Attention-driven Robotic Manipulation (ARM)

Codebase of Q-attention (within the ARM system) and coarse-to-fine Q-attention (within C2F-ARM system) from the following papers:

• Q-attention: Enabling Efficient Learning for Vision-based Robotic Manipulation (ARM system)
• Coarse-to-Fine Q-attention: Efficient Learning for Visual Robotic Manipulation via Discretisation (C2F-ARM system)

Installation (on RLL servers)

conda create -n arm python=3.7 pytorch==1.8.1
conda activate arm 
# If starting from scratch: 
# 1. install CoppeliaSim: 
wget https://www.coppeliarobotics.com/files/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04.tar.xz 
tar -xvf CoppeliaSim_Edu_V4_1_0_Ubuntu18_04.tar.xz 
# NOTE version 4.1 and Ubuntu18_04!
# add to .bashrc:
export COPPELIASIM_ROOT=/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$COPPELIASIM_ROOT
export QT_QPA_PLATFORM_PLUGIN_PATH=$COPPELIASIM_ROOT
source .bashrc
conda activate arm

# 2. VirtualGL rendering 
sudo apt-get install xorg libxcb-randr0-dev libxrender-dev libxkbcommon-dev libxkbcommon-x11-0 libavcodec-dev libavformat-dev libswscale-dev
sudo nvidia-xconfig -a --use-display-device=None --virtual=1280x1024
wget https://sourceforge.net/projects/virtualgl/files/2.5.2/virtualgl_2.5.2_amd64.deb/download -O virtualgl_2.5.2_amd64.deb --no-check-certificate 
sudo dpkg -i virtualgl*.deb
sudo X
nohup sudo X &
export DISPLAY=:0.0

# 3. PyRep 
git clone [email protected]:stepjam/PyRep.git
pip install -r PyRep/requirements.txt
pip install PyRep/.
# make sure example scripts work:
python PyRep/examples/example_baxter_pick_and_pass.py

# 4. RLBench -> YARR -> ARM 
git clone [email protected]:stepjam/RLBench.git
pip install -r RLBench/requirements.txt
pip install RLBench/.


git clone [email protected]:MandiZhao/YARR.git # (notice forked repo)
cd YARR; git checkout dev; cd ..
pip install -r YARR/requirements.txt
pip install -e YARR/.


git clone [email protected]:rll-research/ARM.git
pip install -r ARM/requirements.txt
# (Mandi) Added data/ and log/ folder, changed default settings in yaml
cp -r -n /shared/mandi/rlbench_demo/*  ARM/data/
# test launch
python launch.py method=C2FARM rlbench.task=stack_wine framework.gpu=0

• Not 100% needed for install, save for later:

sudo apt install mesa-utils
# disable QT warning
export QT_LOGGING_RULES='*.debug=false;qt.qpa.*=false'
# clean up nvidia-smi
sudo fuser -v /dev/nvidia*
# then sudo kill -9 
pip install natsort pyquaternion 

• Errors:

1. Rendering:

(arm) mandi@pabamd3:~$ python PyRep/examples/example_baxter_pick_and_pass.py 
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-mandi'
libGL error: No matching fbConfigs or visuals found
libGL error: failed to load driver: swrast

Error: signal 11:

/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04/libcoppeliaSim.so.1(_Z11_segHandleri+0x2b)[0x7f232cb148db]
/lib/x86_64-linux-gnu/libc.so.6(+0x3ef20)[0x7f233b14af20]
/lib/x86_64-linux-gnu/libc.so.6(+0xb1646)[0x7f233b1bd646]
/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04/libcoppeliaSim.so.1(_Z15initGl_ifNeededv+0x116)[0x7f232cc97986]
/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04/libcoppeliaSim.so.1(_ZN11CMainWindowC2Ev+0x648)[0x7f232cf626f8]
/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04/libcoppeliaSim.so.1(_ZN3App16createMainWindowEv+0x46)[0x7f232cc89546]
/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04/libcoppeliaSim.so.1(_Z24simRunSimulator_internalPKciPFvvES2_S2_iS0_b+0x32d)[0x7f232cb29a1d]
/home/mandi/CoppeliaSim_Edu_V4_1_0_Ubuntu18_04/libcoppeliaSim.so.1(_Z29simExtLaunchUIThread_internalPKciS0_S0_+0x91)[0x7f232cb29e01]
/home/mandi/miniconda3/envs/arm/lib/python3.7/site-packages/pyrep/backend/_sim_cffi.cpython-37m-x86_64-linux-gnu.so(+0x5926e)[0x7f232d66526e]
python(_PyMethodDef_RawFastCallKeywords+0x1df)[0x5600e58a8def]

solve by running the sudo X section

ARM is trained using the YARR framework. Head to the YARR github page and follow installation instructions.

ARM is evaluated on RLBench 1.1.0. Head to the RLBench github page and follow installation instructions.

Now install project requirements:

pip install -r requirements.txt

# Data gen:
python ../RLBench/tools/dataset_generator.py --save_path=/home/mandi/ARM/data/ --tasks=take_lid_off_saucepan --image_size=128,128 --renderer=opengl --episodes_per_task=10 --variations=5 --processes=5


#train

python launch.py method=C2FARM rlbench.task=take_lid_off_saucepan rlbench.demo_path=/home/mandi/ARM/data framework.gpu=0

python launch.py method=C2FARM rlbench.task=phone_on_base rlbench.demo_path=/home/mandi/ARM/data framework.gpu=0 rlbench.demos=3

python launch.py rlbench.task=take_umbrella_out_of_umbrella_stand framework.gpu=0

Not that C2F-ARM is the better performing system.

useful thread for fetching from origin: https://stackoverflow.com/questions/5884784/how-to-pull-remote-branch-from-somebody-elses-repo

Running experiments

Be sure to have RLBench demos saved on your machine before proceeding. To generate demos for a task, go to the tools directory in RLBench (rlbench/tools), and run:

python dataset_generator.py --save_path=/mnt/my/save/dir --tasks=take_lid_off_saucepan --image_size=128,128 \
--renderer=opengl --episodes_per_task=100 --variations=1 --processes=1

Experiments are launched via Hydra. To start training C2F-ARM on the take_lid_off_saucepan task with the default parameters on gpu 0:

python launch.py method=C2FARM rlbench.task=take_lid_off_saucepan rlbench.demo_path=/mnt/my/save/dir framework.gpu=0