-
A successive convexification approach for robust receding horizon control
Authors:
Yana Lishkova,
Mark Cannon
Abstract:
A novel robust nonlinear model predictive control strategy is proposed for systems with nonlinear dynamics and convex state and control constraints. Using a sequential convex approximation approach and a difference of convex functions representation, the scheme constructs tubes that contain predicted model trajectories, accounting for approximation errors and disturbances, and guaranteeing constra…
▽ More
A novel robust nonlinear model predictive control strategy is proposed for systems with nonlinear dynamics and convex state and control constraints. Using a sequential convex approximation approach and a difference of convex functions representation, the scheme constructs tubes that contain predicted model trajectories, accounting for approximation errors and disturbances, and guaranteeing constraint satisfaction. An optimal control problem is solved as a sequence of convex programs. We develop the scheme initially in the absence of external disturbances and show that the proposed nominal approach is non-conservative, with the solutions of successive convex programs converging to a locally optimal solution for the original optimal control problem. We extend the approach to the case of additive disturbances using a novel strategy for selecting linearization points. As a result we formulate a robust receding horizon strategy with guarantees of recursive feasibility closed-loop system stability.
△ Less
Submitted 27 January, 2025; v1 submitted 15 February, 2023;
originally announced February 2023.
-
Discrete Lagrangian Neural Networks with Automatic Symmetry Discovery
Authors:
Yana Lishkova,
Paul Scherer,
Steffen Ridderbusch,
Mateja Jamnik,
Pietro Liò,
Sina Ober-Blöbaum,
Christian Offen
Abstract:
By one of the most fundamental principles in physics, a dynamical system will exhibit those motions which extremise an action functional. This leads to the formation of the Euler-Lagrange equations, which serve as a model of how the system will behave in time. If the dynamics exhibit additional symmetries, then the motion fulfils additional conservation laws, such as conservation of energy (time i…
▽ More
By one of the most fundamental principles in physics, a dynamical system will exhibit those motions which extremise an action functional. This leads to the formation of the Euler-Lagrange equations, which serve as a model of how the system will behave in time. If the dynamics exhibit additional symmetries, then the motion fulfils additional conservation laws, such as conservation of energy (time invariance), momentum (translation invariance), or angular momentum (rotational invariance). To learn a system representation, one could learn the discrete Euler-Lagrange equations, or alternatively, learn the discrete Lagrangian function $\mathcal{L}_d$ which defines them. Based on ideas from Lie group theory, in this work we introduce a framework to learn a discrete Lagrangian along with its symmetry group from discrete observations of motions and, therefore, identify conserved quantities. The learning process does not restrict the form of the Lagrangian, does not require velocity or momentum observations or predictions and incorporates a cost term which safeguards against unwanted solutions and against potential numerical issues in forward simulations. The learnt discrete quantities are related to their continuous analogues using variational backward error analysis and numerical results demonstrate the improvement such models can have both qualitatively and quantitatively even in the presence of noise.
△ Less
Submitted 27 January, 2025; v1 submitted 19 November, 2022;
originally announced November 2022.
-
A Multirate Variational Approach to Nonlinear MPC
Authors:
Yana Lishkova,
Mark Cannon,
Sina Ober-Blöbaum
Abstract:
A multirate nonlinear model predictive control (NMPC) strategy is proposed for systems with dynamics and control inputs evolving on different timescales. The proposed multirate formulation of the system model and receding horizon optimal control problem allows larger time steps in the prediction horizon compared to single-rate schemes, providing computational savings while ensuring recursive feasi…
▽ More
A multirate nonlinear model predictive control (NMPC) strategy is proposed for systems with dynamics and control inputs evolving on different timescales. The proposed multirate formulation of the system model and receding horizon optimal control problem allows larger time steps in the prediction horizon compared to single-rate schemes, providing computational savings while ensuring recursive feasibility. A multirate variational model is used with a tube-based successive linearization NMPC strategy. This allows either Jacobian linearization or linearization using quadratic and linear Taylor series approximations of the Lagrangian and generalized forces respectively, providing alternative means for computing linearization error bounds. The two approaches are shown to be equivalent for a specific choice of approximation points and their structure-preserving properties are investigated. Numerical examples are provided to illustrate the multirate approach, its conservation properties and computational savings.
△ Less
Submitted 3 July, 2022; v1 submitted 8 November, 2021;
originally announced November 2021.
-
A multirate variational approach to simulation and optimal control for flexible spacecraft
Authors:
Yana Lishkova,
Sina Ober-Blöbaum,
Mark Cannon,
Sigrid Leyendecker
Abstract:
We propose an optimal control method for simultaneous slewing and vibration control of flexible spacecraft. Considering dynamics on different time scales, the optimal control problem is discretized on micro and macro time grids using a multirate variational approach. The description of the system and the necessary optimality conditions are derived through the discrete Lagrange-d'Alembert principle…
▽ More
We propose an optimal control method for simultaneous slewing and vibration control of flexible spacecraft. Considering dynamics on different time scales, the optimal control problem is discretized on micro and macro time grids using a multirate variational approach. The description of the system and the necessary optimality conditions are derived through the discrete Lagrange-d'Alembert principle. The discrete problem retains the conservation properties of the continuous model and achieves high fidelity simulation at a reduced computational cost. Simulation results for a single-axis rotational maneuver demonstrate vibration suppression and achieve the same accuracy as the single rate method at reduced computational cost.
△ Less
Submitted 30 October, 2020; v1 submitted 12 September, 2020;
originally announced September 2020.
