Scavenger Hunt

Complete codebase for BWI's Scavenger Hunt project.

website/ contains everything relating to the website. All other directories are ROS packages for Scavenger Hunt robots. Some important ones:

scavenger_hunt_api/ - The Scavenger Hunt API. Contains utilities for connecting to the website, uploading proofs, and querying proof feedback. Non-ROSified, though there is an optional ROS wrapper.
bwi_scavenger/ - Drivers for our Scavenger Hunt participant robots.
darksocket/ - Tool for offboard darknet computing.

Installing

In the src/ folder of a catkin workspace (Note from Saeid: remove the older bwi_scavenger package and also delete the build/ and devel/ folder for fresh compilation):

$ git clone https://github.com/bu-air-lab/scavenger_hunt_api.git --recursive

build using catkin tools; $ catkin build in the root of your catkin workspace.

Using the Scavenger Hunt API

The scavenger_hunt_api folder of this repository of the same name contains ROS wrappers for interacting with the Scavenger Hunt website so you can submit proofs and participate in Hunts.

This leaves the implementation of how to carry out the hunts up to the individual.

Please see the API webpage for details on the API.

From Saeid

First localize thge robot using:

roslaunch bwi_launch segbot_bu

And then go to the API website, and create a hunt and tasks. Update the hunt name in bu.cpp file. To let the robot start the scavenger hunt, run:

roslaunch scavenger_hunt scavenger_hunt.launch
rosrun scavenger_hunt server.py
rosrun scavenger_hunt bu_node

TO DO:

Task execution strategy is in server.py. Current code receives the door names o visit locations. There is room to improve it. Also, to find objects, we need to add YOLO or other tolls and a strategy to find food. Robot losing localization at the goal pose also is problematic sometimes. We should ge rid of using the user input to pause the code.

BWI Scavenger Startup

The following sections outlines what is required to run the full Scavenger Hunt codebase on the BWIBot platforms. The BWI Scavenger code may serve as an example for how to implement a scavenger hunt, fetching and starting Hunts, and submitting proofs.

There are two options for image recognition: offloading the image processing to an offboard computer if the robot does not have a GPU, or running YOLO directly on the robot computer if a GPU is available.

In the case that a GPU is not available, darksocket is used to facilitate data flow.

Initial Configuration

Verify launch parameters in bwi_scavenger/config/bwi_scavenger.yaml
- scav_hunt_login - Valid website credentials
- ws_path - Fully-qualified path to the workspace containing these packages
- world - Either sim if running in Gazebo or irl if running in the lab
- topics/perception/xxx - Camera topics should be prefixed with /nav_kinect for IRL and /camera for simulation
Verify launch parameters in scavenger_hunt_api/config/scavenger_hunt.yaml
- website/domain - Domain of Scavenger Hunt website (until the web server is up this is localhost for lab machines and localhost:8080 for robots)
- scratch_path - Valid path that ROS has permission to create files in (we usually use the workspace path)
Verify darksocket config in darksocket/scripts/darksocket.conf
- Robot and offboard machine
  - host - IP of robot (find with hostname -I; should be the second IP printed)
  - port - we usually use 51820
- Offboard machine only
  - darknet_path - Location of darknet binary (built with GPU=1, otherwise you're undermining the purpose of darksocket)
  - cfg_path - Fully-qualified path of cfg file
  - weights_path - Fully-qualified path of weights file
  - meta_path - Fully-qualified path of meta file (which in turn contains the fully-qualified path of the names file)
  - recv_img_path - Fully-qualified path to store a received image in (including image file name)
Verify darksocket launch parameters in darksocket/launch/darksocket_ros.launch
- topics/camera - Camera topics should be prefixed with /nav_kinect for IRL and /camera for simulation
- conf_path - Fully-qualified path to darksocket.conf

Robot Startup

Boot up the scavenger robot. Before any ROS commands, open an SSH tunnel on the robot so it can see the website. This is required only for instances where darksocket is used to offload image processing to another machine.

ssh -D 8080 http://utw10989.utweb.utexas.edu/

Navigate to the workspace and run the following launch files.

roslaunch bwi_launch segbot_v2.launch
roslaunch bwi_scavenger scavenger.launch
rosrun bwi_scavenger absim_ros.py
python3 src/bwi_scavenger/scripts/absim_socket.py

Run darksocket on the offboard machine. If everything was set up correctly, it should load the network and then print "connection established." Again, only required if using darksocket.

python darksocket.py client

Localize the robot in rviz and start a hunt via the do_hunt node.

rosrun bwi_scavenger do_hunt "BWI Lab Hunt"

BWI Lab Hunt is the name of our test hunt, though any hunt registered on the website will work.

bu-air-lab/scavenger_hunt_api