A Novel Model Predictive Control Formulation for Hybrid Systems With Application to Adaptive Behavioral...

Area of Science:

• Control Systems Engineering
• Hybrid Systems Modeling
• Behavioral Health Interventions

Background:

• Traditional model predictive control (MPC) schemes often use move suppression weights for tuning, which can be unintuitive.
• Emerging applications like adaptive behavioral health interventions require robust control with flexible tuning capabilities.
• Linear hybrid systems present unique challenges for control due to their discrete and continuous dynamics.

Purpose of the Study:

• To present a novel multiple-degree-of-freedom model predictive control (MPC) formulation for linear hybrid systems.
• To enable independent adjustment of setpoint tracking, measured disturbance rejection, and unmeasured disturbance rejection.
• To provide a more intuitive and flexible controller tuning method compared to traditional MPC.

Main Methods:

• Developed a novel MPC formulation based on a multiple-degree-of-freedom approach.
• Applied the formulation to a hypothetical adaptive intervention problem inspired by the Fast Track program.
• Simulated the controller's performance under conditions of simultaneous disturbances and plant-model mismatch.

Main Results:

• The proposed MPC formulation allows independent tuning of key control system responses.
• Demonstrated robust performance in simulations involving significant disturbances and model inaccuracies.
• The hybrid MPC approach proved effective for adaptive interventions, handling participant variability.

Conclusions:

• The novel MPC formulation offers enhanced flexibility and intuitiveness for tuning linear hybrid systems.
• This approach is well-suited for demanding applications such as adaptive behavioral health interventions.
• The method successfully addresses challenges posed by real-world complexities like participant variability and model mismatch.