A General-Purpose Software Framework for Dynamic Optimization (Een algemene softwareomgeving voor dynamische optimalisatie)

By Joel A. E. Andersson. 2013

323 citations recorded by the source corpus.

Methods and software for derivative-based numerical optimization in general and simulation-based optimization in particular have seen a large rise in popularity over the past 30 years. Still, due to practical difficulties in implementing many of the methods in a fast and reliable manner, it remains an underused technology both in academia and in industry. To address this, we present a set of methods and tools with the aim of making techniques for dynamic optimization more accessible. In particular, we present CasADi, an open-source software framework for numerical optimization and algorithmic differentiation (AD) that offers a level of abstraction which is lower than algebraic modeling languages, but higher than conventional AD tools. We also discuss several of the many application problems which have already been addressed with CasADi by researchers from diverse fields.

Subfields: Mathematical Optimization

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Mathematical Optimization

Citations in the corpus

FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models (2018, methodology): “This approach is a continuous-time analog to the backpropgation algorithm (Rumelhart et al., 1986; Andersson, 2013) and can be combined with gradient-based optimization to fit the parameters θ by maximum likelihood.”
Neural Ordinary Differential Equations (2018, neutral): “While use of vector-Jacobian products for solving the adjoint method has been explored in optimal control (Andersson, 2013), we explore a fully general integration of black-box ODE solvers into automatic differentiation (Baydin et al.”
Learning an Approximate Model Predictive Controller With Guarantees (2018, methodology): “To create the learning samples, the RMPC is sampled over X with a uniform grid using Casadi 2.4.2 [25] with Python 2.7.2 with grid size of 2 .”
CasADi: a software framework for nonlinear optimization and optimal control (2018, methodology): “Secondly, it can be shown [5] that both forward and reverse mode directional derivatives of this problem can be calculated efficiently by solving a problem that has exactly the same structure.”
Scalable Gradients for Stochastic Differential Equations (2020, methodology): “This method was well-known in the physics, numerical analysis, and control communities for decades [3, 4, 57, 62].”
Learning-Based Model Predictive Control for Safe Exploration (2018, background): “However, we can still obtain exact function values and derivative information by means of algorithmic differentiation, which is at the core of many state-of-the-art optimization software libraries [29].”
Scalable Reversible Generative Models with Free-form Continuous Dynamics (2018, methodology): “This approach is a continuous-time analog to the backpropgation algorithm (Rumelhart et al., 1986; Andersson, 2013) and can be combined with gradientbased optimization methods to fit the parameters θ.”
JuMP: A Modeling Language for Mathematical Optimization (2015, methodology): “CasADi has specialized features for optimal control but, unlike traditional AMLs, does not support linear optimization as a special case.”
Cautious Model Predictive Control Using Gaussian Process Regression (2017, methodology): “Derivative information can be automatically generated using automated differentiation tools, such as CASADI [40].”