Generating collision-free, time-efficient paths in an uncertain dynamic environment poses huge challenges for the formation control with collision avoidance (FCCA) problem a leader-follower structure. In particular, followers have to take both maintenance and into account simultaneously. Unfortunately, most of existing works are simple combinations methods dealing two problems separately. this article, new method based on deep reinforcement learning (RL) is proposed solve FCCA. Especially,...

In this article, a novel method, called attention enhanced reinforcement learning (AERL), is proposed to address issues including complex interaction, limited communication range, and time-varying topology for multi agent cooperation. AERL includes network (CEN), graph spatiotemporal long short-term memory (GST-LSTM), parameters sharing multi-pseudo critic proximal policy optimization (PS-MPC-PPO). Specifically, CEN based on mechanism designed enlarge the agents' range deal with interaction...

Spatial reasoning is an essential problem in embodied AI research. Efforts to enhance spatial abilities through supplementary data and fine-tuning have proven limited ineffective when addressing complex tasks, largely due their dependence on language-based outputs. While some approaches introduced a point-based action space mitigate this issue, they fall short managing more intricate tasks within environments. This deficiency arises from failure fully exploit the inherent thinking...

Deriving a distributed, time-efficient, and connectivity-guaranteed coverage policy in multitarget environment poses huge challenges for multirobot team with limited communication. In particular, the robot needs to cover multiple targets while preserving connectivity. this article, novel deep-reinforcement-learning-based approach is proposed take both connectivity preservation into account simultaneously, which consists of four parts: hierarchical observation attention representation, an...

Making efficient motion decisions for a multi-robot system is challenging problem in target encirclement with collision avoidance. Specifically, each robot local communication has to consider cooperative and avoidance simultaneously. In this paper, distributed transferable policy network framework based on deep reinforcement learning proposed solve the of The able process information uncertain number robots obstacles, which desirable property systems. particular, graph attention mechanism...

When dealing with a series of imminent issues, humans can naturally concentrate on subset these concerning issues by prioritizing them according to their contributions motivational indices, e.g., the probability winning game. This idea concentration offers insights into reinforcement learning sophisticated Large-scale Multi-Agent Systems (LMAS) participated hundreds agents. In such an LMAS, each agent receives long entity observations at step, which overwhelm existing aggregation networks as...

This paper considers a multi-target coverage problem where robot team aims to efficiently cover multi-targets while maintaining connectivity in distributed manner. A novel knowledge-incorporated policy framework is proposed derive distributed, efficient, and guaranteed policy. In particular, knowledge-guided network (KGPnet) designed, which consists of observation attention representation, interaction learning. Giving credit the KGPnet, can be applied different number targets. Moreover,...

Collective behavior of multi-agent systems brings some new problems in control theory and application. Especially, flocking problem with uncertain nonlinear dynamics unknown external disturbances is a challenging problem. Some existing works assume that the intrinsic virtual leader same as those agents, which unreasonable impractical. To solve this issue, we consider an adaptive paper, where allowed to be different from agents. Firstly, approximate each agent, neural network used, whose...

This paper proposes a collaborative policy framework via relational graph reasoning for multi-agent systems to accomplish adversarial tasks. A module consisting of an agent and opponent module, is designed enable each learn mixture state representation enhance the effectiveness policy. In particular, agent, infer different underlying influences from opponents generate agent-level representation. The creatively reason relations their surrounding objects including agents based on latent...

The formation control of multi-robot systems in complex environments is a challenging problem, which needs to handle velocity estimation and obstacle avoidance problems. However, few researches considered these problems simultaneously the literature. In this paper, leader-follower hybrid design for system proposed, solves problem. First, based on architecture, law designed. Second, considering difficulty obtaining whole accurate information, estimator leader designed, Lyapunov function used...

The multi-robot coverage path planning (CPP) is the design of optimal motion sequence robots, which can make robots execute task covering all positions work area except obstacles. In this article, communication capability system applied, and a CPP mechanism proposed to control perform tasks in an unknown environment. mechanism, algorithm based on deep reinforcement learning proposed, generate next action for real-time according current state robots. addition, obstacle avoidance scheme...

Meta-learning enables learning systems to adapt quickly new tasks, similar humans. To emulate this human-like rapid and enhance alignment discrimination abilities, we propose ConML, a universal meta-learning framework that can be applied various algorithms without relying on specific model architectures nor target models. The core of ConML is task-level contrastive learning, which extends from the representation space in unsupervised meta-learning. By leveraging task identity as an...

Inspired by psychological insights into individual behavior, we propose a novel cognition-oriented multiagent reinforcement learning (CORL) framework. CORL equips agents with two distinct types of cognition-situational and self-cognition-derived from local observations. To enhance the informativeness precision these cognition types, introduce information-theoretical regularizers: one to align situational global state other self-cognition each agent's identity for improved role...

Multi-robot task detection and execution is an important direction that people have considered recently. We the dynamic issues related to tasks of multiple robots. This paper presents improved auction algorithm solve this problem combines it with by optimizing capacity utilization load balancing robot. By designing a new cost function, ability robot are added algorithm, matching difficulty assignment execution. In addition, we also add sequence adjustment mechanism avoid redundant loss due...

In this paper, we propose a novel Intrinsic Reward method with Peer Incentives (IRPI) to promote the inter-agent direct interactions and implicitly address credit assignment problem in cooperative multi-agent reinforcement learning (MARL). The IRPI can build mutual incentives between agents by using their causal effect, realize advanced cooperation. Specifically, new intrinsic reward mechanism is conducted, which equips each agent ability other effect them. Moreover, built through neural...

Recently, multiagent reinforcement learning (MARL) has shown great potential for cooperative policies in systems (MASs). However, a noticeable drawback of current MARL is the low sample efficiency, which causes huge amount interactions with environment. Such greatly hinders real-world application MARL. Fortunately, effectively incorporating experience knowledge can assist to quickly find effective solutions, significantly alleviate drawback. In this article, novel multiexperience-assisted...

Multimodal recommendation system has been widely used in short video platform, e-commerce platform and news media. data contains information such as product image text, which is often auxiliary signal to improve the effect of significantly. In order alleviate problems sparsity noise, some researchers construct augmentation use self-supervised learning help model training. These methods have achieved certain results. However, most work based on random ways, masking perturbation. This method...

In ball sports, such as basketball, the coach can guide players to better offend and defend from a holistic perspective win game. Inspired by scenarios, we introduce coach-like concept into decision-making process of cooperative multi-agent systems. We propose new framework Commander-Soldiers Reinforcement Learning (CSRL), for Multi-Agent Specifically, virtual role, Commander, which obtain encode global information every T steps send encoded guidance Soldiers (real agents). Furthermore,...

Learning an effective strategy is challenging for agents in partially observable environment, where the can only observe a part of environment information and make decisions based on local information. Hence, how to effectively utilize achieve efficient cooperation among particularly important. The establish understanding themselves their surrounding historical observation However, lacking global or limited, resulting low performance some complex tasks. To solve problem, situational...

A mobile robot network is a system composed of multiple robots with simple functions through communication and collaboration. To achieve coordination cooperation between robots, designing an efficient mechanism key issue. However, due to the characteristics autonomous movement harsh environment in which perform tasks, facing difficulties such as exhaustion node power, damage, link failure. In order solve these problems, we design intelligent routing for networks based on capabilities aims...

Robots are now essential in unreachable, repeated, and dangerous real-world applications where they take place of human beings. One the important capabilities a multi-robot system is that it should be able to form an autonomous robot network transmit information. However, due limited communication capability single robot, highly variable environment robots work, mobility robots, difficult for them exchange information with each other need. In this article, we propose novel routing protocol...