Videos
Giovanni Franzese, Carlos Celemin, and Jens Kober. Learning Interactively to Resolve Ambiguity in Reference Frame Selection. In Conference on Robot Learning (CoRL), 2020.
Snehal Jauhri, Carlos E. Celemin, and Jens Kober. Interactive Imitation Learning in State-Space. In Conference on Robot Learning (CoRL), 2020.
Rodrigo Pérez Dattari, Carlos Celemin, Giovanni Franzese, Javier Ruiz Del Solar, and Jens Kober. Interactive Learning of Temporal Features for Control. In IEEE Robotics and Automation Magazine (Special Issue on “Deep Learning and Machine Learning in Robotics”), 2020.
Use of State Representation Learning for teaching interactively complex tasks with high dimensional and partial observability. Tasks include a real inverted pendulum, a robot arm as orange selector in a conveyor belt, along with a couple of simulated environments.
Carlos Celemin, Guilherme Maeda, Javier Ruiz-del-Solar, Jan Peters, and Jens Kober. Reinforcement Learning of Motor Skills using Policy Search and Human Corrective Advice. International Journal of Robotics Research, OnlineFirst, 2019.
Learning movement primitives for the "ball in a cup" task with human feedback and Reinforcement Learning
Correcting a trajectory only with human corrective feedback (UR5 robot arm)
Correcting a trajectory only with human corrective feedback (simulated arm)
Carlos Celemin and Jens Kober. Simultaneous Learning of Objective Function and Policy from Interactive Teaching with Corrective Feedback. In IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), 2019.
Training a policy for tracking an object with Reinforcement Learning and a reward function learned from the human corrections
Rodrigo Pérez Dattari, Carlos Celemin, Javier Ruiz Del Solar, and Jens Kober. Continuous Control for High-Dimensional State Spaces: An Interactive Learning Approach. In IEEE International Conference on Robotics and Automation (ICRA), pp. 7611–7617, 2019.
Learning from scratch deep neural networks policies interactively based on corrective human feedback. Simulated environment: Car Racing. Real environments: Robot arm reacher, and pusher, and a Duckie Racing (Self driving car from Duckie Town)
Carlos Celemin, Javier Ruiz Del Solar, and Jens Kober. A fast hybrid reinforcement learning framework with human corrective feedback. In Autonomous Robots, 2019.
Combining Policy Search and Human corrections for training policies in MDPs. Simulated environment: Cart-Pole. Real environments: Swing-Up pendulum, Inverted Wedge, and inverse kinematics of a 3 DoF robot arm
Rodrigo Pérez Dattari, Carlos Celemin, Javier Ruiz Del Solar, and Jens Kober. Interactive Learning with Corrective Feedback for Policies based on Deep Neural Networks. In International Symposium on Experimental Robotics (ISER), 2018.
Learning deep neural networks policies interactively based on corrective human feedback with a pre-trained Autoencoder. Simulated environment: Car Racing. Real environment: Duckie Racing (Self driving car from Duckie Town)
Carlos Celemin, and Javier Ruiz Del Solar. An Interactive Framework for Learning Continuous Actions Policies based on Corrective Feedback. In Journal of Intelligent & Robotic Systems, 2018.
Machine Learning from Human Corrective Advice (COACH). Simulated environment: Cart-Pole, bike balancing, and ball dribbling with the Nao robot. Real environment: ball dribbling with the Nao robot
Carlos Celemin, Rodrigo Pérez Dattari, Javier Ruiz Del Solar, and Manuela Veloso. Interactive Machine Learning Applied to Dribble a Ball in Soccer with Biped Robots. In RoboCup International Symposium, 2017.
Interactive Machine Learning Applied to Dribble a Ball with Nao Robots