A collective codebase for the simulation of an autonomous delivery drone as a part of my submissions for the e-Yantra Robotics Competition organised by IIT Bombay.
- ROS Melodic
- Gazebo
- OpenCV
- pyzbar library
This task includes 2 basic cascading PID scripts to control the attitude (roll, pitch, yaw) and the position (latitude, longitude) of the drone.
In this task, the drone detects the coordinates of the delivery address through the QR Code on the parcel. The drone then goes to that address and drops the package while avoiding obstacles in the way.
In this task, the drone goes to multiple specified locations and executes a search pattern to look for a specific landing marker. After finding the landing marker, the drone hovers over it and then goes to the next location.
In this task, the drone computes the address coordinates for every parcel on the grid square from a manifest document. The drone then delivers them on the closest landing marker at those coordinates.
Task 5 is similar to Task 4 but the manifest documents also contains return parcels that mustbe delivered back to the return grid square. The drone uses a script called Task_5_schedule.py to compute the best order of returns and deliveries such that the drone is flying empty for less time and maximum earnings are acquired in a given time.
Task 6 is the same as Task 5 but with a different manifest to test the performance of the scheduling algorithm.
- Clone repository in the root directory
- On the terminal execute:
catkin build sudo echo 'source ~/delivery_drone/devel/setup.bash' >> ~/.bashrc - Execute launch file for whichever task for example:
roslaunch vitarana_drone task_1.launch
- The drone crashes at random moments on some runs and works perfectly fine on most other runs. This is due to inherent issues with the drone model and Gazebo.
- Drone could be tuned to move faster by tuning the PID of the attitude and position controller scripts better.