Resources

YALMIP

The MATLAB toolbox YALMIP has been developed at Linköping University for more than two decades and is widely used in both academia and industry around the world. YALMIP introduces a natural programming language for representing and developing optimization-based algorithms, dramatically reducing the effort required to test new ideas for researchers and engineers across a wide range of fields.

Over the years, the toolbox has both followed and pushed the frontiers of applied optimization, adding support and pioneering new ideas in areas such as parametric programming, mixed-integer conic programming, semidefinite relaxations, robust optimization, and global optimization. The development is still active, pursuing new ideas relevant to the research in SEDDIT.

yalmip.github.io

Fault Diagnosis Toolbox

Fault Diagnosis Toolbox is a Python package for the analysis and design of fault diagnosis systems for dynamic systems, primarily described by differential-algebraic equations. Key features of the toolbox are extensive support for structural analysis of large-scale dynamic models, fault isolability analysis, sensor placement analysis, and code generation in C/C++ and Python.

Python, documentation
Github, source code
Matlab

Sensor fusion

Dive into the fundamentals of statistical estimation and filtering theory with our comprehensive Sensor Fusion course. Designed for graduate-level students at Master’s level, this course explores key concepts using localization and target tracking applications. Course highlights:

• Short Topic Videos: Concise and informative videos that cover the most essential topics.
• Reading Instructions: Guided readings to deepen your understanding.
• Hands-on Labs: Two practical labs with detailed instructions and datasets to apply your knowledge.

This course is perfect for self-study, providing all the materials you need to master Sensor Fusion at your own pace.

Course TSRT14

Reinforcement learning

Reinforcement learning has a strong theoretical basis. Despite its theoretical complexity, it can be applied to solve many different practical problems. So, while learning RL, one may wonder, “How can such theoretical concepts be translated and applied in practice?” Below, you can find the handout and code for a short course on Reinforcement Learning.

Handout

Code

A more comprehensive version of the course can be found below.

Course URL

Code

Real-time certified QP solvers

As a result of our research on real-time certification and exact complexity certification, summarized in the referenced thesis below, we have developed and released several software tools. The ASCertain.jl software used for exact complexity certification is now publicly available. Additionally, we provide daqp, a real-time certifiable, state-of-the-art, free QP solver tailored for embedded applications such as model predictive control (MPC). Although currently optimized for short prediction horizons without exploiting sparsity, it is among the fastest and most reliable QP solvers available. Its ability to be both exactly and real-time complexity certified makes it highly suitable for safety-critical applications.

Furthermore, we introduce ParametricDAQP.jl, a state-of-the-art multi-parametric QP solver based on our recent publication. A Python implementation is also provided. Parametric programming is especially well-suited for real-time applications like MPC, though it extends to broader use cases. Our solver has demonstrated performance approximately an order of magnitude faster than other popular alternatives.

PhD thesis
ASCertain.jl
daqp QP solver
ParametricDAQP.jl (Julia)
ParametricDAQP (Python)
Publication behind ParametricDAQP