Event
LOPAL: Active Learning with Local Quality Assessment for Imperfect Demonstrations
1 reports · 1 sources
Johannes Heidersberger, Shail Jadav, and Dongheui Lee propose an active learning framework, LOPAL, which uses a Gaussian mixture model to simultaneously encode demonstration trajectories and their local quality assessments, thereby generating trajectories that outperform imperfect demonstrations. In regions with insufficient data, the method actively requests user corrections through a shared autonomy mechanism, while the robot autonomously executes learned behaviors. The authors state that LOPAL has been validated in simulation and real-world experiments; in a real pipeline inspection task, task performance improved by up to 27.31%, while reducing the effort required to collect demonstrations.
Report timeline
- LOPAL: Local Performance-Aware Active Learning from Imperfect Demonstrations
LOPAL is a local performance-aware active learning method that uses a Gaussian Mixture Model (GMM) to jointly encode demonstration trajectories and their local quality assessments, thereby generating trajectories that outperform the imperfect demonstrations. The method also uses a shared autonomy (SA) mechanism to actively request user corrections in areas with insufficient data, while the robot autonomously executes its learned behavior during this process.
arXiv cs.RO ·