Agent Dynamics Research Articles

Adaptive Event-Triggered Consensus of Multi-Agent Systems in Sense of Asymptotic Convergence.

In this paper, the asymptotic consensus control of multi-agent systems with general linear agent dynamics is investigated. A neighbor-based adaptive event-triggering strategy with a dynamic triggering threshold is proposed, which leads to a fully distributed control of the multi-agent system, depending only on the states of the neighboring agents at triggering moments. By using the Lyapunov method, we prove that the states of the agents converge asymptotically. In addition, the proposed event-triggering strategy is proven to exclude Zeno behavior. The numerical simulation results illustrate that the agent states achieve consensus in sense of asymptotic convergence. Furthermore, the proposed strategy is shown to be scalable in case of variable agent numbers.

Efficient Bike-sharing Repositioning with Cooperative Multi-Agent Deep Reinforcement Learning

As an emerging mobility-on-demand service, bike-sharing system (BSS) has spread all over the world by providing a flexible, cost-efficient, and environment-friendly transportation mode for citizens. Demand-supply unbalance is one of the main challenges in BSS because of the inefficiency of the existing bike repositioning strategy, which reallocates bikes according to a pre-defined periodic schedule without considering the highly dynamic user demands. While reinforcement learning has been used in some repositioning problems for mitigating demand-supply unbalance, there are significant barriers when extending it to BSS due to the dimension curse of action space resulting from the dynamic number of workers and bikes in the city. In this paper, we study these barriers and address them by proposing a novel bike repositioning system, namely BikeBrain, which consists of a demand prediction model and a spatio-temporal bike repositioning algorithm. Specifically, to obtain accurate and real-time usage demand for efficient bike repositioning, we first present a prediction model ST-NetPre, which directly predicts user demand considering the highly dynamic spatio-temporal characteristics. Furthermore, we propose a spatio-temporal cooperative multi-agent reinforcement learning method (ST-CBR) for learning the worker-based bike repositioning strategy in which each worker in BSS is considered an agent. Especially, ST-CBR adopts the centralized learning and decentralized execution way to achieve effective cooperation among large-scale dynamic agents based on Mean Field Reinforcement Learning (MFRL), while avoiding the huge dimension of action space. For dynamic action space, ST-CBR utilizes a SoftMax selector to select the specific action. Meanwhile, for the benefits and costs of agents’ operation, an efficient reward function is designed to seek an optimal control policy considering both immediate and future rewards. Extensive experiments are conducted based on large-scale real-world datasets, and the results have shown significant improvements of our proposed method over several state-of-the-art baselines on the demand-supply gap and operation cost measures.

$ H_\infty $ deployment of nonlinear multi-agent systems with Markov switching topologies over a finite-time interval based on T–S fuzzy PDE control

<abstract><p>The deployment of multi-agent systems (MASs) is widely used in the fields of unmanned agricultural machineries, unmanned aerial vehicles, intelligent transportation, etc. To make up for the defect that the existing PDE-based results are overly idealistic in terms of system models and control strategies, we study the PDE-based deployment of clustered nonlinear first-order and second-order MASs over a finite-time interval (FTI). By designing special communication protocols, the collective dynamics of numerous agents are modeled by simple fist-order and second-order PDEs. Two practical factors, external disturbance and Markov switching topology, are considered in this paper to better match actual situations. Besides, T–S fuzzy technology is used to approximate the unknown nonlinearity of MASs. Then, by using boundary control scheme with collocated measurements, two theorems are obtained to ensure the finite-time $ H_\infty $ deployment of first-order and second-order agents, respectively. Finally, numerical examples are provided to illustrate the effectiveness of the proposed approaches.</p></abstract>

Toward model-free safety-critical control with humans in the loop

This vision article shows how to build on the framework of event-triggered Control Barrier Functions (CBFs) to design model-free controllers for safety-critical multi-agent systems with unknown dynamics, including humans in the loop. This event-triggered framework has been shown to be computationally efficient and robust while guaranteeing safety for systems with unknown dynamics. We show how to extend it to model-free safety critical control where a controllable ego agent does not need to model the dynamics of other agents and updates its control based only on events dependent on the error states of agents obtained by real-time sensor measurements. To facilitate the process of real-time sensor measurements critical in this approach, we also present CBF relative degree reduction methods, which can reduce the number of such measurements. We illustrate the effectiveness of the proposed framework on a multi-agent traffic merging decentralized control problem and on highway lane changing control with humans in the loop and relative degree reduction. We also compare the proposed event-driven method to the classical time-driven approach.

Leader-Following Consensus of Linear Multiagent Systems With Aperiodically Sampled Outputs: A Distributed Impulsive-Observer-Based Approach.

This article studies the leader-following consensus problem for a class of linear multiagent systems over a directed graph with aperiodically sampled outputs. First, a novel distributed impulsive-observer-based consensus protocol is designed. This protocol requires only the output measurements at sporadic time instants for the observer and control gain design. Second, by using time-varying Lyapunov function techniques, sufficient conditions for exponential stability of a class of linear impulsive systems are established; subsequently, these stability results are applied for the distributed impulsive observer design. Different from the existing related works, the impulsive observer gain designed in this work is decoupled from the graph properties. As a result, once the impulsive observer gain is designed for one network topology, it can be directly used for other network topologies, as long as the graph properties and the dynamics of local agents satisfy certain conditions. Furthermore, the resilience of the designed protocol is tested under denial of service (DoS) attacks. It is shown that the protocol is robust with respect to low-frequency DoS attacks occurring in the observer communication network. Finally, two examples illustrating the validity and effectiveness of the proposed protocol are included.

Inverse Value Iteration and Q -Learning: Algorithms, Stability, and Robustness.

This article proposes a data-driven model-free inverse Q -learning algorithm for continuous-time linear quadratic regulators (LQRs). Using an agent's trajectories of states and optimal control inputs, the algorithm reconstructs its cost function that captures the same trajectories. This article first poses a model-based inverse value iteration scheme using the agent's system dynamics. Then, an online model-free inverse Q -learning algorithm is developed to recover the agent's cost function only using the demonstrated trajectories. It is more efficient than the existing inverse reinforcement learning (RL) algorithms as it avoids the repetitive RL in inner loops. The proposed algorithms do not need initial stabilizing control policies and solve for unbiased solutions. The proposed algorithm's asymptotic stability, convergence, and robustness are guaranteed. Theoretical analysis and simulation examples show the effectiveness and advantages of the proposed algorithms.

Heterogeneity of agents, transaction costs and oil price dynamics

ABSTRACT In this study, we model the dynamics of oil prices based on a dynamic heterogeneous agent model. Two types of agents including fundamentalists and chartists trade in the market, and the transaction costs are taken into consideration. The empirical results show that it is useful to improve forecasting performance if the model allows the agents to learn from their past performance and considers the transaction costs. Examination of the agents’ dynamic weights shows that fundamentalists and chartists co-existed in the oil market with almost equal weights, and the fundamentalists dominated the market when the oil prices had large deviations from the fundamental level.

Assessing Perfusion in Steno-Occlusive Cerebrovascular Disease Using Transient Hypoxia-Induced Deoxyhemoglobin as a Dynamic Susceptibility Contrast Agent.

Resting brain tissue perfusion in cerebral steno-occlusive vascular disease can be assessed by MR imaging using gadolinium-based susceptibility contrast agents. Recently, transient hypoxia-induced deoxyhemoglobin has been investigated as a noninvasive MR imaging contrast agent. Here we present a comparison of resting perfusion metrics using transient hypoxia-induced deoxyhemoglobin and gadolinium-based contrast agents in patients with known cerebrovascular steno-occlusive disease. Twelve patients with steno-occlusive disease underwent DSC MR imaging using a standard bolus of gadolinium-based contrast agent compared with transient hypoxia-induced deoxyhemoglobin generated in the lungs using an automated gas blender. A conventional multi-slice 2D gradient echo sequence was used to acquire the perfusion data and analyzed using a standard tracer kinetic model. MTT, relative CBF, and relative CBV maps were generated and compared between contrast agents. The spatial distributions of the perfusion metrics generated with both contrast agents were consistent. Perfusion metrics in GM and WM were not statistically different except for WM MTT. Cerebral perfusion metrics generated with noninvasive transient hypoxia-induced changes in deoxyhemoglobin are very similar to those generated using a gadolinium-based contrast agent in patients with cerebrovascular steno-occlusive disease.

Is this a violation? Learning and understanding norm violations in online communities

Using norms to guide and coordinate interactions has gained tremendous attention in the multi-agent community. However, new challenges arise as the interest moves towards dynamic socio-technical systems, where human and software agents interact, and interactions are required to adapt to human's changing needs. For instance, different agents (human or software) might not have the same understanding of what it means to violate a norm (e.g., what characterizes hate speech), or their understanding of a norm might change over time (e.g., what constitutes an acceptable response time). The challenge is to address these issues by learning the meaning of a norm violation from limited interaction data. For this, we use batch and incremental learning to train an ensemble of classifiers. Ensemble learning and data-sampling handle the imbalanced class distribution of the interaction stream. At the same time, the training approaches use different strategies to ensure that the ensemble models reflect the latest community view on the meaning of norm violation. Batch learning uses weight assignment, while incremental learning continuously updates the ensemble models as community members interact. Here, we extend our previous work by creating a different balance strategy for online learning and integrating interpretability to understand norm violations. Additionally, we evaluate the proposed approaches in the context of Wikipedia article edits, where interactions revolve around editing articles, and the norm in question is prohibiting vandalism. Lastly, we conduct ablation studies to compare the ensemble's performance against a single model approach and to examine the behavior of two data sampling techniques. Results indicate that the different machine learning frameworks can learn the meaning of a norm violation in a setting with data imbalance and concept drift, although with significant differences.

The Post-Curing of Waterborne Polyurethane-Acrylate Composite Latex with the Dynamic Disulfide-Bearing Crosslinking Agent.

The development of a dynamic network for commodity polymer systems via feasible methods has been explored in the context of a society-wide focus on the environment and sustainability. Herein, we demonstrate an adaptive post-curing method used to build a self-healable network of waterborne polyurethane-acrylate (WPUA) composite latex. The composite latex was synthesized via the miniemulsion polymerization of acrylates in the dispersion of waterborne polyurethane (PU), with commercial acetoacetoxyethyl methacrylate (AAEM) serving as the functional monomer. Then, a dynamic disulfide (S-S)-bearing diamine was applied as the crosslinking agent for the post-curing of the hybrid latex via keto-amine condensation, which occurred during the evaporation of water for film formation. It was revealed that the microphase separation in the hybrid films was suppressed by the post-curing network. The mechanical performance exhibited a high reliability as regards the contents of the crosslinking agents. The reversible exchange of S-S bonds meant that the film displayed associative covalent-adaptive networks in the range of medium temperature in stress relaxation tests, and ≥95% recovery in both the stress and the strain was achieved after the cut-off films were self-healed at 70 °C for 2 h. The rebuilding of the network was also illustrated by the >80% recovery in the elongation at break of the films after three crushing-hot pressing cycles. These findings offer valuable insights, not only endowing the traditional WPUA with self-healing and reprocessing properties, but broadening the field of study of dynamic networks to polymer hybrid latex.

Introduction to special issue: Advancing disease ecology through eDNA monitoring of infectious agents

AbstractThis special issue focuses on the applications of environmental DNA (eDNA) sequencing for the detection and monitoring of disease‐causing agents, including viruses, bacteria, protozoans, myxozoans, fungi, trematodes, and arthropods. We explore the impact of eDNA technologies, such as metabarcoding and qPCR, in understanding the dynamics of pathogens in various environments as well as their implications for conservation, biosecurity, and veterinary and agricultural health under the “One Health” framework. This issue addresses how molecular sequencing provides innovative solutions to the challenges faced by conventional parasite and pathogen detection methods, enabling a more comprehensive understanding of the spatiotemporal dynamics of disease agents. Finally, we discuss the challenges in eDNA applications, such as primer development and taxonomic resolution, and the opportunities for future research in advancing eDNA methodologies for infectious disease studies. This issue highlights the growing importance of eDNA surveillance in understanding and managing the health of ecosystems and at‐risk species.

An Inexact-Penalty Method for GNE Seeking in Games With Dynamic Agents

We consider a network of autonomous agents whose outputs are actions in a game with coupled constraints. In such network scenarios, agents seek to minimize coupled cost functions using distributed information while satisfying the coupled constraints. Current methods consider the small class of multi-integrator agents using primal-dual methods. These methods can only ensure constraint satisfaction in steady-state. In contrast, we propose an inexact penalty method using a barrier function for nonlinear agents with equilibrium-independent passive dynamics. We show that these dynamics converge to an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\varepsilon$</tex-math></inline-formula> -GNE while satisfying the constraints for all time, not only in steady-state. We develop these dynamics in both the full-information and partial-information settings. In the partial-information setting, dynamic estimates of the others' actions are used to make decisions and are updated through local communication. Applications to optical networks, velocity synchronization of flexible robots and wind farm optimization are provided.

Heterogeneous decision-making dynamics of threshold-switching agents on complex networks.

In the classical two-player decision-making scenario, individuals may have different tendencies to take a certain action, given that there exists a sufficient number of neighbors adopting a particular option. This is ubiquitous in many real-life contexts including traffic congestion, crowd evacuation, and minimal vertex cover problem. Under best-response dynamics, we investigate the decision-making behaviors of heterogeneous agents on complex networks. Results of the networked games are twofold: for networks of uniform degree distribution (e.g., the lattice) and fraction of the strategy is of a linear function of the threshold setting. Moreover, the equilibrium analysis is provided and the relationship between the equilibrium dynamics and the change of the threshold value is given quantitatively. Next, if the games are played on networks with non-uniform degree distribution (e.g., random regular and scale-free networks), influence of the threshold-switching will be weakened. Robust experiments indicate that it is not the value of the average degree, but the degree distribution that influences how the strategy evolves affected by the threshold settings. Our result shows that the decision-making behaviors can be effectively manipulated by tuning the parameters in the utility function (i.e., thresholds) of some agents for more regular network structures.

Modality Exploration, Retrieval and Adaptation for Trajectory Prediction.

Predicting future trajectories of dynamic agents is inherently riddled with uncertainty. Given a certain historical observation, there are multiple plausible future movements people can perform. Notably, these possible movements are usually centralized around a few representative motion patterns, e.g. acceleration, deceleration, turning, etc. In this paper, we propose a novel prediction scheme which explores human behavior modality representations from real-world trajectory data to discover such motion patterns and further uses them to aid in trajectory prediction. To explore potential behavior modalities, we introduce a deep feature clustering process on trajectory features and each cluster can represent a type of modality. Intuitively, each modality is naturally a class, and a classification network can be adopted to retrieve highly probable modalities about to happen in the future according to historical observations. On account of a wide variety of cues existing in the observation (e.g. agents' motion states, semantics of the scene, etc.), we further design a gated aggregation module to fuse different types of cues into a unified feature. Finally, an adaptation process is proposed to adapt a certain modality to specific historical observations and generate fine-grained prediction results. Extensive experiments on four widely-used benchmarks show the superiority of our proposed approach.

Exploring the sentiment in Borsa Istanbul with deep learning

Sentiment analysis holds immense importance in finance and economics, addressing crucial issues such as principal–agent dynamics and information imbalances. The rise of natural language processing signifies a groundbreaking era in sentiment analysis, enabling the effective extraction of insights from textual data. Our research investigates the impact of qualitative financial data on firm valuation, utilizing sentiment extracted from annual financial disclosures, focusing on companies listed on the Borsa Istanbul Stock Exchange from 1998 to 2022. Employing a pre-trained transformer model, we develop sentiment indices and integrate textual data using a system-generalized method of moments. Our study aims to uncover how sentiment expressed in financial disclosures aids in mitigating challenges related to asymmetric information.

Towards explainable motion prediction using heterogeneous graph representations

Motion prediction systems play a crucial role in enabling autonomous vehicles to navigate safely and efficiently in complex traffic scenarios. Graph Neural Network (GNN)-based approaches have emerged as a promising solution for capturing interactions among dynamic agents and static objects. However, they often lack transparency, interpretability and explainability — qualities that are essential for building trust in autonomous driving systems. In this work, we address this challenge by presenting a comprehensive approach to enhance the explainability of graph-based motion prediction systems. We introduce the Explainable Heterogeneous Graph-based Policy (XHGP) model based on an heterogeneous graph representation of the traffic scene and lane-graph traversals. Distinct from other graph-based models, XHGP leverages object-level and type-level attention mechanisms to learn interaction behaviors, providing information about the importance of agents and interactions in the scene. In addition, capitalizing on XHGP’s architecture, we investigate the explanations provided by the GNNExplainer and apply counterfactual reasoning to analyze the sensitivity of the model to modifications of the input data. This includes masking scene elements, altering trajectories, and adding or removing dynamic agents. Our proposal advances towards achieving reliable and explainable motion prediction systems, addressing the concerns of users, developers and regulatory agencies alike. The insights gained from our explainability analysis contribute to a better understanding of the relationships between dynamic and static elements in traffic scenarios, facilitating the interpretation of the results, as well as the correction of possible errors in motion prediction models, and thus contributing to the development of trustworthy motion prediction systems.The code to reproduce this work is publicly available at https://github.com/sancarlim/Explainable-MP/tree/v1.1.

Highly transparent, durable, and omniphobic liquid-like coatings for efficient dynamic de-wetting and self-cleaning applications

Solar cells, which are contributing significantly to achieving carbon neutrality but suffer from the dust deposition problem that limits their power conversion efficiency. Despite the many desirable properties of self-cleaning coatings to improve the efficiency of solar cells, one formidable challenge in designing coatings for their practical applications lies in integrating transparency, durability, and self-cleaning properties. Herein, we report a multifunctional, durable, and omniphobic liquid-like FEVE-g-PDMS coating, which was readily prepared by embedding highly flexible polydimethylsiloxane (PDMS) chains as dynamic de-wetting agent into the fluorinated ethylene-(hydroxyl-alkyl) vinyl ether (FEVE) matrix. The resultant coating exhibited low roughness and high transparency (89.72 % at 550 nm, the air was used as reference), which could be applied to the protective solar cover glass by facile spraying technique. The PDMS chains with high mobility were enriched on the coating surface and could dynamically rotate/bend/stretch to reduce the energy barriers for the moving liquids, endowing coating with unprecedented dynamic de-wetting property. Importantly, the coating demonstrated self-cleaning property, which could readily slide the liquid and dust contaminants off, resulting in a high efficiency recovery ratio for solar cells (95.37 %). Besides, the coating showed strong adhesion, and durability to ultraviolet rays, humidity and thermal atmospheres, and chemical media. Therefore, the synthetic omniphobic liquid-like coating holds great promise for the development of innovative self-cleaning materials with versatility and durability that are not realizable with conventional solid interface engineering.

Deterministic correlations enhance synchronization in oscillator populations with heterogeneous coupling.

Synchronization is a critical phenomenon that displays a pivotal role in a wealth of dynamical processes ranging from natural to artificial systems. Here, we untangle the synchronization optimization in a system of globally coupled phase oscillators incorporating heterogeneous interactions encoded by the deterministic-random coupling. We uncover that, within the given restriction, the added deterministic correlations can profoundly enhance the synchronizability in comparison with the uncorrelated scenario. The critical points manifesting the onset of synchronization and desynchronization transitions, as well as the level of phase coherence, are significantly shaped by the increment of deterministic correlations. In particular, we provide an analytical treatment to properly ground the mechanism underlying synchronization enhancement and substantiate that the analytical predictions are in fair agreement with the numerical simulations. This study is a step forward in highlighting the importance of heterogeneous coupling among dynamical agents, which provides insights for control strategies of synchronization in complex systems.

Integrated investment for improved regional competitiveness of Slovak Regions: Human Agency and Regional Innovation

As regions emerge as critical drivers of economic growth and development, understanding the dynamics of these economic agents becomes imperative for policymakers and stakeholders seeking to enhance regional prosperity. This research delves into the potential roles of existing economic agents/tools, including actors and agencies, for integrated investment proposals. The research is about how to plan integrated investments for improved regional competitiveness for Slovak self-governing regions by considering the soft factors for regional competitiveness with a special focus on underdeveloped regions including regions affected by attenuation of mining activities and negative related side effects. Through a mixed-method approach, our investigation synthesizes relevant literature on regional innovation and competitiveness; we analyze the intricate interactions and alignments of actors and agencies, exploring how their dynamics contribute to the region's ability to attract and leverage investments for promoting innovation and competitiveness. This research contributes to the ongoing dialogue on effective strategies for regional economic development and offers actionable recommendations for policymakers, businesses, and stakeholders to benefit from the potential. Ultimately, our investigation seeks to empower decision-makers with the knowledge needed to place-based/place-sensitive policies and initiatives that facilitate innovation, foster competitiveness, and drive regional success to new dimensions.

TransFuser: Imitation With Transformer-Based Sensor Fusion for Autonomous Driving.

How should we integrate representations from complementary sensors for autonomous driving? Geometry-based fusion has shown promise for perception (e.g., object detection, motion forecasting). However, in the context of end-to-end driving, we find that imitation learning based on existing sensor fusion methods underperforms in complex driving scenarios with a high density of dynamic agents. Therefore, we propose TransFuser, a mechanism to integrate image and LiDAR representations using self-attention. Our approach uses transformer modules at multiple resolutions to fuse perspective view and bird's eye view feature maps. We experimentally validate its efficacy on a challenging new benchmark with long routes and dense traffic, as well as the official leaderboard of the CARLA urban driving simulator. At the time of submission, TransFuser outperforms all prior work on the CARLA leaderboard in terms of driving score by a large margin. Compared to geometry-based fusion, TransFuser reduces the average collisions per kilometer by 48%.