Cloud-Based Model Predictive Control: Establishing a Fully Distributed Architecture for Nonlinear MPC
Over the last two decades, distributed model predictive control (MPC) has become an increasingly popular research topic. Distributed MPC constitutes a valuable tool, providing computationally efficient optimal control algorithms that can successfully account for the interactions of complex, interconnected processes. Achieving closed-loop stability and centralized performance in nonlinear, distributed or decentralized systems remains an obstacle to existing distributed MPC formulations. In this thesis a discrete-time cloud-based model predictive control architecture for application to unstable nonlinear systems under process constraints is developed using a cooperative, distributed optimization approach. Distributed time-varying extremum-seeking, a form of model-free optimal control, is integrated to a conventional MPC scheme to solve the optimization problem at each time step. Each of the control horizon’s inputs is assigned to an agent that shares its local cost information over a network. Agents use this data to estimate the average total cost of the system, which they minimize by a gradient-descent control law based on local approximations of the average total cost gradient. To demonstrate the effectiveness of the discrete-time cloud-based MPC architecture and its stability in closed-loop, three case studies are performed. The first study considers a nonlinear MIMO system subjected to input constraints. The numerical results indicate that the proposed controller recovers the performance of its centralized counterpart. The second case study considers a non-isothermal exothermic CSTR. The simulation demonstrates that the cloud-based architecture can effectively stabilize systems with complex dynamics of higher order. The third case study investigates the architecture’s applicability to economic MPC and shows that the developed controller is capable of stabilizing a system to a periodic orbit when suitable constraints are applied.
URI for this recordhttp://hdl.handle.net/1974/22971
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item:
Except where otherwise noted, this item's license is described as Attribution-NonCommercial 3.0 United States
Showing items related by title, author, creator and subject.
Lewis, Andrew D.; Murray, Richard M. (Society for Industrial and Applied Mathematics, 1995)In this paper we present a definition of "configuration controllability" for mechanical systems whose Lagrangian is kinetic energy with respect to a Riemannian metric minus potential energy. A computable test for this ...
Bullo, Francesco; Lewis, Andrew D. (2002)Controllability and kinematic modeling notions are investigated for a class of mechanical control systems. First, low-order controllability results are given for a class of mechanical control systems. Second, a ...
Towards Controlling Bucket Fill Factor in Robotic Excavation by Learning Admittance Control Setpoints Fernando, Heshan A.; Marshall, Joshua A.; Almqvist, H°akan; Larsson, Johan (2017-09-30)This paper investigates the extension of an admittance control scheme to- ward learning and adaptation of its setpoints to achieve controllable bucket fill factor for robotic excavation of fragmented rock. A previously ...