Two simulated robots map a harbor, share what they find and locate a target together. Undergraduate research at ICMC-USP (CNPq, 2015–2016), built as services on RefSORS, the reference architecture for service-oriented robotic systems from L. B. R. de Oliveira's PhD thesis.
The code is from 2016 and targets ROS Indigo with Gazebo 2. It is kept as it was, apart from portable map paths and package metadata.
-
Live simulation. Build and run the 2016 code in Docker, then watch the mission in a 3D view you can orbit, zoom, pan and point at either robot:
docker build --platform linux/amd64 -f sim/Dockerfile -t refsors-sim . && docker run --rm --name refsors-sim --platform linux/amd64 -p 127.0.0.1:8080:8080 -p 127.0.0.1:6080:6080 refsors-sim
Open http://127.0.0.1:8080. The Gazebo window itself is at http://127.0.0.1:6080/vnc_auto.html?host=127.0.0.1&port=6080. Details and requirements are in sim/README.md.
-
Static viewer. Open viewer/index.html for the mission, node graph, saved maps, world, history and a recorded run, with no ROS or Docker. See viewer/README.md.
- A Husky (larger, Kinect) and a Jackal (smaller, camera) start in a harbor world full of containers.
- Both explore the free areas, each building an occupancy grid from its laser and sharing it through a common knowledge service so they do not search the same area twice.
- Each robot watches its camera for the target, a red object found by HSV thresholding.
- The first robot to see it ends the search for both. If it is the Jackal, the Husky plans a path to the target with A* while the Jackal leaves. The Husky then aligns with the target and both robots exit.
Each robot runs its own set of nodes, under the larger_robot or
smaller_robot namespace:
| Node | Role |
|---|---|
localization |
Publishes the robot's pose from Gazebo's model states |
laser |
Turns the laser scan into the measures the other services use |
mapping |
Builds the robot's occupancy grid (createMap action) |
navigation |
Explores with a force-based occupancy grid, drives to points and aligns with the target (search, driveTo, alignWithBomb actions) |
image_processing |
Detects the target and reports its displacement (processImage action, getBombDisplacement service) |
control |
Coordinates the robot's services for each mission step (exit action) |
Shared by both robots:
| Node | Role |
|---|---|
knowledge |
Shared map, robot positions and the next unexplored goal (addToMap, getMap, getPositions, getNewGoal services) |
path_planning |
Global paths with A* (defineGlobalPath service) |
application |
Runs the mission and hands work from one robot to the other |
Messages, services and actions are in msg, srv and action. Worlds are in world; maps written during a run go to maps.
Needs ROS Indigo (Ubuntu 14.04) with gazebo_ros, husky_gazebo,
jackal_description and jackal_control.
mkdir -p ~/catkin_ws/src && cd ~/catkin_ws/src
git clone https://github.com/rafamarqqqs/refsors-robots.git robot_control
cd ~/catkin_ws && catkin_make && source devel/setup.bash
roslaunch robot_control master.launchMIT. RefSORS is described in L. B. R. de Oliveira's thesis; this repository contains only the robot system built on it.