In the multi-agent systems with switching topology, to address the consistency problem weaken by the high and low density information, a Layered Neighbor Selection(LNS) algorithm is proposed. First, this algorithm divides the neighborhood of the agent into layers, and selects the representative neighbor agents from each layer for communication, state update and state evolution. Then, it designs a layer adjustment strategy and a layer fusion strategy to accelerate the convergence speed, Finally, it designs a layer neighborhood selection consistency protocol, and provides the influence of the number of layers on the convergence. The convergence effect of the traditional consistency protocol is limited to a specific density range and is greatly affected by different densities. Differently, the proposed protocol of this paper can adapt to different density ranges and improve the convergence speed under the condition of system stability. The stability of the consistency protocol is proved by the Lyapunov function method. The effectiveness of the proposed consistency protocol is verified by simulation in comparison with several consistency protocols.

Consensus in multi-agent systems is a key problem in the research of multi-agent cooperative control. Traditional studies on consensus do not consider the overlapping structure in the network topology, neglecting the strength of connections between nodes, resulting in weaker connections being more prone to disconnection during system evolution, thereby affecting system consistency performance. Based on the idea that overlapping nodes in community networks can promote information interconnection between different communities, this paper proposed an overlapping community model (OCM). Firstly, this paper proposed a distributed overlapping node discovery algorithm (DOND) algorithm, which can be used to identify overlapping nodes in all nodes' neighborhood in the system. Secondly, this paper proposed an overlap degree-based topology reweighting algorithm (ODTR) algorithm to quantify the overlap degree between nodes and dynamically assign weights. Finally, this paper proposed an overlapping community model-based multi-agent consensus protocol. The stability of the system is verified by theoretical analysis, and the consensus protocol proposed in this paper is simulated. The experimental results show that the protocol can enhance the system consensus by reducing the number of convergence clusters.

In the mixed traffic environment where Connected Autonomous Vehicles (CAVs) and Human Driving Vehicles (HDV) coexist, the Highway On-Ramp Merging Problem presents challenges. The road contention issue involving different types of vehicles usually impacts traffic flow. Vehicles may contend for positions on the road, including merging, lane changing, and other behaviors, leading to the challenges of accurately predicting and adapting to their actions for CAVs. This increases the risk of merging, resulting in decreased traffic efficiency and traffic congestion. Traditional reinforcement learning algorithms have difficulty in effectively searching for optimal strategies in complex environments, and they are prone to getting stuck in local optima. They are unable to effectively deal with complex traffic situations, leading to imprecise merging decisions. To address these challenges, the Evolutionary Soft Actor-Critic for Discrete Action Settings (ESACD) algorithm is proposed. It maximizes the traffic throughput by adaptively coordinating CAVs to HDV strategies. Firstly, a Rank Selection-based Parent Selection and Crossover Method is introduced to model the interaction population. Secondly, a Multiple Populations with Elastic Training method is designed to enhance CAV adaptability to the changes of the dynamic traffic flow. Finally, a Fitness Evaluation-based Secondary Assessment Mechanism is proposed. Simulation experiments conducted under two different traffic densities demonstrate that the proposed algorithm more efficiently completes the merging task at highway on-ramps for connected vehicles with a significant overall improvement rate when compared with the traditional Soft Actor-Critic (SAC) algorithm. This validates the training efficiency of the proposed algorithm with expanding the traffic throughput.

To counter the risk of eavesdropper attacks inherent in traditional tracking control methods, research on tracking control in multi-agent systems must incorporate privacy protection considerations. This study introduces an event-triggered distributed differential privacy protection tracking control (EDPTC) method in multi-agent systems featuring both leaders and followers, aimed at addressing privacy protection issues in tracking control. The method is designed to ensure the privacy of the states of leaders and followers at all times while achieving mean square tracking. Given the differences in state updates between leaders and followers, privacy protection mechanisms tailored to each party are developed based on the sensitivity upper bound theorem: the decreasing noise mechanism for followers (DNMF) and the random noise mechanism for leaders (RNML) . Moreover, to minimize the impact of random noise on control performance, a leader state estimation algorithm is introduced, and a distributed event trigger is designed to reduce the communication frequency. Additionally, through matrix analysis and probability theory, the privacy of the EDPTC is proven, and sufficient conditions for achieving mean square tracking are derived. Finally, a series of numerical simulations validate the effectiveness of the proposed method.