The rapid proliferation of social networks has revolutionized communication and social interactions, rendering the study of opinion dynamics (OD) within these platforms an essential area of research. OD offers a powerful lens for understanding, simulating, and predicting behavioral patterns and interactions among individuals on social media networks. We conducted an advanced and comprehensive review examining the methodologies and tools used in this domain, focusing on agent-based modeling, network topology, dynamic modeling, and behavioral modeling within multiagent systems (MASs). Key challenges, such as computational complexity, data quality, and model validation, and potential strategies to overcome these limitations are discussed. The review also highlights critical trends and interdisciplinary opportunities, highlighting the integration of emerging technologies and the importance of ethical considerations in research. By studying advancements in simulation, analysis, and prediction in social media networks through OD, this work provides a comprehensive resource for researchers and practitioners to deepen their understanding and develop impactful applications in this field.

This paper introduces a cooperative consensus Q-learning method that targets tumors using micro multi-agent reinforcement learning (MMARL) systems. The micro-agents in this system navigate toward cancerous tumors autonomously in a simulated 2-D vascular environment using Q-learning and cooperative position consensus. In this proposed method, the agents share Q-values and align their movements with each other. Employing this strategy results in coordinated behavior. Under bounded perturbations, theoretical analyses show convergence guarantees for Q-value and cooperative position consensuses. The simulation results demonstrate that this offered technique boosts convergence speed, improves stability, and enhances cumulative rewards compared to the standard consensus method, highlighting its significance for biomedical applications.

In this paper, we introduce Gen-VCA, a generative vision-based covert attack that specifically targets lane keeping driver assistance systems. Gen-VCA drives the AV out of the lane by injecting a defective signal into the steering angle. Then, it uses a customized diffusion-based generative model to create realistic front-view road views, misleading the perception system into believing that the vehicle is correctly aligned with the center line. To detect the malicious behavior of AVs, we propose an adversary detection system that integrates multi-scale forensic image analysis and deep discriminative neural networks. Experiments in real-world driving conditions reveal that Gen-VCA excels in novel center-aligned view synthesis of the road and induces significant lateral steering errors in AVs, while the detection system accurately detects these manipulations with minimal errors.

In this paper, we consider reinforcement learning (RL) techniques to systematically analyze and enhance the levels of disagreement and polarization within social media ecosystems. The proposed methodology employs a Double Deep Q- Learning algorithm to strategically identify individuals within the network. This identification process is aimed at selecting agents for takeover and control, thereby orchestrating a scenario that culminates in the maximization of disagreement and polarization within the network. The social media network is modeled by an asynchronous and synchronous expressed and private opinion dynamics model. The model incorporates a dual-state update mechanism: a synchronous update process for the state representing an individual's private opinion and an asynchronous update process for the state that reflects the individual's publicly expressed opinion. The RL agent's observational capacity is limited to the expressed opinions of individuals and the quantifiable metric of their followers or connections. The proposed model is analyzed for varying topologies and convergence conditions. Simulations are provided to illustrate the results.

This article studies a formation control problem for a group of heterogeneous, nonlinear, uncertain, input-affine, second-order agents modeled by a directed graph. A tunable neural network (NN) is presented, with three layers (input, two hidden, and output) that can approximate an unknown nonlinearity. Unlike one- or two-layer NNs, this design has the advantage of being able to set the number of neurons in each layer ahead of time rather than relying on trial and error. The NN weights tuning law is rigorously derived using the Lyapunov theory. The formation control problem is tackled using a robust integral of the sign of the error feedback and NNs-based control. The robust integral of the sign of the error feedback compensates for the unknown dynamics of the leader and disturbances in the agent errors, while the NN-based controller accounts for the unknown nonlinearity in the multiagent system. The stability and semi-global asymptotic tracking of the results are proven using the Lyapunov stability theory. The study compares its results with two others to assess the effectiveness and efficiency of the proposed method.

This paper investigates the data transmission resilience between agents of a cluster-based, heterogeneous, multi-agent deep reinforcement learning (MADRL) system under gradient-based adversarial attacks. We propose an algorithm using a deep Q-network (DQN) approach and a proportional feedback controller to defend against the fast gradient sign method (FGSM) attack and improve the DQN agent performance. The feedback control system is an auxiliary tool that helps the DQN algorithm reduce system deficiencies. In accordance with the achieved results and under FGSM adversarial attack, the resilience of the developed system is evaluated in three different ways termed robust, semi-robust, and non-robust based on average reward and DQN loss. The data transfer is carried out between agents of a MADRL system in timely and time-delayed manners, for both leaderless and leader-follower scenarios. Simulation results are included to verify the presented results.

In this paper, we consider a discrete private and expressed (synchronous and asynchronous) opinion dynamics model with competitive relationships. Unlike the usual agent-based opinion dynamics models, competition between individuals is investigated in a social media network. The expressed opinions, or states of the individuals in the network, are represented by asynchronous dynamics, where each individual has a choice to express his/her opinion at each time step. Each agent uses a Q-earning algorithm to decide when to express its opinion with the purpose of swaying the opinions of other connected agents to a desired outcome. The private opinions, or states of the individuals, are derived from a combination of their own private opinions and the expressed opinions of connected agents. The dynamics of the social media environment are modeled by private and expressed, both asynchronous and synchronous, opinion dynamics model. The system is investigated for polarization or consensus under different conditions. To illustrate the results, simulations of the system dynamics are provided.

Autonomous vehicles are prone to worms of intelligent cyber-attacks that can use novel deep neural networks to adapt themselves to hosts and remain stealthy for a long period of time. This paper introduces a new, vision-based, covert attack and detection method, called VCAD-GAN, which can be applied to the navigation system of various autonomous vehicles. VCAD-GAN injects a faulty signal into the actuator channel to drive the vehicle out of the lane. To conceal the deviation effects, VCAD-GAN manipulates the camera output using a generative adversarial network such that the vehicle is shown on the road’s center-line and aligned with it. The generator of VCAD-GAN reconstructs a synthesized top view of the road, while the discriminator classifies it as authentic or fake and sends feedback to the generator. A hybrid adversary detection system is also developed for VCAD-GAN using a customized deep neural network, global positioning system data, and an offline map. To evaluate the performance of VCAD-GAN, various 3D simulations were conducted. The simulation results show the validity and effectiveness of the proposed methods.

In this paper, a formation of second-order holonomic agents is made to navigate through an obstacle field using proximal policy optimization (PPO) based deep reinforcement learning (DRL). The angle-based formation is allowed to shrink while maintaining its shape in order to navigate through tight spaces and take the geometric centroid of the formation towards the goal. Two reward schemes are presented, one based on the actions of individual agents and another based on the actions of the team as a whole. For each case, all the agents share a single policy that is trained in a centralized manner. Distance measurements, state information, error information regarding neighboring agents, and simulation information are used for training each policy in an end-to-end fashion. Simulation results for both approaches are compared.

In this paper, we consider a synchronous and an asynchronous discrete opinion dynamics model of social influence networks. The model has a synchronously updating state that represents the individual’s private opinion and an asynchronously updating state that represents the individual’s expressed opinion. The private opinion of each individual in the network is influenced by the expressed opinion of their connected neighbors while the expressed opinion of each individual is influenced by the expressed opinions of the neighboring agents and the social pressure to conform to public opinion. The expressed opinion of each individual in the network is randomly generated at each time step. In addition, the number of agents that choose to express their opinion at each time step is randomly generated. The introduced model is analyzed for social networks with varying topologies and convergence conditions. Simulations are provided to illustrate the results.

Unordered feature sets are a nonstandard data structure that traditional neural networks are incapable of addressing in a principled manner. Providing a concatenation of features in an arbitrary order may lead to the learning of spurious patterns or biases that do not actually exist. Another complication is introduced if the number of features varies between each set. We propose convolutional deep averaging networks (CDANs) for classifying and learning representations of datasets whose instances comprise variable-size, unordered feature sets. CDANs are efficient, permutation-invariant, and capable of accepting sets of arbitrary size. We emphasize the importance of nonlinear feature embeddings for obtaining effective CDAN classifiers and illustrate their advantages in experiments versus linear embeddings and alternative permutation-invariant and -equivariant architectures. We also find that pooling via summation has a significant advantage in CDANs over average- or max-pooling.

This paper introduces an adaptable human-robot interface that uses two types of human-computer interactions: an image processing technique for a right-hand gesture recognition and a smart glove for left-hand commands. A fixed number of gestures is used for specific commands to the vehicle (takeoff, land, hover, etc.), while the smart glove is used for the vehicle motors control. A single shot multi-box detector (SSD) model is used for a hand detection. After removing the cluttered background, the region of interest (RoI) is fed to a convolutional neural network (CNN) for right-hand gesture recognition. We propose three concurrent validation layers including a human-based validation. The validation layers allow the system to adapt to various users including different skin colors and hand shapes. Four flex sensors and a motion processing unit (MPU) are used in the smart glove to measure the bending ratio of each finger and the roll angle of the left hand. These signals are used for a left-hand gesture recognition as well as generation of continuous control signals such as throttle and angle commands of the vehicle. Extensive experimental results are presented that validate the proposed control methods.

This paper deals with a multi-layered formation control of autonomous marine vehicles (AMVs). Each AMV is considered as an agent and modeled with a nonlinear dynamics. The nonlinear dynamics include a square law drag in the velocity of the vehicle and saturation in the acceleration input. The system is stabilized using a rigid graph-based control approach. A Lyapunov candidate is chosen and proved that it satisfies conditions for an energy function of the formulated problem. Using the proposed Lyapunov energy function and related control law, an asymptotic stability with improved rate convergence of the system is rigorously proven. The required constraints for prevention of ambiguous formations that cause failure of convergence in the system have been developed. The simulation results illustrate the effectiveness of the proposed control law.

Positive definite (PD) kernels are an important tool in machine learning that enable efficient solutions to otherwise difficult or intractable problems by implicitly linearizing the problem geometry. In this paper, we develop a set-theoretic interpretation of the earth mover's distance (EMD) and propose earth mover's intersection (EMI), a PD analog to EMD for sets of different sizes. We provide conditions under which EMD or certain approximations to EMD are negative definite. We also present a PD-preserving transformation that can be applied to any kernel and can also be used to derive PD EMD-based kernels and show that the Jaccard index is simply the result of this transformation. Finally, we evaluate kernels based on EMI and the proposed transformation versus EMD in various computer vision tasks and show that EMD is generally inferior even with indefinite kernel techniques.

This paper provides a solution to the nonplanar multiagent formation control problem using graph rigidity. We consider a 3-D multiagent formation control where multiple agents are operating in one plane and some other agents are operating outside of that plane. This can be referred to as a layered formation control where the objective is for all agents to cooperatively acquire a predefined formation shape using a decentralized control law. The proposed control strategy is based on regulating the interagent distances. A rigorous stability analysis is presented that guarantees convergence of these distances to desired values. Simulation results are presented to support the theoretical results.

We present a distance metric based upon the notion of minimum-cost injective mappings between sets. Our function satisfies metric properties as long as the cost of the minimum mappings is derived from a semimetric, for which the triangle inequality is not necessarily satisfied. We show that the Jaccard distance (alternatively biotope, Tanimoto, or Marczewski-Steinhaus distance) may be considered the special case for finite sets where costs are derived from the discrete metric. Extensions that allow premetrics (not necessarily symmetric), multisets (generalized to include probability distributions), and asymmetric mappings are given that expand the versatility of the metric without sacrificing metric properties. The function has potential applications in pattern recognition, machine learning, and information retrieval.

This paper is concerned with the evaluation and comparison of several methods for the classification and recognition of static hand postures from small unlabeled point sets corresponding to physical landmarks, e.g. reflective marker positions in a motion capture environment. We compare various classification algorithms based upon multiple interpretations and feature transformations of the point sets, including those based upon aggregate features (e.g. mean) and a pseudo-rasterization of the space. We find aggregate feature classifiers to be balanced across multiple users but relatively limited in maximum achievable accuracy. Certain classifiers based upon the pseudo-rasterization performed best among tested classification algorithms. The inherent difficulty in classifying certain users leads us to conclude that online learning may be necessary for the recognition of natural gestures.

This work provides a solution to a non-planar multi-agent layered sensing and formation control problem using coning graphs. In particular, the problem addressed consists of one agent coordinating the actions of other interacting agents which are operating in a different plane. The objective is for the agents to cooperatively acquire a pre-defined formation shape using a decentralized control law. The proposed control strategy is based on inter-agent distances for single-integrator agent model and consists of formation acquisition term for non-planar agents. The simulation results that support the proposed approach are presented.

A neural net (NN)-based actuator saturation compensation scheme for the nonlinear systems in Brunovsky canonical form is presented. The scheme that leads to stability, command following, and disturbance rejection is rigorously proved and verified using a general "pendulum type" and a robot manipulator dynamical systems. Online weights tuning law, the overall closed-loop system performance, and the boundedness of the NN weights are derived and guaranteed based on Lyapunov approach. The actuator saturation is assumed to be unknown and the saturation compensator is inserted into a feedforward path. Simulation results indicate that the proposed scheme can effectively compensate for the saturation nonlinearity in the presence of system uncertainty.

One of the most important properties of neural nets (NNs) for control purposes is the universal approximation property. Unfortunately,, this property is generally proven for continuous functions. In most real industrial control systems there are nonsmooth functions (e.g., piecewise continuous) for which approximation results in the literature are sparse. Examples include friction, deadzone, backlash, and so on. It is found that attempts to approximate piecewise continuous functions using smooth activation functions require many NN nodes and many training iterations, and still do not yield very good results. Therefore, a novel neural-network structure is given for approximation of piecewise continuous functions of the sort that appear in friction, deadzone, backlash, and other motion control actuator nonlinearities. The novel NN consists of neurons having standard sigmoid activation functions, plus some additional neurons having a special class of nonsmooth activation functions termed "jump approximation basis function." Two types of nonsmooth jump approximation basis functions are determined- a polynomial-like basis and a sigmoid-like basis. This modified NN with additional neurons having "jump approximation" activation functions can approximate any piecewise continuous function with discontinuities at a finite number of known points. Applications of the new NN structure are made to rigid-link robotic systems with friction nonlinearities. Friction is a nonlinear effect that can limit the performance of industrial control systems; it occurs in all mechanical systems and therefore is unavoidable in control systems. It can cause tracking errors, limit cycles, and other undesirable effects. Often, inexact friction compensation is used with standard adaptive techniques that require models that are linear in the unknown parameters. It is shown here how a certain class of augmented NN, capable of approximating piecewise continuous functions, can be used for friction compensation.

A dynamics inversion compensation scheme is designed for control of nonlinear discrete-time systems with input backlash. The compensator uses the backstepping technique with neural networks (NN) for inverting the backlash nonlinearity in the feedforward path. The technique provides a general procedure for using NN to determine the dynamics pre-inverse of an invertible discrete time dynamical system. A discrete-time tuning algorithm is given for the NN weights so that the backlash compensation scheme becomes adaptive, guaranteeing bounded tracking and backlash errors, and also bounded parameter estimates. A rigorous proof of stability and performance is given and a simulation example verifies the performance. Unlike standard discrete-time adaptive control techniques, no certainty equivalence assumption is needed.