Direct, indirect, and network reciprocities are established mechanisms that sustain cooperation in natural and artificial systems. Yet which mechanism most effectively promotes cooperation on a given network remains unclear. Here, we develop a game-theoretic model to explore the evolution of direct and indirect reciprocity on weighted networks. Unlike classical donor-recipient frameworks, we study symmetric repeated interactions on undirected weighted networks with bilateral reputation updates, capturing heterogeneous tie strengths and accelerating reputation spread. We derive a general condition for reciprocal cooperation that unifies unweighted and weighted cases. Across large ensembles of random and empirical networks, indirect reciprocity consistently enhances cooperation, whereas stronger interactions sharply lower the benefit-to-cost threshold under direct reciprocity. To test the robustness of these insights, we examine competition among six reciprocity strategies and find that indirect reciprocity dominates. Our findings demonstrate that choosing the right reciprocity can promote global cooperation on social networks.

Multi-vehicle–UAV collaborative inspection systems face critical challenges in complex environments, where spatiotemporal node conflicts, coupled scheduling, and congestion severely affect operational efficiency. To address these issues, we define the Multi-UAV–Multi-Vehicle Collaborative Inspection Vehicle Routing Problem in a Bi-Layer Road Network (MUMV-CIVRP-BLRN). The model captures realistic inspection scenarios through (i) spatiotemporal node-conflict constraints that prevent simultaneous vehicle access to occupied nodes and (ii) flexible UAV operations, enabling chained multi-task sorties and cross-vehicle recovery. To solve this NP-hard problem, we propose a Multi-Strategy Improved Adaptive Large Neighborhood Search (M-SIALNS) algorithm. Beyond standard ALNS frameworks, M-SIALNS incorporates a cluster-based initialization method, task-chain destroy–repair operators, and air–ground coordination strategies specifically tailored for the bi-layer structure. These strategies enhance global search, solution feasibility, and robustness. Comprehensive experiments on benchmark datasets and a power-grid case study demonstrate the advantages of M-SIALNS. Compared with state-of-the-art algorithms, it reduces inspection duration by 1.6%–11.0%, consistently delivering statistically significant improvements. Ablation and sensitivity analyses confirm the contribution of tailored operators and provide managerial insights into optimal fleet configurations and resource allocation thresholds. This work advances both the theoretical modeling of bi-layer vehicle–UAV routing and its practical deployment in large-scale inspection missions.

This article addresses the collaborative coverage search problem of intelligent multiagent systems in complex environments. The focus is on a decentralized swarm in which agents independently make decisions for exploration and collaboration under constraints of limited sensor range and restricted communication capabilities. To replicate real-world limitations, the environments considered contain complex obstacles that are initially unknown to the swarm. The primary objective is to cover all unknown areas as rapidly as possible while avoiding collisions with obstacles. We propose a heuristically staged multiagent proximal policy optimization searcher (HMPOS) based on the actor–critic (AC) framework, incorporating a two-stage training process with corresponding neural network modules. The first stage emphasizes local information and individual agent behaviors, whereas the second stage refines the previously trained module with a focus on global information and interagent cooperation. The neural network modules are specifically designed to align with this stage-wise training paradigm. Simulation experiments substantiate the feasibility of our approach, and comparative analyses with existing methods demonstrate the efficiency and superiority of our proposed solution.

Multimodal 3D object detection significantly enhances perception by fusing LiDAR point clouds and RGB images. However, existing methods often fail to adaptively estimate modality confidence under challenging conditions such as heavy occlusion or sparse point clouds, leading to degraded fusion performance. In this paper, we propose GraphFusion, a multimodal framework that integrates cross-modal graph modeling with Bayesian uncertainty-aware fusion for robust 3D object detection. Specifically, a heterogeneous graph driven by geometric and semantic cues aligns 3D points with 2D pixels. A Bayesian attention mechanism then leverages predictive uncertainty to dynamically reweight modalities, prioritizing high-confidence information and enabling noise-resilient and spatially adaptive fusion. The proposed module is highly generalizable and can be seamlessly integrated into existing detectors as a plug-and-play component. Extensive experiments on KITTI and nuScenes demonstrate that GraphFusion achieves significant accuracy improvements with superior robustness and generalization, especially in complex environments.

Unravelling the evolution of cooperation is essential for advancing natural and artificial intelligence (AI) systems. Previous studies have investigated the impact of additional incentives, such as reciprocity and reputation, on cooperative behavior. However, a fundamental question persists: under what conditions do moral preferences evolve and does this evolution subsequently promote cooperation in networked populations of agents? To address this question, we propose a comprehensive framework to systematically explore the co-evolution of moral preferences and cooperative behavior in a networked population. In our framework, the population structure is modeled as a network, with nodes corresponding to AI agents. Moral preferences are modeled through a learning algorithm that adheres to social norms. Prosocial and antisocial behaviors lead to rewards or punishments, and learning agents receive morality scores based on their rewarding behavior toward others. Simulation results demonstrate the effectiveness and robustness of the proposed algorithm in a networked population, showcasing faster convergence. We find that moral preferences enhance cooperation as long as the learning rate is moderate, even in the presence of dominant defectors. This surprising finding also holds for cooperation-inhibiting network structures, provided the critical benefit-cost ratio for cooperation is sufficiently high or below average. Interestingly, moral preferences also co-evolve with cooperation in the populations. Our work not only provides new design methodologies for network algorithms, but also highlights the insight that large-scale evolutionary computation can provide for evolutionary biology and emerging AI-agent populations.

Accurate detection of partial discharge (PD) is critical for ensuring the reliability and safety of high-voltage electrical equipment. This study addresses the challenge of distinguishing PD signals from various sources through unsupervised learning. Acoustic emission sensors were employed to collect PD pulse signals, which were analyzed using a novel cross-domain strategy to extract essential features. In this research, we introduce the self-supervised graph contrastive clustering (SGCC) method, combining graph networks with contrastive learning and residual connections to optimize self-supervised learning. This innovative approach enhances the learning of internode relationships and feature differentiation, effectively minimizing the risk of information homogenization. The temporal dynamic threshold negative sampling method accounts for temporal dynamics and diversity. In addition, we develop a feature contrast function to enhance feature independence and reduce information redundancy in high-dimensional embedding vectors. Clustering of PD pulses is efficiently executed using the Bisecting K-Means algorithm. Our experimental results demonstrate that the proposed features, along with the SGCC method, effectively segregate PD sources, thereby providing substantial support for the safety monitoring of high-voltage systems.

This article presents a novel approach to improving multitarget searching in unknown environments using multirobot systems while ensuring adaptability to changing communication conditions. The proposed method addresses challenges arising from limited scope, dynamic circumstances, and inaccurate decision data due to communication disruptions or interference in real-world scenarios. A comprehensive environmental map is generated using a grid-based mapping methodology, encompassing data related to obstacles, coverage, target occupancy, and communication conditions. Considering the constraints imposed by communication conditions, we develop the adaptive communication condition hierarchical distributed model predictive control framework. This framework incorporates a hierarchical communication strategy for multirobot target search. To assess the effectiveness of our approach, a series of comparative experiments are conducted on three distinct maps, each characterized by unique communication environments, obstacle layouts, and target distributions. These experiments employ four commonly used swarm intelligence algorithms. The research findings indicate that implementing the proposed search framework and communication strategy significantly reduces the time and communication costs associated with locating targets in complex and unfamiliar environments. This is particularly relevant for multirobot systems operating under diverse and limited communication conditions, substantially increasing the task’s success rate.

Reciprocity is a potent mechanism that drives cooperation within human societies. Direct reciprocity involves interacting with individuals, while indirect reciprocity is reliant on anticipated rewards from a third party rather than someone they have interacted with directly. Despite being closely intertwined, current research isolates these two mechanisms, with further division of indirect reciprocity into positive and negative based on the directionality of daily interactions. To capture the coevolution of direct and indirect reciprocities, this work has developed a simple yet comprehensive model, allowing individuals to choose among these reciprocities arbitrarily. A comprehensive equilibrium analysis discovers that the analogous strategy of negative indirect reciprocity, called negative scoring (NSCO), has a stronger control property than the well-known reciprocity strategies. Utilizing evolutionary simulations, we find that directionality can indeed promote the emergence of indirect reciprocity. Overall, our findings shed light on a novel passageway toward a universal theory of direct and indirect reciprocities.

This article proposes a framework for multidifferent-target search in unknown environments based on swarm intelligence. In this framework, the idea of distributed model predictive control is introduced in the target search method. The use of a hierarchical prediction strategy further improves the robot’s path prediction ability in unknown environments. Compared with swarm intelligence methods—adaptive robotic particle swarm optimization (A-RPSO), improved group explosion strategy (IGES), and other existing works, this strategy significantly improves the multidifferent-target search functionality and the task success rate in unknown complex obstacle environments. Moreover, two effective efforts are then introduced to reduce computational complexity and speed up online decision making. One is to select cooperative individuals based on the line of sight, and the other is to reduce both the frequency of decision making and the amount of data transmitted. A comparison between obstacle-free map experiments and obstacle map experiments confirms the effectiveness of the ideas and methods presented in this article.

Forward-looking sonars (FLS) are widely applied in the search of underwater wrecked objects. Intelligent FLS images object segmentation methods can effectively assist this task. However, the low resolution and complex noise interference of FLS images bring great challenges to segmentation. In this paper, we propose a novel semantic segmentation network with multi-level feature fusion capability, called multi-level attention and atrous pyramid nested U-Net (MAANU-Net). We use nested U-structure as the main framework to fuse multi-level features. In addition, we integrate a newly de-signed attention and atrous pyramid (AA) module between encoder and decoder. The proposed method is verified on the dataset acquired by a remotely operated vehicle equipped with a FLS. Experimental results show that the MAANU-Net can overcome noise interference and accurately segment objects, which outperforms the other state-of-the-art methods.

Forward-looking sonar (FLS) is widely applied in underwater operations, among which the search of underwater crash objects and victims is an incredibly challenging task. An efficient detection method based on deep learning can intelligently detect objects in FLS images, which makes it a reliable tool to replace manual recognition. To achieve this aim, we propose a novel Swin Transformer based anchor-free network (STAFNet), which contains a strong backbone Swin Transformer and a lite head with deformable convolution network (DCN). We employ a ROV equipped with a FLS to acquire dataset including victim, boat and plane model objects. A series of experiments are carried out on this dataset to train and verify the performance of STAFNet. Compared with other state-of-the-art methods, STAFNet significantly overcomes complex noise interference, and achieves the best balance between detection accuracy and inference speed.

This work solves the long-tail and few-shot (LTFS) problems faced concurrently in sonar image classification. Although the popular deep transfer learning (TL) alleviates the few-shot problems, it performs poorly in the tail classes. Moreover, current works involving class rebalancing concepts, e.g., resampling and reweighting, are extensively applied to improve tail class accuracy but reduce head class accuracy. Hence, this article investigates the reason for the nonideal performance of the class rebalancing schemes in TL via an empirical study. Impressively, we discover that while using sonar images, these methods hinder the representation fine-tuning but conditionally promote the classifier fine-tuning. Inspired by our discovery, we introduce a two-stage decoupled training approach for the sonar image classification task and propose a novel multibalanced sampling method. Moreover, based on these two key ideas, we suggest a new pipeline entitled balanced ensemble transfer learning (BETL), which simultaneously overcomes the LTFS problems. Extensive experiments on three sonar image datasets of different sizes and imbalance factors demonstrate that BETL significantly outperforms the existing methods. Moreover, BETL’s effectiveness and portability are verified through several experiments. Our code will be available athttps://github.com/Jorwnpay/TGRS_BETL.

How to understand and control the individual choices and cooperative conflicts in multiagent systems is a cross-over study which edges into multiple disciplines. An interdisciplinary mathematical tool to theoretically study the issue of cooperative dilemma is evolutionary game theory. The assumption of graphs that maps social interactions provides a useful approach for studying the dynamics in gaming systems, where its theoretical analysis is drop behind simulations. Here we perform analysis on game dynamics in a spatially distributed system situating on multiple-community networks. In the context of the three paradigmatic game prototypes (dominating games, coexistence games, and the coordination games), the stability of the equilibrium state emerging from this is provided. Further investigation on multicommunity system reveals that the community structure plays a key role in the stability of the strategy evolution. Besides, our explorations toward the multiple strategies adopted by the population whose interactions include n communities, certify that the payoff is crucial for individuals' dominating superiority over others in the system. The results are expected to provide effective perspective to control the individual choice in large systems, and lend itself to multiple applications in control engineering perspective.

Global warming is becoming a knotty problem that cannot be ignored, making many countries face conflicts of interest and the need for cooperation. Evolutionary game theory provides an effective research framework for this kind of cooperation dilemma. From the perspective of control engineering, the mutual influence of environment and system can be utilized to achieve the purpose of self-adaptive adjustment and temperature control. Here, by combining feedback control and evolutionary game theory, we aim to restrict the global temperature to an expected level. Particularly, the controller is motivated by the threshold public goods game, where whether the threshold is realized or not is directly related to the different benefit states of the system. Results are gained where the temperature and cooperation level can be stabilized at the pre-set level. Moreover, we derive the correlation between the temperature trend and parameter settings, which is an efficient guideline to improve the control performance in global warming problems.

We present a model of changing the intensity of interaction based on the individual behavior to study the iterated prisoner's dilemma game in social networks. In this model, each individual has an assessed score of reputation which is obtained by considering the evaluation level of interactive partners for its present behavior. We focus on the effect of evaluation level on the changing intensity of interaction between individuals. For an individual with good behavior, the higher the evaluation level of its partners for its good behavior, the better its reputation, and the higher the probability of surrounding partners interaction with it. On the contrary, for an individual with bad behavior, the lower the evaluation level of its partners for its bad behavior, the worse its reputation, and the less the probability of surrounding neighbors interaction with it. Simulation results show that this effective mechanism can drastically facilitate the emergence and maintenance of cooperation in the population under a treacherous chip. Interestingly, for a small or moderate treacherous chip, the cooperation level monotonously ascends as the evaluation level increases; however, for a higher treacherous chip, existing an optimal evaluation level, which can result in the best promotion of cooperation. Furthermore, we find better agreement between simulation results and theoretical predictions obtained from an extended pair-approximation method, although there are some tiny deviations. We also show some typical snapshots of the system and investigate the reason for appearance and persistence of cooperation. The results further show the importance of evaluation level of individual behavior in coevolutionary relationships.

Increasingly influential leadership is significant to the cooperation and success of human societies. However, whether and how leaders emerge among evolutionary game players still remain less understood. Here, we study the evolution of potential leaders in the framework of evolutionary game theory, adopting the prisoner's dilemma and snowdrift game as metaphors of cooperation between unrelated individuals. We find that potential leaders can spontaneously emerge from homogeneous populations along with the evolution of cooperation, demonstrated by the result that a minority of agents spread their strategies more successfully than others and guide the population behavior, irrespective of the applied games. In addition, the phenomenon just described can be observed more notably in populations situated on scale free networks, and thus implies the relevance of heterogeneous networks for the possible emergence of leadership in the proposed system. Our results underscore the importance of the study of leadership in the population indulging in evolutionary games.