Space Non-Cooperative Object Active Tracking, which means the spacecraft approach to an unknown target only with vision camera. We propose an end-to-end active visual tracking method based on deep Q-learning algorithm, named as DRLAVT. It can guide the chasing spacecraft approach to arbitrary space non-cooperative target merely relied on color or RGBD images, which significantly outperforms PBVS method.
See more details about SNCOAT benchmark in Space Non-Cooperative Object Active Tracking With Deep Reinforcement Learning.
If you use our benchmark or related codes, please cite:
@article{zhou2022space,
title={Space Non-cooperative Object Active Tracking with Deep Reinforcement Learning},
author={Zhou, Dong and Sun, Guanghui and Lei, Wenxiao and Wu, Ligang},
journal={IEEE Transactions on Aerospace and Electronic Systems},
year={2022},
publisher={IEEE}
}
MORE SOURCE CODES ARE COMING SOON ...
- PyTorch ~= 1.6.0
- CoppeliaSim v4.2
- PyRep
- gym
- matplotlib
[Note]: This Program only validated on Ubuntu16.04 and Centos7 platform.
Download scenes at first: Googel Drive | Baidu NetDisk(code:1111)
We construct 18 scenes with different types of space non-cooperative object, including asteroids, capsules, rockets, satellites, and stations.
- SNCOAT-Asteroid-(v0-v5)
- SNCOAT-Capsule-(v0-v2)
- SNCOAT-Rocket-(v0-v2)
- SNCOAT-Satellite-(v0-v2)
- SNCOAT-Station-(v0-v2)
The list of Trackers have been validated in our PBVS framework:
- SiamRPN
- SiamFC
- KCF