Distributed Formation Control Driven by Recursive Equilibrium Networks Under Communication Attacks

Guaranteeing stability for multi-agent formationcontrol under communication constraints and network attacksremains an open challenge. This paper proposes a distributedself-supervised control architecture based on Recursive Bal\u0002anced Networks that ensures exponential stability relying solelyon local neighbor information through contraction theory.The RBN controller employs a parameterized design to au\u0002tomatically satisfy Lyapunov conditions, with computationalcomplexity as low as O(lq\u00b2). Against combined DoS attacks,a developed difficulty-prioritized adversarial training strategymaintains input-to-state stability even when communicationquality deteriorates to 40%. The heterogeneous communicationmechanism based on effective capacity equivalence improvescritical link reliability by 6%. Simulation and hardware exper\u0002iments demonstrate that in 150-timestep trajectory trackingtasks under unknown high-intensity DoS attacks, the multi\u0002UAV formation achieved zero collisions with 0.4-meter averagepositioning accuracy, validating the method\u2019s practical appli\u0002cability in adversarial environments.