/dmpl_experiments

Objectworld Experiments

Primary LanguageMATLAB

DMPL Object World Experiments

Density matching polilcy learning (DMPL) is a policy learning algorithm in a continuous state action space by first estimating the proximal reward and modeling the policy function using model predictive control with respect to the logarithm of the optimized reward. This implementation is based on MATLAB. To compare with other IRL algorithms, IRL-toolkit is used: https://github.com/vvanirudh/IRL-Toolkit.