English | 中文文档
robopal is an open source robot simulation framework based on MuJoCo, mainly used for deep reinforcement learning training and Control algorithm verification. A variety of control schemes and example environments are provided within the framework.
Compared with other frameworks, ours has the following advantages:
- Higher motion accuracy, it is more in line with the motion in the real environment
- High portability, more concise code, easy to learn and use
- Rich task environment, such as compliant control, motion imitation, visual servoing, etc.
- Windows (10+) / Linux
- Python 3.9+
- MuJoCo-2.3.7
- pinocchio 2.6.20
$ git clone https://github.com/NoneJou072/robopal
$ cd robopal
$ pip install -e .
The development version mainly includes the following controllers:
- Joint Space Controller
Use Ruckig to plan joint positions, velocities, accelerations for smooth motion, and calculate torques based on pd control and dynamics - Cartesian Space Controller
Position/rotation control in Cartesian space based on pd control and kinematics