A new approach to adaptive unfalsified control

In this paper, we present a new approach to adaptive unfalsified control. Similar to previous work, the goal is to combine the unfalsified control strategy with an optimization of the controller parameters. The new idea pursued here is to evolve the set of candidate controllers in the unfalsification approach by an evolutionary algorithm. This enables to stabilize an unknown plant even if there is no stabilizing controller in the initial set. The application of optimization strategies to the unfalsified control idea requires a modification of the cost function used because the original cost function does not indicate instability of potential controllers reliably before they are applied to the plant. We demonstrate and analyse why the original cost function may remain bounded for destabilizing candidate controllers and propose a more reliable cost function. The appraoch is demonstrated for the example of a linear unknown plant.