Search

In the design of adaptive IIR filters, the multi-modal nature of the error surfaces can limit the use of gradient-based and other iterative search methods. Stochastic learning automata have previously been shown to have global optimisation properties making them suitable for the optimisation of filter coefficients. Continuous action reinforcement learning automata are presented as an extension to the standard automata which operate over discrete parameter sets. Global convergence is claimed, and demonstrations are carried out via a number of computer simulations.

A reinforcement learning strategy is applied to the problem of the dynamic roll control of a full-body vehicle system fitted with semi-active suspension under digital control. The simulation model used in this study is based upon realistic vehicle hardware. Prior engineering knowledge of the non-linear actuation system is used to develop a control structure. Parameters in this structure are then obtained using Continuous Action Reinforcement Learning Automata (CARLA), an extension of the interconnected learning automata methodology. No model-based information is used in the controller synthesis.