DDPG based disinfectant robot

In India the main problem is population density (382/sq km). In such a country this virus can spread like a wildfire, but due to the lock down it has slowed down. As businesses reopen the spread of this disease poses a great threat to the people.

So, we designed a simple yet effective robot that attempts to disinfect the entire room.

If we make a remote controlled robot we need to assign a person specifically for the task which will become too monotonous and tedious for the operator. This is the reason why we adopted reinforcement learning to build the bot.

The idea revolves around the simple concept pixel to action. So this model takes in the pixels(image) directly and gives out the possible actions.

The model we are going to use is a DDPG with experience replay. So, looking at an object the robot will try to sanitize it. Also a small LIDAR system that would help it navigate across the room.

The bot is a simple cart with 4 UVC lights mounted on top of it. In our experimentation phase we are trying out the mechanism with a 4W Philips light to get the proof of concept ready.

So, this is the basic outline of our project.