Mixed-Integer Model Predictive Control
As part of MIT’s course 16.413 I worked on a team to derive a model-predictive controller for a simulated quadrotor in a non-convex planning space. I used SCIP, an open-source optimization software in conjunction with simplified quadrotor dynamics model to perform the path-planning subject to obstacle avoidance constraints.
Because objects could be placed arbitrarily within the environment, the planning problem was inherently non-convex. For this reason the problem was modeled as a mixed-integer problem.
In this repo all of our work is shown in a Jupyter Notebook, along with a brief explanation of some relevant topics in optimization and a derivation of the quadrotor dynamics.