基于领航–跟随策略的多智能体协同目标定位

Leader-follower-based cooperative target localization for multi-agent systems

  • 摘要: 本文针对平面内多智能体系统协同目标定位问题,提出了一种基于分布式目标估计的领航–跟随控制策略. 现有方法多依赖复杂的几何分析或精确的距离测量,限制了其在复杂环境下的应用. 基于此,本文设计了一种新型分布式目标估计器,适用于任意部署的静止智能体与目标. 不失一般性,假设两个领航者能观测到目标,且它们与目标三者不共线. 虽然该假设隐含地确定了目标的位置,但该位置对于所有智能体均是未知的. 与常见的协同控制器不同,目标估计器不仅需要各个智能体的目标估计值一致,还必须收敛到目标位置. 因此,本文提出的估计器通过引入投影矩阵,将领航智能体获取的局部方位角信息嵌入分布式估计算法. 该操作使得领航者对目标的估计沿视线方向收敛于目标位置. 基于此,系统无需进行任何距离测量,仅利用两个方位角即可驱动所有智能体对目标位置的估计实现指数收敛,从而实现了简单信息条件下的全局定位. 此外,投影矩阵因其半正定的特性增加了理论分析的困难,缩放因子的引入保证了协同控制算法的收敛性,并基于缩放因子提供了量化参数,在一定程度上量化了协同控制算法的收敛速度. 最后,本文给出了数值仿真结果,以验证所提出的基于领航-跟随的分布式协同控制策略的有效性.

     

    Abstract: This study addresses the problem of cooperative target localization for stationary multi-agent systems, aiming to localize a stationary target within a specifically designed planar environment by proposing a novel leader-follower-based distributed cooperative control strategy that does not rely on distance measurements. Existing approaches to the cooperative target localization problem frequently rely on the geometric relationships between adjacent agents and the target, or the measurement distances from some agents to the target, to design cooperative pointing controllers. However, these methods generally require complex geometric analyses or precise distance measurements, which inherently limit their applicability within complex environments. To overcome the limitations of complex geometric analyses and precise distance measurements, a novel distributed target estimator was designed in this study. This estimator requires only two leader agents to provide the direction angle and scaling factor of the observed target, without the need for direct distance measurement, and it is suitable for any deployed stationary agent or stationary target. In this setup, each agent possesses knowledge of its own global position, and its deployment is arbitrary. The designed distributed target estimator requires only directional information (orientation angles) of the target, as perceived by the two designated leader agents, along with a carefully chosen scaling factor, eliminating the need for other distance measurements related to the target. Without loss of generality, it is assumed that the two leader agents measure the target's orientation angles, and that the three entities –two leader agents and the target – are not collinear. Although this assumption implicitly determines the target's position, the spatial information regarding the target remains unknown to the leader agents and other agents. Unlike common cooperative controllers, the target estimator imposes a dual requirement: not only must the individual target estimates from each agent converge to a consistent value, but this consistent estimate must also asymptotically approach the target’s actual location. Therefore, the proposed estimator embeds the local orientation angle information acquired by the leader agents into a distributed estimation algorithm by introducing a projection matrix. This operation ensures that the leader’s estimates of the target converge along the line of sight toward the target position. Thus, the system eliminates the need for distance measurement. Only two orientation angles are utilized to facilitate the exponential convergence of the estimates of all agents for the target position, thereby achieving global localization under simple information conditions. Furthermore, the projection matrix, owing to its positive semi-definite property, increases the difficulty and complexity of theoretical analysis. The introduction of a scaling factor ensures convergence of the cooperative control algorithm and provides quantifiable parameters based on this scaling factor, thereby quantifying its convergence speed of the cooperative control algorithm to a certain extent. Theoretical analysis has proven that followers gradually approach and converge on the target position through local information exchange. Finally, the efficacy and performance of the proposed leader-follower-based distributed cooperative control strategy are demonstrated via a numerical simulation of an agent system with 11 agents in a planar environment of sufficient scope.

     

/

返回文章
返回