arXiv subjects

Ali Eslami

Publications and source records attributed to Ali Eslami.

At least 19 recordsLinked to original sources

arXiv2026

Resilient Control of Switched Vehicle Platoons under False Data Injection Attacks

This paper investigates resilient control design for leader--follower vehicle platoons with mode-dependent powertrain dynamics subject to False Data Injection (FDI) attacks on vehicle-to-vehicle (V2V) communication channels. The longitudinal motion of each vehicle is described by a switched third-order model that captures changes in the powertrain dynamics across different operating modes. To estimate the attack signals that are injected into the communication channels, each vehicle is equipped with an auxiliary system, and a dedicated observer is implemented for each communication link. The resulting attack estimates are then used to mitigate the effects of the attacks through our proposed resilient controller. For attacks with bounded rates but potentially unbounded amplitudes, the closed-loop platoon is shown to be uniformly ultimately bounded. For a predecessor-following topology, string stability is established for the nominal switched platoon, while the effect of nonzero attack-estimation errors on acceleration propagation is shown to be bounded. Numerical case studies demonstrate the effectiveness of the proposed approach.

eess.SY↗

arXiv2026

When Agents Meet Electric Bus Fleet Operations: Pricing Behavior, Trade-offs, and Policy Implications in an Aggregator Framework

Agentic systems are changing how complex operational tasks are coordinated, introducing a new paradigm for connecting heterogeneous data sources and automating processes. Electric bus fleets provide a relevant test case. Their operation requires continuous coordination between service reliability, battery state-of-charge, charger availability, electricity prices, route-energy uncertainty, and vehicle-to-grid (V2G) opportunities. This paper proposes an agentic aggregator framework that streamlines this decision environment by coupling an optimization-based electric bus scheduling model with supervisory agents for disturbance detection, tariff adaptation, and schedule evaluation. The optimization core enforces physical feasibility across routes, chargers, batteries, and V2G exchanges, while the agentic layer interprets changing operating conditions, triggers real-time re-optimization when needed, and defines how flexibility value is allocated between the aggregator and the public transport operator (PTO). A realistic depot case study evaluates day-ahead and real-time operations under profit-based and operation-based coordination modes, considering service delays, route-energy deviations, electricity price shocks, and combined disturbances. The results show that agentic aggregation can support adaptive fleet-grid coordination by maintaining feasible schedules, activating re-optimization selectively, and improving the use of charging and V2G flexibility. However, they also reveal a critical trade-off: the same agentic capability that reduces operational complexity can extract value from the PTO when configured around profit-oriented pricing. These findings suggest that agentic aggregators can become useful for managing electric bus V2G operations, but their deployment in public-fleet contexts requires transparent coordination modes, auditable tariff-setting, and explicit value-sharing rules.

cs.AI↗

arXiv2026

Stable but Unsafe: Agent-Driven Cyber-Physical Systems Under Gain Manipulation Attacks

AI agents are increasingly being connected to Cyber-Physical Systems (CPS) to generate or modify control-relevant parameters at runtime, including feedback gains, cost weights, and reference signals. These updates create a parameter channel: a pathway between the agent and the controller that is structurally distinct from classical sensor and actuator channels. Among the parameters carried by this channel, feedback gains are especially high-leverage: under linear state feedback, a single gain matrix determines closed-loop eigenvalue placement for the entire system. Consequently, malicious gain updates can reshape the closed-loop dynamics without producing the signal-level inconsistencies targeted by residual-based monitors. We formalize this attack surface through a three-axis attacker model and a taxonomy of Gain Manipulation Attacks (GMA). Two impact classes are identified: stability-margin erosion under sustained gain drift and transient amplification under one-shot gain replacement. We demonstrate that an attacker can drive the system past its safe physical operating limits while maintaining mathematical stability, proving that stability verification alone is insufficient to bound the physical impact. Using Bauer--Fike eigenvalue bounds and the Kreiss matrix theorem, we derive exact stealthiness conditions and worst-case impact certificates for each class. Finally, we propose preliminary detection directions and validate our framework through a vehicle lateral dynamics case study.

eess.SY↗

arXiv2026

Stealthy Cyber-Attacks on Vehicle Lateral Dynamics: A System-Theoretic Analysis

This paper studies the vehicle bicycle model under three classes of stealthy cyber-attacks: replay attacks, zero dynamics attacks, and covert attacks. Using a system-theoretic framework, we analyze the feasibility and impact of these attacks on vehicle lateral dynamics. The investigation considers different measurement configurations, including yaw rate, lateral acceleration, and longitudinal acceleration outputs, to evaluate how sensor selection influences attack detectability and system vulnerability. Each attack class is characterized in terms of required system knowledge, communication access, and impact. The analysis shows that replay attacks remain largely model-agnostic, while zero dynamics attacks are fundamentally constrained by control-oriented design choices, particularly output selection, which can eliminate unstable zero dynamics and limit the attack impact. In contrast, covert attacks, enabled by coordinated actuator and sensor manipulation, allow sustained and stealthy deviation of lateral states when sufficient access and system knowledge are available. The effects of actuator and tire saturation are also examined, revealing attack-dependent impacts on stealthiness and effectiveness. Finally, simulation case studies are conducted by using CarSim-Simulink co-simulation to validate and verify the theoretical results.

eess.SY↗

arXiv2026

A Control-Theoretic Foundation for Agentic Systems

This paper develops a control-theoretic framework for analyzing agentic systems embedded within feedback control loops, where an AI agent may adapt controller parameters, select among control strategies, invoke external tools, reconfigure decision architectures, and modify control objectives during operation. These capabilities are formalized by interpreting agency as hierarchical runtime decision authority over elements of the control architecture, leading to an augmented closed-loop representation in which physical states, internal memory, tool outputs, interaction signals, and design variables evolve as a coupled dynamical system. A five-level hierarchy of agency is defined, ranging from fixed control laws to runtime synthesis of control architectures and objectives. The analysis shows that increasing agency introduces interacting dynamical mechanisms such as time-varying adaptation, endogenous switching, decision-induced delays, and structural reconfiguration. The framework is developed in both nonlinear and linear settings, providing explicit design constraints for AI-enabled control systems in safety-critical applications.

eess.SY↗

arXiv2025

Security Risks of Agentic Vehicles: A Systematic Analysis of Cognitive and Cross-Layer Threats

Agentic AI is increasingly being explored and introduced in both manually driven and autonomous vehicles, leading to the notion of Agentic Vehicles (AgVs), with capabilities such as memory-based personalization, goal interpretation, strategic reasoning, and tool-mediated assistance. While frameworks such as the OWASP Agentic AI Security Risks highlight vulnerabilities in reasoning-driven AI systems, they are not designed for safety-critical cyber-physical platforms such as vehicles, nor do they account for interactions with other layers such as perception, communication, and control layers. This paper investigates security threats in AgVs, including OWASP-style risks and cyber-attacks from other layers affecting the agentic layer. By introducing a role-based architecture for agentic vehicles, consisting of a Personal Agent and a Driving Strategy Agent, we will investigate vulnerabilities in both agentic AI layer and cross-layer risks, including risks originating from upstream layers (e.g., perception layer, control layer, etc.). A severity matrix and attack-chain analysis illustrate how small distortions can escalate into misaligned or unsafe behavior in both human-driven and autonomous vehicles. The resulting framework provides the first structured foundation for analyzing security risks of agentic AI in both current and emerging vehicle platforms.

cs.AI↗

arXiv2025

Resource-Aware Stealthy Attacks in Vehicle Platoons

Connected and Autonomous Vehicles (CAVs) are transforming modern transportation by enabling cooperative applications such as vehicle platooning, where multiple vehicles travel in close formation to improve efficiency and safety. However, the heavy reliance on inter-vehicle communication makes platoons highly susceptible to attacks, where even subtle manipulations can escalate into severe physical consequences. While existing research has largely focused on defending against attacks, far less attention has been given to stealthy adversaries that aim to covertly manipulate platoon behavior. This paper introduces a new perspective on the attack design problem by demonstrating how attackers can guide platoons toward their own desired trajectories while remaining undetected. We outline conditions under which such attacks are feasible, analyze their dependence on communication topologies and control protocols, and investigate the resources required by the attacker. By characterizing the resources needed to launch stealthy attacks, we address system vulnerabilities and informing the design of resilient countermeasures. Our findings reveal critical weaknesses in current platoon architectures and anomaly detection mechanisms and provide methods to develop more secure and trustworthy CAV systems.

eess.SY↗

arXiv2025

Agentic Vehicles for Human-Centered Mobility: Definition, Prospects, and Synergistic Co-Development with Vehicle Autonomy

Autonomy, from the Greek autos (self) and nomos (law), refers to the capacity to operate according to internal rules without external control. Autonomous vehicles (AuVs) are therefore understood as vehicular systems that perceive their environment and execute tasks with minimal human intervention, consistent with the direction indicated by the SAE levels of automated driving. However, recent research and deployments increasingly showcase vehicular capabilities that, while not contradicting autonomy, are not entailed by it, including ambiguous goal handling, purposeful social engagement, external tool use, proactive problem solving, continuous learning, and context-sensitive reasoning in unseen and ethically salient situations, enabled in part by multimodal language models. These developments reveal a gap between technical autonomy and the broader social cognitive functions required for human-centered mobility, which are more precisely captured by the notion of agency. Therefore, rather than adding increasingly elaborate modifiers to "autonomous," we introduce agentic vehicles (AgVs) and suggest that autonomy and agency are intertwined but conceptually distinct: if autonomy concerns what to do and how to do it (task executions under internal rules), agency pertains to why to do it and what else can be done (goal-directed, adaptive actions). We present autonomy and agency as orthogonal yet synergistic dimensions with co-development implications. Vehicle agency marks a novel dimension of mobility service intelligence, heralding vehicles as purposeful actors in society.

cs.CY↗

arXiv2025

Zero Dynamics Attack Detection and Isolation in Cyber-Physical Systems with Event-triggered Communication

This paper investigates the problem of Zero Dynamics (ZD) cyber-attack detection and isolation in Cyber-Physical Systems (CPS). By utilizing the notion of auxiliary systems with event-based communications, we will develop a detection mechanism capable of detecting and isolating the ZD cyber-attack even when the attackers have full knowledge of the dynamics of the auxiliary system and can launch False Data Injection (FDI) attacks on all the communication channels. More specifically, we will utilize a self-triggering rule for the communication channels connecting the auxiliary system with the Command & Control (C&C) center, leveraging its properties to detect the ZD cyber-attack. Finally, the effectiveness and capabilities of our approach are verified and demonstrated through simulation case studies.

eess.SY↗

arXiv2025

Advancing Geometry with AI: Multi-agent Generation of Polytopes

Polytopes are one of the most primitive concepts underlying geometry. Discovery and study of polytopes with complex structures provides a means of advancing scientific knowledge. Construction of polytopes with specific extremal structure is very difficult and time-consuming. Having an automated tool for the generation of such extremal examples is therefore of great value. We present an Artificial Intelligence system capable of generating novel polytopes with very high complexity, whose abilities we demonstrate in three different and challenging scenarios: the Hirsch Conjecture, the k-neighbourly problem and the longest monotone paths problem. For each of these three problems the system was able to generate novel examples, which match or surpass the best previously known bounds. Our main focus was the Hirsch Conjecture, which had remained an open problem for over 50 years. The highly parallel A.I. system presented in this paper was able to generate millions of examples, with many of them surpassing best known previous results and possessing properties not present in the earlier human-constructed examples. For comparison, it took leading human experts over 50 years to handcraft the first example of a polytope exceeding the bound conjectured by Hirsch, and in the decade since humans were able to construct only a scarce few families of such counterexample polytopes. With the adoption of computer-aided methods, the creation of new examples of mathematical objects stops being a domain reserved only for human expertise. Advances in A.I. provide mathematicians with yet another powerful tool in advancing mathematical knowledge. The results presented demonstrate that A.I. is capable of addressing problems in geometry recognized as extremely hard, and also to produce extremal examples different in nature from the ones constructed by humans.

math.CO↗

arXiv2023

Consensus, dissensus and synergy between clinicians and specialist foundation models in radiology report generation

Radiology reports are an instrumental part of modern medicine, informing key clinical decisions such as diagnosis and treatment. The worldwide shortage of radiologists, however, restricts access to expert care and imposes heavy workloads, contributing to avoidable errors and delays in report delivery. While recent progress in automated report generation with vision-language models offer clear potential in ameliorating the situation, the path to real-world adoption has been stymied by the challenge of evaluating the clinical quality of AI-generated reports. In this study, we build a state-of-the-art report generation system for chest radiographs, $\textit{Flamingo-CXR}$, by fine-tuning a well-known vision-language foundation model on radiology data. To evaluate the quality of the AI-generated reports, a group of 16 certified radiologists provide detailed evaluations of AI-generated and human written reports for chest X-rays from an intensive care setting in the United States and an inpatient setting in India. At least one radiologist (out of two per case) preferred the AI report to the ground truth report in over 60$\%$ of cases for both datasets. Amongst the subset of AI-generated reports that contain errors, the most frequently cited reasons were related to the location and finding, whereas for human written reports, most mistakes were related to severity and finding. This disparity suggested potential complementarity between our AI system and human experts, prompting us to develop an assistive scenario in which Flamingo-CXR generates a first-draft report, which is subsequently revised by a clinician. This is the first demonstration of clinician-AI collaboration for report writing, and the resultant reports are assessed to be equivalent or preferred by at least one radiologist to reports written by experts alone in 80$\%$ of in-patient cases and 60$\%$ of intensive care cases.

eess.IV↗

arXiv2020

Game Plan: What AI can do for Football, and What Football can do for AI

The rapid progress in artificial intelligence (AI) and machine learning has opened unprecedented analytics possibilities in various team and individual sports, including baseball, basketball, and tennis. More recently, AI techniques have been applied to football, due to a huge increase in data collection by professional teams, increased computational power, and advances in machine learning, with the goal of better addressing new scientific challenges involved in the analysis of both individual players' and coordinated teams' behaviors. The research challenges associated with predictive and prescriptive football analytics require new developments and progress at the intersection of statistical learning, game theory, and computer vision. In this paper, we provide an overarching perspective highlighting how the combination of these fields, in particular, forms a unique microcosm for AI research, while offering mutual benefits for professional teams, spectators, and broadcasters in the years to come. We illustrate that this duality makes football analytics a game changer of tremendous value, in terms of not only changing the game of football itself, but also in terms of what this domain can mean for the field of AI. We review the state-of-the-art and exemplify the types of analysis enabled by combining the aforementioned fields, including illustrative examples of counterfactual analysis using predictive models, and the combination of game-theoretic analysis of penalty kicks with statistical learning of player attributes. We conclude by highlighting envisioned downstream impacts, including possibilities for extensions to other sports (real and virtual).

cs.AI↗

arXiv2019

Local Voting Games for Misbehavior Detection in VANETs in Presence of Uncertainty

Cooperation between neighboring vehicles is an effective solution to the problem of malicious node identification in vehicular ad hoc networks (VANETs). However, the outcome is subject to nodes' beliefs and reactions in the collaboration. In this paper, a plain game-theoretic approach that captures the uncertainty of nodes about their monitoring systems, the type of their neighboring nodes, and the outcome of the cooperation is proposed. In particular, one stage of a local voting-based scheme for identifying a target node is developed using a Bayesian game. In this context, incentives are offered in expected utilities of nodes in order to promote cooperation in the network. The proposed model is then analyzed to obtain equilibrium points, ensuring that no node can improve its utility by changing its strategy. Finally, the behavior of malicious and benign nodes is studied by extensive simulation results. Specifically, it is shown how the existing uncertainties and the designed incentives impact the strategies of the players and, consequently, the correct target-node identification.

cs.GT↗

arXiv2019

Attentive Neural Processes

Neural Processes (NPs) (Garnelo et al 2018a;b) approach regression by learning to map a context set of observed input-output pairs to a distribution over regression functions. Each function models the distribution of the output given an input, conditioned on the context. NPs have the benefit of fitting observed data efficiently with linear complexity in the number of context input-output pairs, and can learn a wide family of conditional distributions; they learn predictive distributions conditioned on context sets of arbitrary size. Nonetheless, we show that NPs suffer a fundamental drawback of underfitting, giving inaccurate predictions at the inputs of the observed data they condition on. We address this issue by incorporating attention into NPs, allowing each input location to attend to the relevant context points for the prediction. We show that this greatly improves the accuracy of predictions, results in noticeably faster training, and expands the range of functions that can be modelled.

cs.LG↗

arXiv2019

Risk Assessment of Autonomous Vehicles Using Bayesian Defense Graphs

Recent developments have made autonomous vehicles (AVs) closer to hitting our roads. However, their security is still a major concern among drivers as well as manufacturers. Although some work has been done to identify threats and possible solutions, a theoretical framework is needed to measure the security of AVs. In this paper, a simple security model based on defense graphs is proposed to quantitatively assess the likelihood of threats on components of an AV in the presence of available countermeasures. A Bayesian network (BN) analysis is then applied to obtain the associated security risk. In a case study, the model and the analysis are studied for GPS spoofing attacks to demonstrate the effectiveness of the proposed approach for a highly vulnerable component.

cs.CR↗

arXiv2018

Generating Diverse Programs with Instruction Conditioned Reinforced Adversarial Learning

Advances in Deep Reinforcement Learning have led to agents that perform well across a variety of sensory-motor domains. In this work, we study the setting in which an agent must learn to generate programs for diverse scenes conditioned on a given symbolic instruction. Final goals are specified to our agent via images of the scenes. A symbolic instruction consistent with the goal images is used as the conditioning input for our policies. Since a single instruction corresponds to a diverse set of different but still consistent end-goal images, the agent needs to learn to generate a distribution over programs given an instruction. We demonstrate that with simple changes to the reinforced adversarial learning objective, we can learn instruction conditioned policies to achieve the corresponding diverse set of goals. Most importantly, our agent's stochastic policy is shown to more accurately capture the diversity in the goal distribution than a fixed pixel-based reward function baseline. We demonstrate the efficacy of our approach on two domains: (1) drawing MNIST digits with a paint software conditioned on instructions and (2) constructing scenes in a 3D editor that satisfies a certain instruction.

cs.LG↗

arXiv2018

A Micro-Scale Mobile-Enabled Implantable Medical Sensor

Micro-scale implantable medical devices (IMDs) extend the immense benefits of sensors used in health management. However, their development is limited by many requirements and challenges, such as the use of safe materials, size restrictions, safe and efficient powering, and selection of suitable wireless communication technologies. Some of the proposed wireless communication technologies are the terahertz (THz) radio frequency (RF), and ultrasound. This paper investigates the use of {\em magnetic induction-based backscatter communication} as an alternative technology. In particular, the goal is to provide a practical design for a micro-scale IMD, referred to as a "biomote" here, that can communicate with a wearable or handheld device such as a cell phone, tablet, or smart watch.First, it is demonstrated that communication via magnetic induction can be established between a biomote and such an external reader. Then, low-power modulation and error-correction coding schemes that can be implemented in micro-scale are explored for the mote. With the aim of increasing reliability and accuracy of measurements through mass deployment of biomotes, suitable low-power media access control (MAC) schemes are proposed, and the feasibility of their implementation in micro-scale is highlighted. Next, assuming that the human body is an additive white Gaussian noise (AWGN) channel, the performance of the mote is simulated and analyzed. Results of this analysis asserts that a communication range of at least few centimeters is achievable with an acceptable bit error rate. Finally, from the analysis of the MAC schemes, the optimum number of motes to be deployed for various read delays and transmission rates is obtained.

cs.IT↗

arXiv2017

Error Correction Coding Meets Cyber-Physical Systems: Message-Passing Analysis of Self-Healing Interdependent Networks

Coupling cyber and physical systems gives rise to numerous engineering challenges and opportunities. An important challenge is the contagion of failure from one system to another, which can lead to large-scale cascading failures. However, the self-healing ability emerges as a valuable opportunity where the overlaying cyber network can cure failures in the underlying physical network. To capture both self-healing and contagion, this paper considers a graphical model representation of an interdependent cyber-physical system, in which nodes represent various cyber or physical functionalities, and edges capture the interactions between the nodes. A message-passing algorithm is proposed for this representation to study the dynamics of failure propagation and healing. By conducting a density evolution analysis for this algorithm, network reaction to initial disruptions is investigated. It is proved that as the number of message-passing iterations increases, the network reaches a steady-state condition that would be either a complete healing or a complete collapse. Then, a sufficient condition is derived to select the network parameters to guarantee the complete healing of the system. The result of the density evolution analysis is further employed to jointly optimize the design of cyber and physical networks for maximum resiliency. This analytical framework is then extended to the cases where propagation of failures in the physical network is faster than the healing responses of the cyber network. Such scenarios are of interest in many real-life applications such as smart grid. Finally, extensive numerical results are presented to verify the analysis and investigate the impact of the network parameters on the resiliency of the network.

cs.SI↗