Pixelate 2021

Computer Vision based event organised by Robotics Club of IIT (BHU) Varanasi

Problem Statement

• There is a 12*12 Grid made up of cells of different colors white, black, blue, green, yellow, red, pink.
• There are exactly 2 pink tiles one with a dark blue circle and the other, a dark blue square on top of it. There are exactly 2 light blue tiles one with a dark blue circle and the other, a dark blue square on top of it.
• Circle on pink tile represents a covid patient and circle on a blue tile represents covid hospital while similarly the square corresponds to non covid patients and hospitals.
• Each colored tile has a certain cost with it red-4, yellow-3, green-2, white-1, black-invalid cell.
• There are certain cells which can be entered from only one direction specified by a red triangle on top of it.
• Basically our bot has to go to each patient and take it to the corresponding hospital(covid/non-covid) while taking the valid path with minimum cost.
• The bot has an ArUco marker on top of it and we are given a static overhead camera feed.

The Pybullet Arena

Approach

• We used Computer Vision OpenCV functions to read the camera image and preprocess the grid data by creating a custom mask for every colour and segmenting it out and then applying shape detection on it to detect a square, circle and triangle.
• Following that a 12*12 adjacency matrix was made representing a directed weighted graph.
• The minimum path to each patient is caluclated using Djikstra's Algorithm, and then the bot is made to move towards it.
• After picking up the patient, we check whether the patient is covid/non-covid by removing the pink cover and then he is taken to the appropriate hospital, again following the minimum path.
• The program terminates after both the patients have been taken to their respective hospitals.
• For movement of the bot, on every cell the camera feed is called and the position of the bot and its orientation are detected using OpenCV functions, based on which the Proportional Controller calculates the torque to be given, to each wheel.

Working

You can check out the final video of the Pybullet simulation here.

Setup

• Clone the repo and create a virtual environment and activate it using

python3 -m venv venv
source venv/bin/activate

• Then execute the folowing command

pip install -e pixelate_arena

• Execute our solution b11.py using

python3 b11.py

Team B11

Sandeepan Ghosh

Raghavansh Singla