Veröffentlichungen - Institut Informatik

anthropomorphic robot arm arm movement model artificial skin attractor dynamics approach Autonomous robotics BCI behavior generation direct physical interaction Dynamical systems force sensing grippers haptic interface human anatomy human motor behavior human perception inverse kinematics Machine Learning man machine interaction man-machine systems man-machine-interaction manipulators movement model object recognition operators gaze operators gesture pervasive design robot manipulator control robotic assistant robots gripper sensory channels 
Zeige alle
2021
1.
 Fidencio, Aline Xavier;  Glasmachers, Tobias;  Naro, Daniele
Application of Reinforcement Learning to a Mining System Proceedings Article 
In: 2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI), S. 000111–000118, 2021.
Abstract | Links | BibTeX | Schlagwörter: Control Applications, Industrial Application, Machine Learning, Machine learning algorithms, Mining Industry, Reinforcement learning
@inproceedings{fidencioApplicationReinforcementLearning2021,

title = {Application of Reinforcement Learning to a Mining System},

author = {Aline Xavier Fidencio and Tobias Glasmachers and Daniele Naro},

doi = {10.1109/SAMI50585.2021.9378663},

year  = {2021},

date = {2021-01-01},

urldate = {2021-01-01},

booktitle = {2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)},

pages = {000111--000118},

abstract = {Automation techniques have been widely applied in different industry segments, among others, to increase both productivity and safety. In the mining industry, with the usage of such systems, the operator can be removed from hazardous environments without compromising task execution and it is possible to achieve more efficient and standardized operation. In this work a study case on the application of machine learning algorithms to a mining system example is presented, in which reinforcement learning algorithms were used to solve a control problem. As an example, a machine chain consisting of a Bucket Wheel Excavator, a Belt Wagon and a Hopper Car was used. This system has two material transfer points that need to remain aligned during operation in order to allow continuous material flow. To keep the alignment, the controller makes use of seven degrees of freedom given by slewing, luffing and crawler drives. Experimental tests were done in a simulated environment with two state-of-the-art algorithms, namely Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC). The trained agents were evaluated in terms of episode return and length, as well as alignment quality and action values used. Results show that, for the given task, the PPO agent performs quantitatively and qualitatively better than the SAC agent. However, none of the agents were able to completely solve the proposed testing task.},

keywords = {Control Applications, Industrial Application, Machine Learning, Machine learning algorithms, Mining Industry, Reinforcement learning},

pubstate = {published},

tppubtype = {inproceedings}

}

Schließen
Automation techniques have been widely applied in different industry segments, among others, to increase both productivity and safety. In the mining industry, with the usage of such systems, the operator can be removed from hazardous environments without compromising task execution and it is possible to achieve more efficient and standardized operation. In this work a study case on the application of machine learning algorithms to a mining system example is presented, in which reinforcement learning algorithms were used to solve a control problem. As an example, a machine chain consisting of a Bucket Wheel Excavator, a Belt Wagon and a Hopper Car was used. This system has two material transfer points that need to remain aligned during operation in order to allow continuous material flow. To keep the alignment, the controller makes use of seven degrees of freedom given by slewing, luffing and crawler drives. Experimental tests were done in a simulated environment with two state-of-the-art algorithms, namely Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC). The trained agents were evaluated in terms of episode return and length, as well as alignment quality and action values used. Results show that, for the given task, the PPO agent performs quantitatively and qualitatively better than the SAC agent. However, none of the agents were able to completely solve the proposed testing task.
Schließen
doi:10.1109/SAMI50585.2021.9378663
Schließen
Impressum