Project Type: Course
Course: Pitt Senior Design
Robot Used: Custom Raspberry Pi Mobile Robot
Software Used: Python | ROS | OpenCV
Date(s): Sep 2017 - Dec 2017
Forest fires are becoming a huge issue that have the ability to destroy countless lives and property. The key to mitigating the destructuive
force of forest fires is early detection and response. Autonomous drones that can continuously scan remote, inaccessible regions for the smallest sign of a
fire would be invaluable. An additional key feature of these drones would be to carry water tanks to swarm around the fire and begin extinguishing it before
human firefighters can reach the area.
With this project we aimed to research the use of swarm robotic motion algorithms in the detection and response to forrest fires. To simplify
the problem we used small mobile ground robots to replace large drones, a flashlight to replace a fire, and an eye-in-the-sky vision system to replace GPS and
provide location data to each robot. However, many of the lesons and motion planning software developed for this system can be extrapolated for drone
use in a real-world larger setting.
One of my main responsibilties for this project was writing the low-level software for each individual robot.
This included sampling data from the onboard light sensors and controlling each individual motor
in order to carryout the received velocity commands.
My second main task was to develop the eye-in-the-sky "GPS" system. An Xbox Kinect sensor was suspended above the
work area of the robots capturing RGBD data of the entire environment. I processed the RGB image using the OpenCV library to calculate
a pixel location for the center of each robot. The depth image was used to convert these to real-world position coordinates. These coordinates
were then sent over the ROS network to each robot. In a large scale implementation of this project the vision system could easily be
swapped out with a GPS chip as long as the GPS driver publishes the same position message type.
Address
Phone #
Boston, MA
781 812 8630
joelynch523@gmail.com