., there might exist more than one attack choice over each received observation, and adopt our previously proposed modeling framework, where such an attack is captured by a standard finite-state automaton. For a given supervisor S and a smart sensor attack model A, another smart attack model A' is called attack equivalent to A with respect to S, if the resulting compromised supervisor, defined as the composition of the supervisor S and attack model A', is control equivalent to the original compromised supervisor, defined as the composition of S and A. Following the spirit of supervisor reduction that relies on the concept of control congruence, we will show that, this problem of synthesizing a reduced smart attack model A' that is attack equivalent to A with respect to S, can be transformed to a classical supervisor reduction problem, making all existing synthesis tools available for supervisor reduction directly applicable to our problem. A simplified and ideally minimum-state attack model can reveal all necessary observation sequences for the attacker to be successful, thus, reminds system designers to take necessary precautions in advance, which may improve system resilience significantly. An example is presented to show the effectiveness of our proposed attack model reduction technique to identify the system vulnerability.

eess.SY↗

arXiv2021

A Topological Approach for Computing Supremal Sublanguages for Some Language Equations in Supervisory Control Theory

In this paper, we shall present a topological approach for the computation of some supremal sublanguages, often specified by language equations, which arise from the study of the supervisory control theory. The basic idea is to identify the solutions of the language equations as open sets for some (semi)-topologies. Then, the supremal sublanguages naturally correspond to the supremal open subsets, i.e., the interiors. This provides an elementary and uniform approach for computing various supremal sublanguages encountered in the supervisory control theory and is closely related to a theory of approximation, known as the rough set theory, in artificial intelligence.

cs.FL↗

arXiv2021

Retrieval & Interaction Machine for Tabular Data Prediction

Prediction over tabular data is an essential task in many data science applications such as recommender systems, online advertising, medical treatment, etc. Tabular data is structured into rows and columns, with each row as a data sample and each column as a feature attribute. Both the columns and rows of the tabular data carry useful patterns that could improve the model prediction performance. However, most existing models focus on the cross-column patterns yet overlook the cross-row patterns as they deal with single samples independently. In this work, we propose a general learning framework named Retrieval & Interaction Machine (RIM) that fully exploits both cross-row and cross-column patterns among tabular data. Specifically, RIM first leverages search engine techniques to efficiently retrieve useful rows of the table to assist the label prediction of the target row, then uses feature interaction networks to capture the cross-column patterns among the target row and the retrieved rows so as to make the final label prediction. We conduct extensive experiments on 11 datasets of three important tasks, i.e., CTR prediction (classification), top-n recommendation (ranking) and rating prediction (regression). Experimental results show that RIM achieves significant improvements over the state-of-the-art and various baselines, demonstrating the superiority and efficacy of RIM.

cs.IR↗

arXiv2021

Observation-Assisted Heuristic Synthesis of Covert Attackers Against Unknown Supervisors

In this work, we address the problem of synthesis of covert attackers in the setup where the model of the plant is available, but the model of the supervisor is unknown, to the adversary. To compensate the lack of knowledge on the supervisor, we assume that the adversary has recorded a (prefix-closed) finite set of observations of the runs of the closed-loop system, which can be used for assisting the synthesis. We present a heuristic algorithm for the synthesis of covert damage-reachable attackers, based on the model of the plant and the (finite) set of observations, by a transformation into solving an instance of the partial-observation supervisor synthesis problem. The heuristic algorithm developed in this paper may allow the adversary to synthesize covert attackers without having to know the model of the supervisor, which could be hard to obtain in practice. For simplicity, we shall only consider covert attackers that are able to carry out sensor replacement attacks and actuator disablement attacks. The effectiveness of our approach is illustrated on a water tank example adapted from the literature.

eess.SY↗

arXiv2021

Synthesis of Maximally Permissive Covert Attackers Against Unknown Supervisors by Using Observations

In this paper, we consider the problem of synthesis of maximally permissive covert damage-reachable attackers in the setup where the model of the supervisor is unknown to the adversary but the adversary has recorded a (prefix-closed) finite set of observations of the runs of the closed-loop system. The synthesized attacker needs to ensure both the damage-reachability and the covertness against all the supervisors which are consistent with the given set of observations. There is a gap between the de facto maximal permissiveness, assuming the model of the supervisor is known, and the maximal permissiveness that can be attained with a limited knowledge of the model of the supervisor, from the adversary's point of view. We consider the setup where the attacker can exercise sensor replacement/deletion attacks and actuator enablement/disablement attacks. The solution methodology proposed in this work is to reduce the synthesis of maximally permissive covert damage-reachable attackers, given the model of the plant and the finite set of observations, to the synthesis of maximally permissive safe supervisors for certain transformed plant, which shows the decidability of the observation-assisted covert attacker synthesis problem. The effectiveness of our approach is illustrated on a water tank example adapted from the literature.

eess.SY↗

arXiv2021

AutoFT: Automatic Fine-Tune for Parameters Transfer Learning in Click-Through Rate Prediction

Recommender systems are often asked to serve multiple recommendation scenarios or domains. Fine-tuning a pre-trained CTR model from source domains and adapting it to a target domain allows knowledge transferring. However, optimizing all the parameters of the pre-trained network may result in over-fitting if the target dataset is small and the number of parameters is large. This leads us to think of directly reusing parameters in the pre-trained model which represent more general features learned from multiple domains. However, the design of freezing or fine-tuning layers of parameters requires much manual effort since the decision highly depends on the pre-trained model and target instances. In this work, we propose an end-to-end transfer learning framework, called Automatic Fine-Tuning (AutoFT), for CTR prediction. AutoFT consists of a field-wise transfer policy and a layer-wise transfer policy. The field-wise transfer policy decides how the pre-trained embedding representations are frozen or fine-tuned based on the given instance from the target domain. The layer-wise transfer policy decides how the high?order feature representations are transferred layer by layer. Extensive experiments on two public benchmark datasets and one private industrial dataset demonstrate that AutoFT can significantly improve the performance of CTR prediction compared with state-of-the-art transferring approaches.

cs.IR↗

arXiv2021

Dual Graph enhanced Embedding Neural Network for CTR Prediction

CTR prediction, which aims to estimate the probability that a user will click an item, plays a crucial role in online advertising and recommender system. Feature interaction modeling based and user interest mining based methods are the two kinds of most popular techniques that have been extensively explored for many years and have made great progress for CTR prediction. However, (1) feature interaction based methods which rely heavily on the co-occurrence of different features, may suffer from the feature sparsity problem (i.e., many features appear few times); (2) user interest mining based methods which need rich user behaviors to obtain user's diverse interests, are easy to encounter the behavior sparsity problem (i.e., many users have very short behavior sequences). To solve these problems, we propose a novel module named Dual Graph enhanced Embedding, which is compatible with various CTR prediction models to alleviate these two problems. We further propose a Dual Graph enhanced Embedding Neural Network (DG-ENN) for CTR prediction. Dual Graph enhanced Embedding exploits the strengths of graph representation with two carefully designed learning strategies (divide-and-conquer, curriculum-learning-inspired organized learning) to refine the embedding. We conduct comprehensive experiments on three real-world industrial datasets. The experimental results show that our proposed DG-ENN significantly outperforms state-of-the-art CTR prediction models. Moreover, when applying to state-of-the-art CTR prediction models, Dual graph enhanced embedding always obtains better performance. Further case studies prove that our proposed dual graph enhanced embedding could alleviate the feature sparsity and behavior sparsity problems. Our framework will be open-source based on MindSpore in the near future.

cs.IR↗

arXiv2021

Privacy-Preserving Co-synthesis Against Sensor-Actuator Eavesdropping Intruder

In this work, we investigate the problem of privacy-preserving supervisory control against an external passive intruder via co-synthesis of dynamic mask, edit function, and supervisor for opacity enforcement and requirement satisfaction. We attempt to achieve the following goals: 1) the system secret cannot be inferred by the intruder, i.e., opacity of secrets against the intruder, and the existence of the dynamic mask and the edit function should not be discovered by the intruder, i.e., covertness of dynamic mask and edit function against the intruder; 2) the closed-loop system behaviors should satisfy some safety and nonblockingness requirement. We assume the intruder can eavesdrop both the sensing information generated by the sensors and the control commands issued to the actuators, and we refer to such an intruder as a sensor-actuator eavesdropping intruder. Our approach is to model the co-synthesis problem as a distributed supervisor synthesis problem in the Ramadge-Wonham supervisory control framework, and we propose an incremental synthesis heuristic to incrementally synthesize a dynamic mask, an edit function, and a supervisor, which consists of three steps: 1) we first synthesize an ensemble ME of dynamic mask and edit function to ensure the opacity and the covertness against a sensor eavesdropping but command non-eavesdropping intruder, and marker-reachability; 2) we then decompose ME into a dynamic mask and an edit function by using a constraint-based approach, with the help of a Boolean satisfiability (SAT) solver; 3) finally, we synthesize a supervisor such that opacity and covertness can be ensured against the sensor-actuator eavesdropping intruder, and at the same time safety and nonblockingness requirement can be ensured. The effectiveness of our approach is illustrated on an example about the enforcement of location privacy for an autonomous vehicle.

eess.SY↗

arXiv2021

Privacy-Preserving Supervisory Control of Discrete-Event Systems via Co-Synthesis of Edit Function and Supervisor for Opacity Enforcement and Requirement Satisfaction

This paper investigates the problem of co-synthesis of edit function and supervisor for opacity enforcement in the supervisory control of discrete-event systems (DES), assuming the presence of an external (passive) intruder, where the following goals need to be achieved: 1) the external intruder should never infer the system secret, i.e., the system is opaque, and never be sure about the existence of the edit function, i.e., the edit function remains covert; 2) the controlled plant behaviors should satisfy some safety and nonblockingness requirements, in the presence of the edit function. We focus on the class of edit functions that satisfy the following properties: 1) the observation capability of the edit function in general can be different from those of the supervisor and the intruder; 2) the edit function can implement insertion, deletion, and replacement operations; 3) the edit function performs bounded edit operations, i.e., the length of each string output of the edit function is upper bounded by a given constant. We propose an approach to solve this co-synthesis problem by modeling it as a distributed supervisor synthesis problem in the Ramadge-Wonham supervisory control framework. By taking the special structure of this distributed supervisor synthesis problem into consideration and to improve the possibility of finding a non-empty distributed supervisor, we propose two novel synthesis heuristics that incrementally synthesize the supervisor and the edit function. The effectiveness of our approach is illustrated on an example in the enforcement of the location privacy.

eess.SY↗

arXiv2021

Networked Supervisor Synthesis Against Lossy Channels with Bounded Network Delays as Non-Networked Synthesis

In this work, we study the problem of supervisory control of networked discrete event systems. We consider lossy communication channels with bounded network delays, for both the control channel and the observation channel. By a model transformation, we transform the networked supervisor synthesis problem into the classical (non-networked) supervisor synthesis problem (for non-deterministic plants), such that the existing supervisor synthesis tools can be used for synthesizing networked supervisors. In particular, we can use the (state-based) normality property for the synthesis of the supremal networked supervisors, whose existence is guaranteed by construction due to our consideration of command non-deterministic supervisors. The effectiveness of our approach is illustrated on a mini-guideway example that is adapted from the literature, for which the supremal networked supervisor has been synthesized in the synthesis tools SuSyNA and TCT.

eess.SY↗

arXiv2021

Bounded Synthesis of Resilient Supervisors

In this paper, we investigate the problem of synthesizing resilient supervisors against combined actuator and sensor attacks, for the subclass of cyber-physical systems that can be modelled as discrete-event systems. We assume that the attackers can carry out actuator enablement and disablement attacks as well as sensor replacement attacks. We consider both risky attackers and covert attackers in the setup where the (partial-observation) attackers may or may not eavesdrop the control commands (issued by the supervisor). A constraint-based approach for the bounded synthesis of resilient supervisors is developed, by reducing the problem to the Quantified Boolean Formulas (QBF) problem. The bounded synthesis problem can then be solved either with a QBF solver or with repeated calls to a propositional satisfiability (SAT) solver, by employing maximally permissive attackers, which can be synthesized with the existing partial-observation supervisor synthesis procedures, as counter examples in the counter example guided inductive synthesis loop.

eess.SY↗

arXiv2021

Synthesis of Covert Actuator Attackers for Free

In this paper, we shall formulate and address a problem of covert actuator attacker synthesis for cyber-physical systems that are modelled by discrete-event systems. We assume the actuator attacker partially observes the execution of the closed-loop system and is able to modify each control command issued by the supervisor on a specified attackable subset of controllable events. We provide straightforward but in general exponential-time reductions, due to the use of subset construction procedure, from the covert actuator attacker synthesis problems to the Ramadge-Wonham supervisor synthesis problems. It then follows that it is possible to use the many techniques and tools already developed for solving the supervisor synthesis problem to solve the covert actuator attacker synthesis problem for free. In particular, we show that, if the attacker cannot attack unobservable events to the supervisor, then the reductions can be carried out in polynomial time. We also provide a brief discussion on some other conditions under which the exponential blowup in state size can be avoided. Finally, we show how the reduction based synthesis procedure can be extended for the synthesis of successful covert actuator attackers that also eavesdrop the control commands issued by the supervisor.

eess.SY↗

arXiv2021

Synthesis of Covert Sensor Attacks in Networked Discrete-Event Systems with Non-FIFO Channels

In this paper, we investigate the covert sensor attack synthesis problem in the framework of supervisory control of networked discrete-event systems (DES), where the observation channel and the control channel are assumed to be non-FIFO and have bounded network delays. We focus on the class of sensor attacks satisfying the following properties: 1) the attacker might not have the same observation capability as the networked supervisor; 2) the attacker aims to remain covert, i.e., hide its presence against the networked monitor; 3) the attacker could insert, delete, or replace compromised observable events; 4) it performs bounded sensor attacks, i.e., the length of each string output of the sensor attacker is upper bounded by a given constant. The solution methodology proposed in this work is to solve the covert sensor attack synthesis problem for networked DES by modeling it as the well studied Ramadge-Wonham supervisor synthesis problem, and the constructions work for both the damage-reachable attacks and the damage-nonblocking attacks. In particular, we show the supremal covert sensor attack exists in the networked setup and can be effectively computed by using the normality property based synthesis approach.

eess.SY↗

arXiv2020

Overview of Networked Supervisory Control with Imperfect Communication Channels

This paper presents an overview of the networked supervisory control framework for discrete event systems with imperfect communication networks, which can be divided into the centralized supervisory control setup and the decentralized supervisory control setup. We review the state-of-art networked control frameworks with observation channel delays and control channel delays, for untimed and timed models. Data losses in communication channels are also considered. The review of the state-of-art networked control frameworks will be focused on the following parts: 1) the construction of the networked control closed-loop system 2) the condition to ensure the existence of a networked supervisor 3) the synthesis procedure for networked-delay resilient supervisor 4) the possibility of improving the synthesis efficiency.

eess.SY↗

arXiv2020

Correlative infrared optical coherence tomography and hyperspectral chemical imaging

Optical coherence tomography (OCT) is a high-resolution three-dimensional imaging technique that enables non-destructive measurements of surface and subsurface microstructures. Recent developments of OCT operating in the mid-infrared (MIR) range (around 4 μm) lifted fundamental scattering limitations and initiated applied material research in formerly inaccessible fields. The MIR spectral region, however, is also of great interest for spectroscopy and hyperspectral imaging, which allow highly selective and sensitive chemical studies of materials. In this contribution, we introduce an OCT system (dual-band, central wavelengths of 2 μm m and 4 μm) combined with MIR spectroscopy that is implemented as a raster scanning chemical imaging modality. The fully-integrated and cost-effective optical instrument is based on a single supercontinuum laser source (emission spectrum spanning from 1.1 μm to 4.4 μm). Capabilities of the in-situ correlative measurements are experimentally demonstrated by obtaining complex multidimensional material data, comprising morphological and chemical information, from a multi-layered composite ceramic-polymer specimen.

physics.optics↗

arXiv2019

A Novel Fully Informed Water Cycle Algorithm for Solving Optimal Power Flow Problems in Electric Grids

Optimal power flow (OPF) is a key tool for planning and operations in energy grids. The line-flow constraints, generator loading effect, piece-wise cost functions, emission, and voltage quality cost make the optimization model non-convex and computationally cumbersome to solve. Metaheuristic techniques for solving the problem have emerged as a promising solution to solve the complex OPF problem. Recently, the water cycle algorithm (WCA), a method inspired by the observation of the water cycle process and the surface run-off model was proposed for solving optimization problems. This paper proposes an improved version of WCA that uses the concept of sharing global and local information among individuals to improve the exploitation ability compared with the standard WCA. The so called fully informed WCA (FIWCA) is tested against standard WCA and other metaheuristic techniques studied in the literature on IEEE 30 and 57 bus systems for various scenarios. Comparison and discussion regarding the performance and reliability of the metaheuristics approaches studied in literature are discussed. The obtained optimization results show that the better performance of proposed FIWCA comparing with the WCA and other algorithms especially in term of stability performance over replications. Consequently, it emerges as a tool for solving OPF in a reliable and efficient manner.

math.OC↗

arXiv2019

A Method for Quickly Bounding the Optimal Objective Value of an OPF Problem using a Semidefinite Relaxation and a Local Solution

Optimal power flow (OPF) is an important problem in the operation of electric power systems. Due to the OPF problem's non-convexity, there may exist multiple local optima. Certifiably obtaining the global solution is important for certain applications of OPF problems. Many global optimization techniques compute an optimality gap that compares the achievable objective value corresponding to the feasible point from a local solution algorithm with the objective value bound from a convex relaxation technique. Rather than the traditional practice of completely separating the local solution and convex relaxation computations, this paper proposes a method that exploits information from a local solution to speed the computation of an objective value bound using a semidefinite programming (SDP) relaxation. The improvement in computational tractability comes with the trade-off of reduced tightness for the resulting objective value bound. Numerical experiments illustrate this trade-off, with the proposed method being faster but weaker than the SDP relaxation and slower but tighter than second-order cone programming (SOCP) and quadratic convex (QC) relaxations for many large test cases.

math.OC↗