IFAC 2026 Workshop: Hands-on Notebooks

Runnable notebooks from GPU Acceleration in Optimization and Optimal Control, pre-congress workshop WS-12 at the 23rd IFAC World Congress (BEXCO, Busan, Sunday 23 August 2026).

Workshop page, lecture slides, and rendered notebooks: https://madsuite.org/ifac2026/

Notebook	Contents
1-gpu-computing.ipynb	GPU computing in the Julia language: arrays, broadcasting, map/reduce, linear algebra, KernelAbstractions.jl kernels, batched simulation
2-optimal-control-and-estimation.ipynb	Direct transcription, pendulum swing-up, Goddard's rocket by orthogonal collocation, particle steering, parameter estimation for the Boehm STAT5 model
reductions-solution.ipynb	Worked solution to the interior-point reductions exercise
montecarlo-pi-solution.ipynb	Worked solution to the Monte-Carlo π exercise
minimum-time-solution.ipynb	Worked solution to the minimum-time swing-up exercise
particle-steering-solution.ipynb	Worked solution to the particle steering exercise

The notebooks ship without outputs, so every output you see is one you generated. Executed versions, with outputs from the workshop's GPU servers (NVIDIA Quadro GV100), are rendered on the workshop page.

Running them yourself

The GPU parts need an NVIDIA GPU; the CPU parts (Ipopt solves, plotting) run anywhere. During the workshop itself, a hosted notebook server with this environment preinstalled is provided in the room, so none of this setup is needed there.

1. Install Julia 1.12.

2. Instantiate the project environment (Project.toml pins the package versions the notebooks were executed with):

   julia --project=. -e 'using Pkg; Pkg.instantiate()'

3. Install the Jupyter kernel, pointed at this environment:

   using Pkg; Pkg.add("IJulia")
   using IJulia
   installkernel("julia-ifac2026", "--project=@.")

4. Start Jupyter in this directory and select the julia-ifac2026 kernel:

   jupyter lab

The notebooks read boehm_data.jl and figs/ by relative path, so run them from this directory.

License

MIT