Community Extension: Optimization framework for sensor placement (FastGaussianPuffOpt)

[dietmarwo]

Hello FastGaussianPuff team,

First, thank you for providing this efficient implementation of the Gaussian Puff model. I have been using it as the forward model for a sensor placement optimization project and wanted to share the results with you.

I have created an extension repository, FastGaussianPuffOpt, which wraps your FastGaussianPuff implementation to solve inverse problems and design optimization tasks.

Key features of the extension:

• Sensor Optimization: Uses the puff model to determine optimal sensor coordinates $(x, y, z)$ for maximizing detection or minimizing error.
• Advanced Solvers: Integrates derivative-free optimization libraries (specifically fcmaes and Optuna) to handle the non-convex nature of the dispersion landscape.
• Constraint Handling: Includes logic for handling complex 3D spatial constraints (e.g., obstacles or valid placement zones).

I thought this might be of interest to you or other users of this library who are looking to apply the model to optimal experimental design or source reconstruction tasks. I’m also curious about other optimization problems you are interested in. FastGaussianPuff’s efficiency is impressive, and because it is single-threaded, it fits extremely well into parallel optimization frameworks where multiple independent simulations need to run simultaneously.

Best regards,
Dietmar Wolz / dietmarwo