Image Based Visual Servoing

A simple IBVS simulation in pybullet built as a learning experiment as a part of my undergraduate research.

The initial code was built off of @roadrollerdafjorst's IBVS sim but has deviated since.

Instructions

• docs/env.yml contains the conda environment config file. Create an environment (here, called ibvs) and load it in as usual:

cd docs
conda env create -f env.yml
conda activate ibvs

• run the pybullet sim in DIRECT mode, with images saved to src/img/.

cd src
python main.py

• if required, convert the images into an MP4 file for later analysis

bash create_vid.sh

Demo

Target Image	Output GIF

TODO

• port to airobot to get rid of view matrix complexity
• use a better target and scene
• experiment with more complex dof.
  • Currently simply taking the first three points' (x, y) coords as our 6 dof.
• find out why the obstacles fall slowly in the p.DIRECT sim even though it does not seem to happen in p.GUI.