Explore the vast range of titles available.

Multi-agent perceptual map construction and long-term maintenance constitute an important paradigm for improving adaptability and real-world applicability. With the outstanding capability of 3D Gaussian Splatting in preserving fine-grained texture details, a number of 3DGS-based real-time mapping approaches have recently emerged. However, these methods often struggle to cope with complex dynamics in real-world environments and lack the generalization needed to scale to multi-agent systems. Existing solutions typically rely on direct parameter concatenation or locally confined optimization, which are unable to explicitly model cross-agent observation reliability under temporal asynchrony and dynamic inconsistency, and therefore tend to amplify conflicting updates rather than resolve them. To address these limitations, we propose DGOMapping, an online system for multi-agent dynamic perceptual mapping. DGOMapping leverages an uncertainty-coupled 4DGS scene representation and a collaborative interaction mechanism via Gaussian perception-score exchange, enabling both real-time 4DGS construction and long-term map memory adjustment. Experiments on multiple real-world datasets demonstrate that DGOMapping effectively suppresses dynamic interference and exploits multi-agent collaboration, achieving state-of-the-art performance in both tracking and reconstruction. The proposed system therefore provides a practical sensing-oriented solution for collaborative perception and real-time dynamic environment mapping.

This research presents an intelligent, environment-aware navigation framework for smart electric vehicles (EVs), focusing on multi-agent mapping, real-time obstacle recognition, and adaptive route optimization. Unlike traditional navigation systems that primarily minimize cost and distance, this research emphasizes how EVs perceive, map, and interact with their surroundings. Using a distributed mapping approach, multiple EVs collaboratively construct a topological representation of their environment, enhancing spatial awareness and adaptive path planning. Neural Radiance Fields (NeRFs) and machine learning models are employed to improve situational awareness, reduce positional tracking errors, and increase mapping accuracy by integrating real-time traffic conditions, battery levels, and environmental constraints. The system intelligently balances delivery speed and energy efficiency by dynamically adjusting routes based on urgency, congestion, and battery constraints. When rapid deliveries are required, the algorithm prioritizes faster routes, whereas, for flexible schedules, it optimizes energy conservation. This dynamic decision making ensures optimal fleet performance by minimizing energy waste and reducing emissions. The framework further enhances sustainability by integrating an adaptive optimization model that continuously refines EV paths in response to real-time changes in traffic flow and charging station availability. By seamlessly combining real-time route adaptation with energy-efficient decision making, the proposed system supports scalable and sustainable EV fleet operations. The ability to dynamically optimize travel paths ensures minimal energy consumption while maintaining high operational efficiency. Experimental validation confirms that this approach not only improves EV navigation and obstacle avoidance but also significantly contributes to reducing emissions and enhancing the long-term viability of smart EV fleets in rapidly changing environments.

This article introduces multi-agent strategy enabling multiple agents to explore an unknown environment with many obstacles, while generating a topological map in a cooperative way. Once a topological map is built, it can be used for various purposes, such as path planning or intruder capture. Multiple agents generate a Voronoi graph as a topological map of the environment, while dropping indexed markers at Voronoi vertices. Each agent has range sensors to detect nearby obstacles, thus can move along a Voronoi edge. Also, each agent stores the boundary for the explored region thus far, and unite its boundary with the boundary of another agent if some conditions are met. In this way, multiple agents can explore the entire workspace in a time efficient manner. The proposed exploration strategy doesn’t require localization of an agent or a marker in global coordinate systems. To the best of our knowledge, this article is unique in addressing a multi-agent exploration and map building strategy, such that each agent drops indexed markers for generating a topological map of the environment. The effectiveness of the proposed exploration and mapping strategy is demonstrated utilizing MATLAB simulations.

Reinforcement learning (RL), as an emerging interdisciplinary field formed by the integration of artificial intelligence and control science, is currently demonstrating a cross-disciplinary development trend led by artificial intelligence and has become a research hotspot in the field of optimal control. This paper systematically reviews the development context of RL, focusing on the intrinsic connection between single-agent reinforcement learning (SARL) and multi-agent reinforcement learning (MARL). Firstly, starting from the formation and development of RL, it elaborates on the similarities and differences between RL and other learning paradigms in machine learning, and briefly introduces the main branches of current RL. Then, with the basic knowledge and core ideas of SARL as the basic framework, and expanding to multi-agent system (MAS) collaborative control, it explores the coherence characteristics of the two in theoretical frameworks and algorithm design. On this basis, this paper reconfigures SARL algorithms into dynamic programming, value function decomposition and policy gradient (PG) type, and abstracts MARL algorithms into four paradigms: behavior analysis, centralized learning, communication learning and collaborative learning, thus establishing an algorithm mapping relationship from single-agent to multi-agent scenarios. This innovative framework provides a new perspective for understanding the evolutionary correlation of the two methods, and also discusses the challenges and solution ideas of MARL in solving large-scale MAS problems. This paper aims to provide a reference for researchers in this field, and to promote the development of cooperative control and optimization methods for MAS as well as the advancement of related application research.

The growing complexity of production systems requires appropriate control architectures that allow flexible adaptation during their runtime. Although cyber-physical production systems (CPPS) provide the means to cope with complexity and flexibility, the migration with existing control systems is still a challenge. The term CPPS denotes a mechatronic system (physical world) coupled with software entities and digital information (cyber part), both enabling the smart factory concept for the Industry 4.0 (I4.0) paradigm. In this regard, design patterns could help developers to build their software with common solutions for manufacturing control derived from experiences. We provide a description and comparison of the already existing multi-agent systems (MAS) design patterns, which were collected and classified by introducing two classification criteria to support MAS developers. The applicability of these criteria is shown in the case of specific example architectures from the lower and higher control levels. The authors, together with experts from the German Agent Systems committee FA 5.15, gathered more than twenty MAS patterns, evaluated, and compared four selected patterns with the presented criteria and terminology. The main contribution is a CPPS architecture that fulfills requirements related to the era of smart factories, as well as the Reference Architectural Model I4.0 (RAMI 4.0). The conclusions indicate that agent-based patterns greatly benefit the CPPS design. In addition, it is shown that manufacturing based on MAS is a good way to address complex requests of the CPPS development.

Mobile robot path planning involves designing optimal routes from starting points to destinations within specific environmental conditions. Even though there are well-established autonomous navigation solutions, it is worth noting that comprehensive, systematically differentiated examinations of the critical technologies underpinning both single-robot and multi-robot path planning are notably scarce. These technologies encompass aspects such as environmental modeling, criteria for evaluating path quality, the techniques employed in path planning and so on. This paper presents a thorough exploration of techniques within the realm of mobile robot path planning. Initially, we provide an overview of eight diverse methods for mapping, each mirroring the varying levels of abstraction that robots employ to interpret their surroundings. Furthermore, we furnish open-source map datasets suited for both Single-Agent Path Planning (SAPF) and Multi-Agent Path Planning (MAPF) scenarios, accompanied by an analysis of prevalent evaluation metrics for path planning. Subsequently, focusing on the distinctive features of SAPF algorithms, we categorize them into three classes: classical algorithms, intelligent optimization algorithms, and artificial intelligence algorithms. Within the classical algorithms category, we introduce graph search algorithms, random sampling algorithms, and potential field algorithms. In the intelligent optimization algorithms domain, we introduce ant colony optimization, particle swarm optimization, and genetic algorithms. Within the domain of artificial intelligence algorithms, we discuss neural network algorithms and fuzzy logic algorithms. Following this, we delve into the different approaches to MAPF planning, examining centralized planning which emphasizes decoupling conflicts, and distributed planning which prioritizes task execution. Based on these categorizations, we comprehensively compare the characteristics and applicability of both SAPF and MAPF algorithms, while highlighting the challenges that this field is currently grappling with.

In multi-agent path finding, coordinates reflect the positions of agents in the environment and play a crucial role in preventing collisions between them. A key issue that needs to be addressed is how to establish efficient communication and learning strategies between agents to share this coordinate information. However, current methods face problems such as information fusion loss and redundant shared information. Therefore, this paper proposes CoMP-LG, a multi-agent path finding method based on learnable communication graphs. First, the SCDCN network is used to extract feature information from the agents’ observations, and SCA is applied to enhance the coordinate information within these observation features, reducing redundancy in the spatial dimensions while improving the representational ability of the coordinate features. Next, the communication problem between agents is conceptualized as a learnable graph, and the transformer is used to optimize this graph to reduce communication redundancy. Finally, this communication graph is utilized to enable information sharing between agents, thereby reducing uncertainty in local decisions during multi-agent path finding. Empirical results demonstrate that CoMP-LG achieves strong performance in both accuracy and FlowTime across maps of varying sizes and numbers of agents.

Agent-based modelling (ABM) is becoming a widely explored method for investigating human–water systems, given its ability to represent heterogeneous actors and their decisions. ABM can simulate how humans interact and co-adapt with their environment, which is beneficial for understanding the effects of humans’ decisions in the face of hazards and climate change. ABMs can serve as tools for examining the effects of current and future hydrological hazard management strategies. However, the implementation of hydrological hazard management in ABMs has not yet been systematically evaluated for floods and droughts. To map the current status of ABMs in hydrological hazard modelling and facilitate a discussion on further potential, we conducted a systematic mapping review based on the ROSES protocol. In this review, we investigate what kinds of hydrological hazards and management strategies that are represented in ABMs. Additionally, we synthesise current practices regarding agent types and their decision-making. A total of 377 articles were screened, and 77 articles were analysed in full text. Our findings indicate that hydrological hazard management strategies in ABMs include both structural and non-structural measures. However, there is an emphasis on the complexity of individual agents’ decision-making in implementing these measures, whereas collective agents (e.g. governments) performing non-individual hazard management are implemented more simplistically, often as static scenarios or collective agents with ad-hoc or rational decision-making. Conversely, individual agents are commonly implemented with human-like behaviour. Our study highlights that the simplicity of hazard management in these models could restrict the potential of ABMs as policy and predictive tools, as the implemented hazard management does not capture the full dynamics of human–water systems. Involving stakeholders, adopting interdisciplinary methods, or incorporating bounded-rational decision-making could represent a significant shift to further enhance the explanatory power of ABM for addressing challenges in hydrological hazard management.

Collaborative robots represent an evolution in the field of swarm robotics that is pervasive in modern industrial undertakings from manufacturing to exploration. Though there has been much work on path planning for autonomous robots employing floor plans, energy-efficient navigation of autonomous robots in unknown environments is gaining traction. This work presents a novel methodology of low-overhead collaborative sensing, run-time mapping and localization, and navigation for robot swarms. The aim is to optimize energy consumption for the swarm as a whole rather than individual robots. An energy- and information-aware management algorithm is proposed to optimize the time and energy required for a swarm of autonomous robots to move from a launch area to the predefined destination. This is achieved by modifying the classical Partial Swarm SLAM technique, whereby sections of objects discovered by different members of the swarm are stitched together and broadcast to members of the swarm. Thus, a follower can find the shortest path to the destination while avoiding even far away obstacles in an efficient manner. The proposed algorithm reduces the energy consumption of the swarm as a whole due to the fact that the leading robots sense and discover respective optimal paths and share their discoveries with the followers. The simulation results show that the robots effectively re-optimized the previous solution while sharing necessary information within the swarm. Furthermore, the efficiency of the proposed scheme is shown via comparative results, i.e., reducing traveling distance by 13% for individual robots and up to 11% for the swarm as a whole in the performed experiments.

Natural disasters cause significant losses. Unmanned aerial vehicles (UAVs) are valuable in rescue missions but need to offload tasks to edge servers due to their limited computing power and battery life. This study proposes a task offloading decision algorithm called the multi-agent deep deterministic policy gradient with cooperation and experience replay (CER-MADDPG), which is based on multi-agent reinforcement learning for UAV computation offloading. CER-MADDPG emphasizes collaboration between UAVs and uses historical UAV experiences to classify and obtain optimal strategies. It enables collaboration among edge devices through the design of the ’critic’ network. Additionally, by defining good and bad experiences for UAVs, experiences are classified into two separate buffers, allowing UAVs to learn from them, seek benefits, avoid harm, and reduce system overhead. The performance of CER-MADDPG was verified through simulations in two aspects. First, the influence of key hyperparameters on performance was examined, and the optimal values were determined. Second, CER-MADDPG was compared with other baseline algorithms. The results show that compared with MADDPG and stochastic game-based resource allocation with prioritized experience replay, CER-MADDPG achieves the lowest system overhead and superior stability and scalability.