Safe Formation Control of Uncertain Multiagent Systems From a Bayesian Perspective

In this study, we address the formation control of uncertain multiagent systems with a safety requirement of collision avoidance. The control inputs are determined through quadratic programming (QP) solvers, where both the control Lyapunov function condition and control barrier function condition serve as constraints in the QP problem to achieve the control and safety objectives, respectively. In this work, the Bayesian theorem plays a crucial role in two key aspects. First, the unknown uncertainties are estimated using Gaussian process models, which provide mean and variance information incorporated into the QP framework to determine the control inputs. The integration ensures the attainment of desired objectives with high probability. Second, the Bayesian optimization algorithm is employed to optimize some hyperparameters. These selected hyperparameters enhance the solvability of the QP problem and simultaneously improve the control performance of multiagent systems.