Model Predictive Control in the browser with WebAssembly | garethx
The demo app does work on mobile, but the experience is improved on a big screen. For some ideas of what to try in the sim, see Interacting with the simulator.
First, a caveat: My background is not in control theory, but rather computer vision. The purpose of this project was to improve my practical understanding of trajectory optimization and Model Predictive Control (MPC) by building a working demonstration in simulation. The implementation detailed in this article might be suboptimal in some ways - it is ultimately a toy implementation for my own edification. That said, perhaps it can provide some educational value to others treading the same path.
Additionally, I had been looking for an opportunity to build something with emscripten - a C++ compiler that targets WebAssembly (WASM). Compiling a control algorithm for the browser may seem an unusual choice, but it does afford us some nice properties:
Lastly, the project served as an opportunity to add some new features to wrenfold. wrenfold is a symbolic code-generation framework designed to ease the quick implementation of objective functions for nonlinear optimization problems. (Disclaimer: I am the author of wrenfold).
