Reduced Process Control Complexity using Explicit Model Predictive Control
Nonlinear Model Predictive Control (NMPC) has become a widely recognized method to model the behavior of a wide range of complex processes such as automotive mechatronics, compressor control and combustion control in engines. Within this aspect, the need for execution of a numerical optimization algorithm on-line in real-time can be avoided with an explicit feedback law introduced in Explicit MPC. This leads to computational efficiency for on-line computations, as well as verifiability of the implementation and low software and hardware complexity for the design of embedded control systems. With these highly desirable advantages in further reducing the complexity of systems, research in advanced Explicit MPC methodologies has the potential to impact microcontroller implementations like seen in >$20B Internal Combustion Engine industry.
Cost-Efficient Explicit Model Predictive Control with Probabilistic Region Selection Using Markov Chain A computationally efficient Explicit Model Predictive Control method has been developed in the Aerospace Engineering Department of the University of Michigan. The proposed method uses a probabilistic method rather than a deterministic one to determine the state space region. Based on a Markov chain (MC), the method predicts the system state space region membership and therefore, the associated controller feedback gain a-priori offline with reduced computation cost according to likelihoods captured in the chain. This is then adapted on-line in real-time for the microcontroller. Furthermore, real-time computational constraints can be satisfied with the resulting faster online computation time that could be a desirable feature leading to inexpensive control implementations.
Applications • Controllers utilizing Explicit Model Predictive Control • Internal Combustion Engines, Automotive Mechatronics • Compressor Control
Advantages • Faster online computation time • Low software and hardware complexity of the system