search

utiasDSL / safe-control-gym

PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and RL
https://www.dynsyslab.org/safe-robot-learning/
MIT License
636 stars 132 forks source link

Looking to reproduce a result from IROS 2022 competition paper #145

Closed SapanaChaudhary closed 2 months ago

SapanaChaudhary commented 10 months ago

Hello!

I was hoping to reproduce Fig. 2 (below) from your IROS 2022 competition paper titled A Remote Sim2real Aerial Competition: Fostering Reproducibility and Solutions’ Diversity in Robotics Challenges.

Screenshot 2024-01-06 at 12 08 25 PM

I had a few questions regarding the figure above:

  1. Which controller was used to produce the figure - traditional control vs RL?
  2. I looked into the beta-iros-competition branch, but, could only see implementations for ["circle", "square", "figure8"] trajectories. Does this mean if I were to try to reproduce the result above using PPO, I would have to implement that trajectory in https://github.com/utiasDSL/safe-control-gym/blob/beta-iros-competition/safe_control_gym/envs/benchmark_env.py#L491?

Thank you, and looking forward to your reply!

JacopoPan commented 10 months ago

@SapanaChaudhary those are results comparing the use of the Bitcraze Crazyflie firmware cmdFullState() interface in simulation and in real-life, @spencerteetaert might correct me but there should be several examples in folder dev-sim2real/

Federico-PizarroBejarano commented 2 months ago

Closing due to staleness. If you have any further issues @SapanaChaudhary feel free to let me know, either in a new issue or I can reopen this one.