Adaptive Control of Uncertain Systems with Input Delay Based on Active Inference

Active inference as a theory involving both perception and control has been popular for its advantages in fast learning, control accuracy and computational efficiency in the field of unmanned system control. This work firstly implement the active inference control (AIFC) framework for adaptive control of uncertain unmanned systems with input delay for trajectory tracking. Compared with radial basis function neural network control (RBFNNC), AIFC can achieve better control performance in both convergence speed and task accuracy. Secondly, considering the non-negligible input delay of a practical unmanned system, the predictive state of the current state is introduced into the active inference framework to design a delay feedback AIFC for the uncertain system with delay. The delay feedback AIFC is applied to a trajectory tracking problem for an unmanned car. The simulation results validate our claimed advantages of the AIFC on convergence speed and tracking accuracy with various delay size.