The Roboticist's Workbench

Under the umbrella of the Roboticist's Workbench we place effort on integrating and hardening research prototypes into a robust and versatile infrastructure that can support further research. We have mostly worked on laser range data so far, investigating human pattern recognition (ECAI 2012) and autonomously collecting supervision for improving the pattern recognition models (Machine Learning - Cognitive Systems Workshop, 2012). RoboHPR, an off-line human pattern recognition prototype has been released.

Furthermore, we have released the first version of RoboMAE, a visual environment for annotating multimodal perception channels and for training and testing perception components (MMEDIA 2014).

On-going projects include using intensity information to improve the segmentation of laser range data, fusing vision and laser range data to identify landmarks in SLAM, and transferring methods from simulated racing games to robot path planning.