Research Summary

In the future, robots are expected to work as companions with humans in various areas including domestic scenarios such as care-giving. However, even with well-engineered robots, it would be unrealistic to move robots directly from factories to home environments to perform complex tasks due to safety. Moreover, robots also have to continuously adapt to new environments to avoid hazardous actions since using experts to program a robot for every environment is impossible. Hence, we need adaptive learning algorithms. Spoken language can be considered one of the most effective communication channels to warn robots about threats. A human can guide the robot by a verbal utterance toward a safer interaction. Our goal is to train a robot to safely perform complex tasks with the ability of processing environmental feedback, including guidance and warnings by a human, to shape a proper signal for updating its own policy. Therefore, this research project is focused on three areas: a) given the verbal instructions, generating high-level actions, b) learning low-level actions to fulfill the high-level action, c) learning warnings using the prosodic/sentiment features of the human speaker.

Publications

H. Beik-Mohammadi, M. A. Zamani, Matthias Kerzel, Stefan Wermter. (2019)
Mixed-Reality Deep Reinforcement Learning for a Reach-to-grasp Task

Artificial Neural Networks and Machine Learning – ICANN 2019, pages 611-623 - Sep 2019. 

Egor Lakomkin, Mohammad Ali Zamani, Cornelius Weber, Sven Magg, Stefan Wermter

(2019, May)

Incorporating End-to-End Speech Recognition Models for Sentiment Analysis (pdf)

2019 International Conference on Robotics and Automation (ICRA), Montreal, QC, Canada, 2019, pp. 7976-7982.

 Mohammad Ali Zamani, Sven Magg, Cornelius Weber, Di Fu, and Stefan Wermter

(2018, Nov)

Deep Reinforcement Learning using Compositional Representations for Performing Instructions. 

Paladyn. Journal of Behavioral Robotics, Volume 9, pages 358-–373 - Nov 2018.

Egor Lakomkin, Mohammad Ali Zamani, Cornelius Weber, Sven Magg, Stefan Wermter

(2018, Oct)

On the Robustness of Speech Emotion Recognition for Human-Robot Interaction with Deep Neural Networks (pdf)

2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, 2018, pp. 854-860.

Hadi Beik-Mohammadi, Matthias Kerzel, Michael Görner, Mohammad Ali Zamani, Manfred Eppe, Stefan Wermter  (2018, Oct)

Neural End-to-End Learning of Reach for Grasp Ability with a 6-DoF Robot Arm

IROS 2018 Workshop on Machine Learning in Robot Motion Planning - Sep 2018.

Matthias Kerzel, Hadi Beik-Mohammadi, Mohammad Ali Zamani, and Stefan Wermter (2018, July)

Accelerating Deep Continuous Reinforcement Learning through Task Simplification.
International Joint Conference on Neural Networks (IJCNN), 2018.

Egor Lakomkin, Mohammad Ali Zamani, Cornelius Weber, Sven Magg, Stefan Wermter

EmoRL:Real-time Acoustic Emotion Classification using Deep Reinforcement Learning
International Conference on Robotics and Automation (ICRA) - May 2018.

 Mohammad Ali Zamani, Sven Magg, Cornelius Weber, Stefan Wermter

Language-modulated Safer Actions using Deep Reinforcement Learning

ICRA PhD Forum - May 2018.

Mohammad Ali Zamani, Hadi Beik-Mohammadi, Matthias Kerzel,Sven Magg, Stefan Wermter
Learning Spatial Representation for Secure Human-Robot Collaboration in Joint Manual Tasks (pdf)

 ICRA workshop on the WORKplace is better with intelligent, collaborative, robot MATEs (WORKMATE) - May 2018.

Zamani, M.A., Magg, S., Weber, C., and Wermter, S. (2017, August).

Deep Reinforcement Learning using Symbolic Representation for Performing Spoken Language Instructions.

In 2nd Workshop on Behavior Adaptation, Interaction and Learning for Assistive Robotics (BAILAR) on Robot and Human Interactive Communication (RO-MAN), 2017 26th IEEE International Symposium on.

Short Curriculum Vitae