避障
多智能体系统
计算机科学
障碍物
融合
控制理论(社会学)
钥匙(锁)
控制工程
分布式计算
传感器融合
人工智能
避碰
功能(生物学)
工程类
作者
Weihong Wang,Yan Xie,Sha Fan,Chao Deng,Bohui Wang,Valeriy V. Vyatkin
标识
DOI:10.1109/tcyb.2026.3736680
摘要
This article addresses the problem of distributed formation and collision avoidance for a class of nonlinear multiagent systems (MASs) in cluttered environments. Unlike traditional collision avoidance methods that constrain the feasible workspace by relying on simplified geometric obstacle models, this article achieves a more precise representation of irregular obstacles to improve space utilization. To achieve these improvements, a hierarchical perception-control method consisting of a point-cloud fusion algorithm in the perception layer, distributed command governors in the navigation layer, and adaptive controllers in the control layer is proposed. Specifically, the point-cloud fusion algorithm is designed to extract closed obstacle boundaries by integrating data from LiDAR and RGB-D cameras to ensure that all obstacles are detected and their boundaries are identified with high precision. Based on the extracted boundary information, distributed command governors utilizing a$p$-norm-based metric to construct a smooth and differentiable obstacle model are designed to generate a collision-free reference signal for each agent to follow. Furthermore, adaptive controllers are proposed by using the backstepping method. It is shown that collision avoidance and the preservation of the prescribed formation shape can be achieved by means of the developed method. Numerical simulations and experimental results are presented to verify the effectiveness of the proposed hierarchical control method.
科研通智能强力驱动
Strongly Powered by AbleSci AI