ROS packages:


ar_track_alvar: Software for tracking and estimating the pose of AR tags, a type of visual fiducial. Supports integration of multiple tags and kinect depth data (optional) for improved pose estimates.

dmp: A ROS service implementing Dynamic Movement Primitives. Allows DMPs to be learned from demonstration and then used for fast planning and re-planning.

ml_classifiers: A ROS service that provides an plugin-based interface for supervised learning algorithms. Comes with nearest neighbor and SVM classifiers.

pr2_lfd_utils: ROS utilities for capturing and processing user demonstrations.

changepoint: A ROS service implementing CHAMP: Bayesian changepoint detection over various types of data supported by plugins, including a zero-mean Gaussian model and several types of articulated motion models.

active_articulation: A ROS package implementing a particle filter-based approach to representing and actively reducing uncertainty over articulated motion models.


Other code:


C++ code to reproduce the experiements in Efficient Probabilistic Performance Bounds for Inverse Reinforcement Learning.

Matlab code for automatic segmentation of demonstrations using the BP-AR-HMM.

C code for using PushGP + RL to run experiments like those described in Genetic Programming for Reward Function Search


Tutorials:


S. Niekum. A Brief Introduction to Bayesian Nonparametric Methods for Clustering and Time Series Analysis. Technical report CMU-RI-TR-15-02, Robotics Institute, Carnegie Mellon University, January 2015. [bibtex]