Challenges in set-valued model-predictive control

In this abstract we describe a framework for computationally-aware computing through set-valued model predictive control. Model-predictive control (MPC) can enable multi-objective optimization in real-time, though it depends on accurate models through which future state values can be predicted. This abstract improves upon existing MPC approaches in that it considers the state to be a set (rather than a singleton in the state), allowing the trajectories to be given by a sequence of sets. The framework is beneficial for physical systems control where the uncertainty in future projection can be attributed to both model error, and environmental or sensor uncertainty, thus providing guarantees of performance, robustly. We provide an overview of the framework, and include discussion for its advantages.