A distributed formation control scheme based on epipolar geometry.
The instructions here provided were tested in Ubuntu 20.04 and 22.04 operating systems, in an x86 architecture. To run the demo, please follow the steps below:
- Make sure you have Python Poetry and Python 3.8 installed
- Clone this repository to your local home folder
cd ~
git clone git@github.com:KTH-DHSG/epic.git- Run
poetry installinside the cloned folder
cd ~/epic
poetry install- Run the demo!
To run the provided example, run:
cd ~/epic
poetry run python demo/multi_agent_dynamic.pyThis will run a simulation of a group of 6 agents in a simulated environment with a randomized initial and target pose in the vicinity of the poses in the file demo/6_agent_geometry.json.
In the first example, we show the simulated formation control of a group of 6 agents in a simulated environment. We consider camera types with different distortion models.
In the second example, we show the real formation control of a group of 3 agents. The leader agent, to the left, is manually teleoperated, while the followers are controlled by the EpiC framework. The agents are equipped with monocular Flir Blackfly USB 3.0 Cameras.
Here follow some details of the experimental conditions observed in the videos. For any other question, reach out to the repository maintainer:
- Features: SIFT
- Feature Matching: SIFT descriptors with RANSAC geometric filtering
- Cameras:
- Leader: FLIR Blackfly USB 3
- Follower 1: FLIR Blackfly USB 3
- Follower 2: IMX323 USB Camera
- Resolution used: 640 x 480 px
- Target Formation (position with respect to leader)
- Leader: [0.0, 0.0, 0.0, −0.5235, −0.4628, 0.4753, 0.5345]
- Follower 1: [−0.1136, 1.5824, 0.0370, −0.4499, −0.5410, 0.5492, 0.4506]
- Follower 2: [−1.0820, 0.7141, 0.0190, 0.5003, 0.5035, −0.4985, −0.4976]
- Format: [px py pz qx qy qz qw] (position x,y,z, quaternion x,y,z,scalar)
- Onboard Computers: Intel NUC Core i3 16GB RAM (2020 model)