Safe Control of Multiagent Systems via Low-Complexity Control Barrier Functions

• 成果类型: Article
• 署名作者: Min, Xiao; Baldi, Simone; Shi, Yang; Yu, Wenwu
• 署名单位: Nanjing University of Posts & Telecommunications; Southeast University - China; University of Victoria
• 刊物名称: IEEE TRANSACTIONS ON AUTOMATIC CONTROL
• ISSN/ISSBN: 0018-9286
• DOI: 10.1109/TAC.2025.3566652
• 发表日期: 2025
• 页码: 6831-6845
• 关键词: safety; Complexity theory; Multi-agent systems; control systems; uncertainty; systematics; Backstepping; trajectory; quadratic programming; system dynamics; Control barrier function (CBF); funnel control; higher order systems; low-complexity control; multiagent systems (MASs); safety

摘要

In its standard formulation, the control barrier function (CBF) method scales poorly as the order of the system dynamics increases. On the one hand, the need of recursively extending the control with higher-order terms along the backstepping procedure inevitably results in dynamically increasing complexity. On the other hand, the lack of an explicit form of the CBF leads to implicit formulations of the safety conditions, to be solved numerically via quadratic programming. These high complexity and poor scalability issues further amplify in multiagent systems. This work proposes a low-complexity framework for safe control of higher order multiagent systems within the philosophy of funnel control. Different from the state-of-the-art, safety is expressed in terms of a CBF that is explicitly constructed from funnel functions, thus providing a solution to the well-known problem of lack of systematic methods for constructing a CBF. Low complexity comes from the fact that the control does not involve complex dynamical extensions during the backstepping procedure. Notably, the proposed CBF is shown to automatically satisfy by design the safety conditions, without resorting to any quadratic programming. Lyapunov analysis shows that the design allows to consider individual terms for stability and safety, so as to accomplish these tasks in a seamless integrated fashion. Simulation studies and comparisons with the state-of-the-art further demonstrate the method.