https://github.com/linesd/pilco-disentangling-uncertainty
Disentangling Sources of Uncertainty for Active Exploration (Reinforcement Learning)
https://github.com/linesd/pilco-disentangling-uncertainty
cart-double-pendulum cart-pole control control-systems disentangling-sources dynamical-systems gaussian-processes learning-control machine-learning monte-carlo-samples pendulum pilco probabilistic-inference reinforcement-learning reinforcement-learning-algorithms uncertainty
Last synced: 3 months ago
JSON representation
Disentangling Sources of Uncertainty for Active Exploration (Reinforcement Learning)
- Host: GitHub
- URL: https://github.com/linesd/pilco-disentangling-uncertainty
- Owner: linesd
- License: other
- Created: 2019-05-22T10:38:33.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-12-16T10:42:39.000Z (almost 6 years ago)
- Last Synced: 2025-06-08T12:38:13.467Z (4 months ago)
- Topics: cart-double-pendulum, cart-pole, control, control-systems, disentangling-sources, dynamical-systems, gaussian-processes, learning-control, machine-learning, monte-carlo-samples, pendulum, pilco, probabilistic-inference, reinforcement-learning, reinforcement-learning-algorithms, uncertainty
- Language: MATLAB
- Homepage:
- Size: 140 MB
- Stars: 8
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: license.txt
Awesome Lists containing this project
README
# PILCO - disentangling sources of uncertainty
This repository contains the orgininal PILCO MATLAB code (see pilco_readme.txt and license) with added functionality that disentangles the sources of uncertainty in the cost function for a given policy by Monte Carlo sampling.
Uncertainty analysis implemented for the following scenarios:
* Pendulum
* Cart-Pole
* Pendubot
* Cart-Double-Pendulum## Run
Scenarios can be run from from the scenarios folder:
* pendulum_learn.m
* cartPole_learn.m
* pendubot_learn.m
* cartDouble_learn.mSet the number of Monte Carlo samples in the scenario settings files.
## Plot
Plots of the Monte Carlo samples and uncertainty decomposition can be made for each environment from the scripts in the uncertainty_plots folder.
## Output
example plots for the **pendubot scenario**:


example plots for the **cartDoublePendulum scenario**:


