Getting to Grips with Model-Free Deep Reinforcement Learning for Dexterous Robotic Manipulation: Benchmarks, Analyses, Challenges
In order to have an exact replication of the robot used in this proposal, you will need the following materials:
| Material | Brand | Quantity |
|---|---|---|
| ServoMotors | Dynamixel XL-320 | 4 |
| WebCam | Standard | 1 |
| STL Files | link | |
| USD2 PHB | Dynamixel | 1 |
| U2D2 INT | Dynamixel | 1 |
Two state-space representations were used as state input for all the algorithms. Image represention and vector representation
The action-space controls the joint angles of each servomotors
For testing purposes, this robot is trained to complete a manipulation task, translation of an object (5cm cube).
Results using image representation
Results using vector representation
A video showing the robot executing the tasks during the training process can be found at: https://www.youtube.com/watch?v=ReDdVq_bQuQ
If you use either the code or data in your paper, please kindly star this repo and cite our paper
Cite this paper as: Coming soon
Please feel free to contact us or open an issue if you have questions or need additional explanations.