colinkohler / bulletarm Goto Github PK

BulletArm is a benchmark and learning-environment for robotic manipulation.

License: MIT License

Python 99.76% HTML 0.22% CMake 0.02%

bulletarm's Introduction

Documentation | Paper | License

BulletArm is a benchmark and learning-environment for robotic manipulation. BulletArm provides a set of standardized benchmark tasks in simulation alongside a collection of baseline algorithms.

Announcements
Installation
Getting Started
Contributions
Cite

Announcements

June 24, 2022

Version 1.0.0 is live!

Install

Clone this repo

git clone https://github.com/ColinKohler/BulletArm.git
cd BulletArm

Install dependencies
```
pip install -r requirements.txt 
```

Install this package

pip install .

Or add to your PYTHONPATH

export PYTHONPATH=/path/to/BulletArm/:$PYTHONPATH

Run the block stacking demo

python tutorials/block_stacking_demo.py

Getting Started

To get started with BulletArm please see the Getting Started section of the documentation.

Contributions

We hope that other researchers will help us extend this package by adding new tasks, robots, etc. Please see the Contributions section of the documentation for additional details.

Cite

The development of this package was part of the work done for our paper: BulletArm: An Open-Source Robotic Manipulation Benchmark and Learning Framework. Please cite this work if you use our code:

@misc{https://doi.org/10.48550/arxiv.2205.14292,
      doi = {10.48550/ARXIV.2205.14292},
      url = {https://arxiv.org/abs/2205.14292},
      author = {Wang, Dian and Kohler, Colin and Zhu, Xupeng and Jia, Mingxi and Platt, Robert},
      keywords = {Robotics (cs.RO), FOS: Computer and information sciences, FOS: Computer and information sciences},
      title = {BulletArm: An Open-Source Robotic Manipulation Benchmark and Learning Framework},
      publisher = {arXiv},
      year = {2022},
      copyright = {arXiv.org perpetual, non-exclusive license}
}

Feel free to contact us.

License

BulletArm is distributed under GNU General Public License. See License.

bulletarm's People

Contributors

Stargazers

Watchers

Forkers

zxp-s-works angelazhx pointw xxs90 saulbatman kamilkurach zihesong tinzzyli hai-h-nguyen dian-bdai gyyer karmenlu yctan70 chamzaskonstantinos tw-ilson arsht

bulletarm's Issues

tkinter error

Hi, I usually get Tkinter problems as below when training. After this happened, I had to reboot my device to free GPU memory cause I could not find the process in the GPU processes table to kill.

100%|████████████████████████████████████████████████| 100/100 [01:16<00:00, 1.31it/s] Action Step:4015; Episode: 992.0; Reward:0.520; Eval Reward:0.000; Explore:0.01; Loss:0Exception ignored in: <function Image.__del__ at 0x7f01fad40440>00:03<00:41, 2.36it/s] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 3507, in __del__ self.tk.call('image', 'delete', self.name) RuntimeError: main thread is not in main loop Action Step:4045; Episode: 1000.0; Reward:0.480; Eval Reward:0.000; Explore:0.01; LossException ignored in: <function Variable.__del__ at 0x7f01fad239e0>:12<01:14, 1.19it/s] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4050; Episode: 1001.0; Reward:0.470; Eval Reward:0.000; Explore:0.01; Loss:Exception ignored in: <function Variable.__del__ at 0x7f01fad239e0> Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4055; Episode: 1001.0; Reward:0.470; Eval Reward:0.000; Explore:0.01; LossException ignored in: <function Variable.__del__ at 0x7f01fad239e0>:14<01:40, 1.15s/it] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4060; Episode: 1001.0; Reward:0.470; Eval Reward:0.000; Explore:0.01; Loss:Exception ignored in: <function Variable.__del__ at 0x7f01fad239e0> Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4070; Episode: 1005.0; Reward:0.450; Eval Reward:0.000; Explore:0.01; Loss:Exception ignored in: <function Image.__del__ at 0x7f01fad40440>00:15<01:26, 1.01s/it] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 3507, in __del__ self.tk.call('image', 'delete', self.name) RuntimeError: main thread is not in main loop Action Step:4100; Episode: 1010.0; Reward:0.410; Eval Reward:0.000; Explore:0.01; LossException ignored in: <function Variable.__del__ at 0x7f01fad239e0>:25<01:22, 1.07s/it] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4105; Episode: 1011.0; Reward:0.410; Eval Reward:0.000; Explore:0.01; LossException ignored in: <function Variable.__del__ at 0x7f01fad239e0>:26<01:17, 1.02s/it] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4110; Episode: 1011.0; Reward:0.410; Eval Reward:0.000; Explore:0.01; Loss:Exception ignored in: <function Variable.__del__ at 0x7f01fad239e0> Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4115; Episode: 1012.0; Reward:0.410; Eval Reward:0.000; Explore:0.01; Loss:Exception ignored in: <function Variable.__del__ at 0x7f01fad239e0>27<01:20, 1.07s/it] Traceback (most recent call last): File "/home/huy/anaconda3/envs/bullet-arm/lib/python3.7/tkinter/__init__.py", line 332, in __del__ if self._tk.getboolean(self._tk.call("info", "exists", self._name)): RuntimeError: main thread is not in main loop Action Step:4120; Episode: 1012.0; Reward:0.410; Eval Reward:0.000; Explore:0.01; Loss:./run.sh: line 4: 70195 Aborted (core dumped) python bulletarm_baselines/fc_dqn/scripts/main_goal.py --algorithm=dqn --architecture=equi_asr --env=block_stacking --fill_buffer_deconstruct --planner_episode=5

Gym wrapper -- action_space

Hello,
Thanks for the awesome work and open-sourcing the code! Appreciate it.

I was wondering if there was a gym wrapper for the environment? I was specifically trying to generate random episodes which would involve (in a gym environment) something like obs, *_ = env.step(env.action_space.sample()).

What would be the equivalent to this?

Thanks!

Edit: @ColinKohler -- please look at this whenever you get a chance. Thanks!

single process bugs

when run /bulletarm_baselines/fc_dqn/scripts/main.py with num_process = 0, some errors relate to tensor shape occurs. For example:

Traceback (most recent call last):
File "bulletarm_baselines/fc_dqn/scripts/main.py", line 288, in
train()
File "bulletarm_baselines/fc_dqn/scripts/main.py", line 157, in train
planner_actions_star_idx, planner_actions_star = agent.getActionFromPlan(plan_actions)
File "/BulletArm/bulletarm_baselines/fc_dqn/agents/agents_3d/base_3d.py", line 20, in getActionFromPlan
x = plan[:, 0:1]
IndexError: too many indices for tensor of dimension 1

here plan_actions.shape is torch.Size([4]) while expected to be torch.Size([1, 4]) in getActionFromPlan

about tutorials

Thanks for your excellent work，in the tutorials，baseline_demo.py ,creating_a_new_robot.py is empty，when you update？

Missing cv2 in requirements.txt

To reproduce:

Follow installation in a fresh conda environment: https://colinkohler.github.io/BulletArm/api/getting_started.html

Solved also by running: pip install -r baseline_requirements.txt

how to get image observation while running headless

Hi, I wonder is it possible to get image observation of the environment while running headless?

Thanks!

EOFError raised when creating environment

really good framework for robot manupulation!
but there are some environment files with no init.py file that casued issue
now I'm facing issue with
File "/home/usr1/bash/envs/bulletarm/lib/python3.9/multiprocessing/connection.py", line 388, in _recv
raise EOFError
for both open loop and close loop environments Benchmarks, any ideas how to fix this? thanks a lot

Infinite loop in tutorial train_dummy_agent.py

num_episodes is never incremented in https://github.com/ColinKohler/BulletArm/blob/main/tutorials/train_dummy_agent.py, resulting in an infinite loop.

Also, I was not able to ever trigger done_idxs = np.nonzero(dones)[0]. Perhaps I did not wait long enough or there are some issues with how done is flagged.

Why does the test Block Stacking Demo fail to run after installation?

Whether I ran it directly in Pycharm or typed: python tutorials/block_stacking_demo.py on the terminal, the result would look like this