Collision‐free flocking control of uncertain multi‐agent systems: Combining current learning adaptive control and projection operator

Abstract Adaptive flocking control of multi‐agent systems faces challenges in handling uncertainties and ensuring safety. This paper aims to address these issues based on the hypothesis that the uncertain parameters are bounded. First, a concurrent learning adaptive control method relaxes the persistently excitation condition for parameter convergence, enabling adaptability with interval excitation only. Second, an element‐wise projection operator bounds parameter estimates within known intervals, precomputing collision avoidance conditions, and guaranteeing safety. Third, combining with the aforementioned methods, a distributed flocking algorithm incorporates limited sensing range in a moving region, achieving collision avoidance, connectivity, and cohesion via bounded potential functions. LaSalle's invariance principle shows that parameter estimates converge within bounds, collision avoidance conditions hold, and system stability is achieved. Simulations validate enhanced adaptability, guaranteed safety, and the expected cooperative flocking motion. The proposed approach addresses critical challenges for real‐world deployment of swarm technology.