High-Performance Reference Governors for Faster Constraint Management

This paper proposes an extension of the reference governor, which we refer to as the High-Performance Reference Governor (HP-RG), for constraint management of linear systems. This extension is motivated by the fact that reference governors typically respond slower than model-predictive controllers, for example in response to a step change in the tracking setpoint. The newly-developed scheme introduces two modes: the “standard mode”, and the “fast mode”. In the standard mode, the scheme behaves the same as a standard reference governor. In the fast mode, however, the computed control command is less conservative than what a standard reference governor would compute. This control command is computed such that a constraint-admissible transition back to the standard mode is possible in the next timestep. This guarantees recursive feasibility and constraint satisfaction. To guarantee stability and prevent unnecessary mode switches, a stability-inducing mechanism is introduced, which allows transition to fast mode only when large variations in the setpoint are detected. To further reduce the computational footprint of HP-RG, an extension is presented that can be implemented using an explicit algorithm. Finally, the effectiveness of the developed algorithms is verified using numerical examples.