Spot-Sim2Real is a modular library for development of Spot for embodied AI tasks (e.g., Language-guided Skill Coordination (LSC), Adaptive Skill Coordination (ASC)) -- configuring Spot robots, controlling sensorimotor skills, and coordinating Large Language Models (LLMs).
Please refer to the setup instructions page for information on how to setup the repo. Note that this repo by-default does not track dirty status of submodules, if you're making any intentional changes within the third-party packages be sure to track them separately.
Computer can be connected to the robot in one of the following modes.
- Ethernet (Gives best network speed, but it is cluttery :sad: )
This mode can be used to create a wired connection with the robot. Useful for teleoperating the robot via computer - Access Point Mode
This is a wireless mode where robot creates its wifi network. Connect robot to this mode for teleoperating it using controller over long distances. Robot is in Access Point mode if you see a wifi with name likespot-BD-***********
(where * is a number) - Client Mode (Gives 2nd best network speed, we usually prefer this)
This is a wireless mode where robot is connected to an external wifi network (from a nearby router). Computer should be connected to this same network, wired connection between router and computer will be faster than wireless connection.
Follow the steps from Spot's Network Setup page by Boston Dynamics to connect to the robot.
After setting up spot in correct network configuration, please add its IP inside bashrc
echo "export SPOT_IP=<spot's ip address>" >> ~/.bashrc
source ~/.bashrc
Test and ensure you can ping spot
ping $SPOT_IP
If you get response like this, then you are on right network
(spot_ros) user@linux-machine:~$ ping $SPOT_IP
PING 192.168.1.5 (192.168.1.5) 56(84) bytes of data.
64 bytes from 192.168.1.5: icmp_seq=1 ttl=64 time=8.87 ms
64 bytes from 192.168.1.5: icmp_seq=2 ttl=64 time=7.36 ms
Before starting to run the code, you need to ensure that all ROS env variables are setup properly inside bashrc. Please follow the steps from Setting ROS env variables for proper ROS env var setup.
Go to the repository
cd /path/to/spot-sim2real/
The code for the demo lies inside the main
branch.
# Check your current git branch
git rev-parse --abbrev-ref HEAD
# If you are not in the `main` branch, then checkout to the `main` branch
git checkout main
- Since we do not have a physical emergency stop button (like the large red push buttons), we need to run an e-stop node.
python -m spot_wrapper.estop
- Keep this window open at all the times, if the robot starts misbehaving you should be able to quickly press
s
orspace_bar
to kill the robot
- Ensure you have the Estop up and running in one terminal. Follow these instructions for e-stop
- Run keyboard teleop with this command in a new terminal
spot_keyboard_teleop
-
Before running scripts on the robot, waypoints should be recorded. These waypoints exist inside file
spot-sim2real/spot_rl_experiments/configs/waypoints.yaml
-
Before recording receptacles, make the robot sit at home position then run following command
spot_reset_home
-
There are 2 types of waypoints that one can record,
- clutter - These only require the (x, y, theta) of the target receptacle
- place - These requre (x, y, theta) for target receptable as well as (x, y, z) for exact drop location on the receptacle
-
To record a clutter target, teleoperate the robot to reach near the receptacle target (using joystick). Once robot is at a close distance to receptacle, run the following command
spot_rl_waypoint_recorder -c <name_for_clutter_receptacle>
-
To record a place target, teleoperate the robot to reach near the receptacle target (using joystick). Once robot is at a close distance to receptacle, use manipulation mode in the joystick to manipulate the end-effector at desired (x,y,z) position. Once you are satisfied with the end-effector position, run the following command
spot_rl_waypoint_recorder -p <name_for_place_receptacle>
-
In a new terminal, run the executable as
spot_rl_launch_local
This command starts 4 tmux sessions\n
- roscore
- img_publishers
- proprioception
- tts
-
You can run
tmux ls
in the terminal to ensure that all 4 tmux sessions are running. You need to ensure that all 4 sessions remain active until 70 seconds after running thespot_rl_launch_local
. If anyone of them dies before 70 seconds, it means there is some issue and you should rerunspot_rl_launch_local
. -
You should try re-running
spot_rl_launch_local
atleast 2-3 times to see if the issue still persists. Many times roscore takes a while to start due to which other nodes die, re-running can fix this issue. -
You can verify if all ros nodes are up and running as expected if the output of
rostopic list
looks like the following(spot_ros) user@linux-machine:~$ rostopic list /filtered_hand_depth /filtered_head_depth /hand_rgb /mask_rcnn_detections /mask_rcnn_visualizations /raw_hand_depth /raw_head_depth /robot_state /rosout /rosout_agg /text_to_speech
-
If you don't get the output as follows, one of the tmux sessions might be failing. Follow the debugging strategies described in ISSUES.md for triaging and resolving these errors.
-
This is the image visualization tool that helps to understand what robot is seeing and perceiving from the world
spot_rl_ros_img_vis
-
Running this command will open an image viewer and start printing image frequency from different rosotopics.
-
If the image frequency corresponding to
mask_rcnn_visualizations
is too large and constant (like below), it means that the bounding box detector has not been fully initialized yetraw_head_depth: 9.33 raw_hand_depth: 9.33 hand_rgb: 9.33 filtered_head_depth: 11.20 filtered_hand_depth: 11.20 mask_rcnn_visualizations: 11.20 raw_head_depth: 9.33 raw_hand_depth: 9.33 hand_rgb: 9.33 filtered_head_depth: 11.20 filtered_hand_depth: 8.57 mask_rcnn_visualizations: 11.20 raw_head_depth: 9.33 raw_hand_depth: 9.33 hand_rgb: 9.33 filtered_head_depth: 8.34 filtered_hand_depth: 8.57 mask_rcnn_visualizations: 11.20
Once the
mask_rcnn_visualizations
start becoming dynamic (like below), you can proceed with next stepsraw_head_depth: 6.87 raw_hand_depth: 6.88 hand_rgb: 6.86 filtered_head_depth: 4.77 filtered_hand_depth: 5.01 mask_rcnn_visualizations: 6.14 raw_head_depth: 6.87 raw_hand_depth: 6.88 hand_rgb: 6.86 filtered_head_depth: 4.77 filtered_hand_depth: 5.01 mask_rcnn_visualizations: 5.33 raw_head_depth: 4.14 raw_hand_depth: 4.15 hand_rgb: 4.13 filtered_head_depth: 4.15 filtered_hand_depth: 4.12 mask_rcnn_visualizations: 4.03 raw_head_depth: 4.11 raw_hand_depth: 4.12 hand_rgb: 4.10 filtered_head_depth: 4.15 filtered_hand_depth: 4.12 mask_rcnn_visualizations: 4.03
-
This is an important step. Ensure robot is at its start location and sitting, then run the following command in a new terminal
spot_reset_home
-
The waypoints that were recorded are w.r.t the home location. Since the odometry drifts while robot is moving, it is necessary to reset home before start of every new run
- Follow the steps described in e-stop section
-
Ensure you have correctly added the waypoints of interest by following the intructions to record waypoints
-
In a new terminal you can now run the code of your choice
-
To run Sequencial experts
spot_rl_mobile_manipulation_env
-
To run Adaptive skill coordination
spot_rl_mobile_manipulation_env -m
-
To run Language instructions with Sequencial experts, ensure the usb microphone is connected to the computer
python spot_rl_experiments/spot_rl/envs/lang_env.py
-
-
If you are done with demo of one of the above code and want to run another code, you do not need to re-run other sessions and nodes. Running a new command in the same terminal will work just fine. But make sure to bring robot at home location and reset its home using
spot_reset_home
in the same terminal
To run the current spot-aria code, follow these steps
# Go to the aria directory
cd aria_data_loaders/
# Run the aria_streamer.py file
python aria_streamer.py --data-path="/path/to/directory/containing/vrs/and/mps/files" --vrs-name="name_of_vrs_file" --dry-run=False --verbose=True
- Args:
--data-path
(Input: string): This is the path to the directory where VRS & MPS files reside. Make sure your VRS & MPS files for a trajectory co-exist in the same folder.vrs-file
(Input: string): This is the name of the VRS file. Ignore the.vrs
in the name of the file. For eg. if the name of VRS file issample_trajectory.vrs
, then the input will be like--vrs-file="sample_trajectory"
. This file will be searched for within the directory specified with--data-path
--dry-run
(Input: bool): This is used to indicate if it should command robot to move or not. If dry run is true, then it will not send the commands to the robot to make it move. But this will still require you to have robot watching the QR code for localizing itself in the system.--verbose
(Input: bool): This flag is for updating the visual stream with detections from both QR & object detectors.
We thank Naoki Yokoyama for setting up the foundation of the codebase, and Joanne Truong for polishing the codebase. Spot-Sim2Real is built upon Naoki's codebases: bd_spot_wrapper and spot_rl_experiments , and with new features (LLMs, pytest) and improving robustness.
If you find this repository helpful, feel free to cite our papers: Adaptive Skill Coordination (ASC) and Language-guided Skill Coordination (LSC).
@article{yokoyama2023adaptive,
title={Adaptive Skill Coordination for Robotic Mobile Manipulation},
author={Yokoyama, Naoki and Clegg, Alexander William and Truong, Joanne and Undersander, Eric and Yang, Tsung-Yen and Arnaud, Sergio and Ha, Sehoon and Batra, Dhruv and Rai, Akshara},
journal={arXiv preprint arXiv:2304.00410},
year={2023}
}
@misc{yang2023adaptive,
title={LSC: Language-guided Skill Coordination for Open-Vocabulary Mobile Pick-and-Place},
author={Yang, Tsung-Yen and Arnaud, Sergio and Shah, Kavit and Yokoyama, Naoki and Clegg, Alexander William and Truong, Joanne and Undersander, Eric and Maksymets, Oleksandr and Ha, Sehoon and Kalakrishnan, Mrinal and Mottaghi, Roozbeh and Batra, Dhruv and Rai, Akshara},
howpublished={\url{https://languageguidedskillcoordination.github.io/}}
}
Spot-Sim2Real is MIT licensed. See the LICENSE file for details.