<p>In this paper, a reinforcement learning-based distributed appointed-time prescribed performance optimal formation control strategy is proposed for multiple-QUAVs with model uncertainties and unknown disturbances. By integrating appointed-time prescribed performance functions and neural networks, a reinforcement learning-based optimal formation controller is proposed to achieve prescribed transient and steady-state performance at an appointed time. An actor-critic reinforcement learning neural network and a self-structuring neural network (SSNN) are employed to address the Hamilton–Jacobi–Bellman (HJB) equation and unknown dynamics, respectively. Additionally, a self-structuring neuron update strategy is proposed for deployment in these neural networks to enable the optimal number of neurons to be adjusted online without affecting the approximation performance. The Lyapunov analysis method is used to prove that all the error signals are semiglobally uniformly ultimately bounded (SGUUB). Finally, the effectiveness of the proposed control strategy is demonstrated via numerical simulation experiments.</p>