Idle speed control using linear time varying model predictive control and discrete time approximations

Real time model predictive idle speed control of ultra-lean burn engines: experimental results

“…In order to counter the computational complexity due to nonlinear MPC associated with the solution of (9)-(13), the linear time varying (LTV) MPC approach presented in (Falcone et al (2008), Sharma et al (2010)) is applied to devise an idle speed controller for ultra-lean burn engines.…”

“…In (Sharma et al (2010)), results of (Nešić and Teel (2004)) and (Falcone et al (2008)) are collectively used to counter the effects of discrete time approximations and limitations of real time computational abilities. For completeness, the LTV-MPC formulation for the idle speed control of ultralean burn engines is summarized here which is experimentally validated in Section 4 on an H 2 ICE.…”

“…The aim of this paper is twofold: firstly, an idle speed controller for ultra-lean burn engines is developed using model predictive control theory developed in (Sharma et al (2010)). The control design is based on the successive online linearizations of an approximate discrete time nonlinear engine model about the current operating point so as to alleviate the challenges of computational burden that arises from nonlinear model predictive control.…”

Data-driven predictive control of idle speed control for SI engine

“…The low fuel economy is considered as the main reason of the global energy crisis. In urban area, a car with passenger needs about one third of its on board fuel in normal city driving during engine's idling state, with traffic loads increased this percentage will further increase [1]. Therefore it is significant to optimize vehicle and powertrain operations with respect to often-conflicting requirements of improved fuel economy, reduced emissions, guaranteed combustion stability at idle.…”

Distributed Model Predictive Control of Nonlinear Systems Based on Price-Driven Coordination

“…Remark 1. A generic-type finite-time MPC, based on the successively linearized model (eqs 5 ) at time instant k can be formulated as subject to for l = 0, ..., H p , where H p is the prediction horizon and H u is the control horizon; and, x̂ and û are states and manipulated input variables inside the controller, respectively; and . In addition, X ( k ) = [ x̂ ( k + 1| k ) T , ..., x̂ ( k + H p | k ) T ] T is the vector of the predicted state trajectory; U ( k ) = [ û ( k | k ) T , ..., û ( k + H p – 1| k ) T ] T is the vector of the calculated manipulated variable moves; Q is a positive definite block-diagonal weighting matrix for the states (i.e., Q = diag­{ Q ii }); R is a positive definite block-diagonal weighting matrix for the manipulated variables of the overall system (i.e., R = diag­{ R ii }); and P is a positive definite block-diagonal weighting matrix for the terminal cost of the overall system (i.e., P = diag­{ P ii }).…”