AUVControl

Control of a hovering AUV simulated using HoloOcean. This implements the full robotics stack include various algorithms for state estimation, low-level control, and planning. HoloOcean is simple to use, and the HoveringAUV is easy to control, making this a perfect platform for experimenting and testing of various robotics algorithms.

Python Module

The code is organized into the python module auv_control. This consists of 3 submodules - estimation, control, and planning - along with various other helper functions. The following algorithms have been implemented:

• Base module
  • State class - wrapper class to standardize how our state is represented.
  • scenario - python dictionary with the example HoloOcean scenario to run.
• Estimation module
  • InEKF - Implementation of the invariant EKF with the IMU sensor as the process model and GPS, DVL, magnetometer, and pressure sensor as the update step.
• Control Module
  • LQR - Simple LQR controller using linearized state dynamics and euler angles. (Simple improvement to this would be something using Lie Groups)
• Planning Module
  • RRT - Loads up various obstacles in the environment, and plans a trajectory from one side of the environment to the other using RRT.
  • Traj - Various predefined trajectories to test controllers on.

Running

To run the simulation, first install all dependencies

pip install -r requirements.txt

Then simply run the script

using Demo_sac.ipynb
or
python run_sac.py