A primal-dual interior-point method for robust optimal control of linear discrete-time systems (original) (raw)
Related papers
Efficient robust optimization for robust control with constraints
Mathematical Programming, 2008
This paper proposes an efficient computational technique for the optimal control of linear discrete-time systems subject to bounded disturbances with mixed polytopic constraints on the states and inputs. The problem of computing an optimal state feedback control policy, given the current state, is non-convex. A recent breakthrough has been the application of robust optimization techniques to reparameterise this problem as a convex program. While the reparameterised problem is theoretically tractable, the number of variables is quadratic in the number of stages or horizon length N and has no apparent exploitable structure, leading to computational time of O(N 6 ) per iteration of an interior-point method. We focus on the case when the disturbance set is ∞-norm bounded or the linear map of a hypercube, and the cost function involves the minimization of a quadratic cost. Here we make use of state variables to regain a sparse problem structure that is related to the structure of the original problem, that is, the policy optimization problem may be decomposed into a set of coupled finite horizon control problems. This decomposition can then be formulated as a highly structured quadratic program, solvable by primal-dual interior-point methods in which each iteration requires O(N 3 ) time. This cubic iteration time can be guaranteed using a Riccati-based block factorization technique, which is standard in discrete-time optimal control. Numerical results are presented, using a standard sparse primal-dual interior point solver, which illustrate the efficiency of this approach.
Application of a merit function based interior point method to linear model predictive control
International Journal of Information Technology, Modeling and Computing, 2014
This paper presents robust linear model predictive control (MPC) technique for small scale linear MPC problems. The quadratic programming (QP) problem arising in linear MPC is solved using primal dual interior point method. We present a merit function based on a path following strategy to calculate the step length α, which forces the convergence of feasible iterates. The algorithm globally converges to the optimal solution of the QP problem while strictly following the inequality constraints. The linear system in the QP problem is solved using LDL T factorization based linear solver which reduces the computational cost of linear system to a certain extent. We implement this method for a linear MPC problem of undamped oscillator. With the help of a Kalman filter observer, we show that the MPC design is robust to the external disturbances and integrated white noise.
SIAM Journal on Optimization, 2012
A method is proposed for reducing the cost of computing search directions in an interior point method for a quadratic program. The KKT system is partitioned and modified, based on the ratios of the slack variables and dual variables associated with the inequality constraints, to produce a smaller, approximate linear system. Analytical and numerical results are included that suggest the distribution of eigenvalues of the new, approximate system matrix is improved, which makes it more amenable to being solved with an iterative linear solver. ...
New optimization methods in predictive control
2011
This thesis is mainly concerned with the efficient solution of a linear discrete-time finite horizon optimal control problem (FHOCP) with quadratic cost and linear constraints on the states and inputs. In predictive control, such a FHOCP needs to be solved online at each sampling instant. In order to solve such a FHOCP, it is necessary to solve a quadratic programming (QP) problem. Interior point methods (IPMs) have proven to be an efficient way of solving quadratic programming problems. A linear system of equations needs to be solved in each iteration of an IPM. The ill-conditioning of this linear system in the later iterations of the IPM prevents the use of an iterative method in solving the linear system due to a very slow rate of convergence; in some cases the solution never reaches the desired accuracy. A new well-conditioned IPM, which increases the rate of convergence of the iterative method is proposed. The computational advantage is obtained by the use of an inexact Newton ...
Design and Implementation of Interior-Point Method Based Linear Model Predictive Controller
Communications in Computer and Information Science, 2013
Linear model predictive control (MPC) assumes a linear system model, linear inequality constraints and a convex quadratic cost function. Thus, it can be formulated as a quadratic programming (QP) problem. Due to associated computational complexity of QP solving algorithms, its applicability is restricted to relatively slow dynamic systems. This paper presents an interiorpoint method (IPM) based QP solver for the solution of optimal control problem in MPC. We propose LU factorization to solve the system of linear equations efficiently at each iteration of IPM, which renders faster execution of MPC. The approach is demonstrated practically by applying MPC to QET DC Servomotor for position control application.
Preconditioners for inexact interior point methods for predictive control
2010
This paper presents a new method for solving a linear discrete-time finite horizon optimal control problem (FHOCP) with quadratic cost and linear constraints on the states and inputs. Such a FHOCP needs to be solved online, at each sampling instant, in predictive control. In order to solve such a FHOCP, it is necessary to solve a quadratic programming (QP) problem. The proposed technique uses an inexact interior-point method (IIPM) to solve the QP problem. This new technique is computationally more efficient than the Riccati Recursion method of Rao, Wright and Rawlings (Journal of Optimization Theory and Applications, 1998), when measured in terms of the number of floating point operations. The computational advantage is obtained by the use of an inexact Newton method, and with the use of novel preconditioners in the minimum residual (MINRES) method. The computational performance of this method is demonstrated by numerical results.
A New Hot-start Interior-point Method for Model Predictive Control
2011
In typical model predictive control applications, a finite-horizon optimal control problem, in the form of a quadratic program (QP), must be solved at each sampling instant with a known initial state. We present a new hot-start strategy to solve such QPs using interior-point methods, where the first interior-point iterate is constructed from a backward time-shifting of the solution to the QP at the previous time-step. There are two difficulties with such a strategy. First, a naive backward shifting of a previous solution can yield an initial iterate on the boundary of the primal-dual feasible region, leading to blocking of the search direction and consequently to very small and inefficient interior-point steps. Second, a backward shifted solution does not provide a set of strictly feasible terminal KKT conditions. In order to address both of these issues, we propose a modification to the basic backwardshifting method which provides simultaneously an initial iterate that satisfies strict feasibility conditions and a strictly feasible set of primal and dual terminal decision variables. Numerical results indicate that the proposed technique yields convergence in fewer iterations than a coldstart interior-point method.
A Duality Approach for Solving Control-Constrained Linear-Quadratic Optimal Control Problems
SIAM Journal on Control and Optimization, 2014
We use a Fenchel duality scheme for solving control-constrained linear-quadratic optimal control problems. We derive the dual of the optimal control problem explicitly, where the control constraints are embedded in the dual objective functional, which turns out to be continuously differentiable. We specifically prove that strong duality and saddle point properties hold. We carry out numerical experiments with the discretized primal and dual formulations of the problem, for which we implement powerful existing finite-dimensional optimization techniques and associated software. We illustrate that by solving the dual of the optimal control problem, instead of the primal one, significant computational savings can be achieved. Other numerical advantages are also discussed.
Model predictive control based on linear programming~ the explicit solution
… on Automatic Control, 2002
For discrete-time linear time-invariant systems with constraints on inputs and states, we describe a method to determine explicitly, as a function of the initial state, the solution to optimal control problems that can be formulated using a linear program (LP). In particular, we focus our attention on performance criteria based on a mixed 1/∞-norm, namely 1-norm with respect to time and ∞-norm with respect to space. We show that the optimal control profile is a piecewise affine and continuous function of the initial state. Thus, when optimal control profiles are computed at each time step as in model predictive control (MPC) schemes, the explicit piecewise affine form allows to eliminate on-line LP, as the computation associated with MPC becomes a simple function evaluation. Therefore the proposed technique is attractive for a wide range of applications where simple on-line computation is a crucial requirement. Besides practical advantages, the availability of the explicit structure of the MPC controller provides an insight into the type of control action in different regions of the state space, and highlights possible conditions of degeneracies of the LP, such as multiple optima.
1998
Primal-dual interior-point path-following methods for semide nite programming (SDP) are considered. Several variants are discussed, based on Newton's method applied to three equations: primal feasibility, dual feasibility, and some form of centering condition. The focus is on three such algorithms, called respectively the XZ, XZ+ZX and Q methods. For the XZ+ZX and Q algorithms, the Newton system is well-de ned and its Jacobian is nonsingular at the solution, under nondegeneracy assumptions. The associated Schur complement matrix has an unbounded condition number on the central path, under the nondegeneracy assumptions and an additional rank assumption. Practical aspects are discussed, including Mehrotra predictor-corrector variants and issues of numerical stability. Compared to the other methods considered, the XZ+ZX method is more robust with respect to its ability to step close to the boundary, converges more rapidly, and achieves higher accuracy.