A distributed adaptive architecture with the nonlinear reference model for safe finite-time control of uncertain multiagent systems

We propose a new distributed adaptive control architecture for finite-time control of uncertain nonlinear multiagent systems. The proposed architecture employs three key components for each agent; a nonlinear reference model, a weight update rule, and an adaptive control signal. Predicated on agent-wise reference model state exchange, an ideal finite-time behaviour of overall multiagent system is captured by nonlinear reference models. The weight update rule of each agent, which is driven by a local error signal between the actual uncertain state of an agent and its reference model state, then adjusts agent-wise controller parameters in real-time to drive the local error signals of each agent to zero in finite-time. That is, not only the states of agents converge to their nonlinear reference model states in finite-time, but also the latter states converge to the given ideal behaviour in finite-time. Considering safety, the distinct feature of our architecture is that it does not rely on agent-wise actual state exchange between agents, which involves the effect of system uncertainties. This implies that when a subset of agents exhibits, for example, Byzantine behaviour, then their behaviour do not affect the rest of multiagent system from functioning.