Takes a single pendulum (with a torque actuator) and models it as a Markov Decision Process (MDP), using linear barycentric interpolation over a uniform grid. Then, value iteration is used to compute the optimal policy, which is then displayed as a plot. Finally, a simulation is run to demonstrate how to evaluate the optimal policy.
Matthew Kelly (2021). Markov Decision Process - Pendulum Control (https://github.com/MatthewPeterKelly/MDP_Pendulum), GitHub. Retrieved .
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!Create scripts with code, output, and formatted text in a single executable document.
@Hazim -- This work is loosely based on materia from Russ Tedrake's underactuated class at MIT, and some of his other papers on the topic. I think these course notes are a good place to start: http://underactuated.mit.edu/dp.html. The barycentric interpolation implementation is based (roughly) on some code from the an early version of the Drake optimization toolbox, but rewritten to be fast in Matlab.
Hi Matthew. Great work! Could you please share your references?