Water Dispenser for Dogs: Machine Learning

I wanted to go a little creative with ideas in trying to find interesting ways to use machine learning. I am the owner of three dogs and I always feel a little icky about leaving their food and water in the open for them to approach their meals as and when they like.

It annoys me to see their water bowl out in the open, or even a little dirt going into their bowl. And hence, it gave me the idea of creating a water dispenser that gives out water when my dogs bark/ or perform a certain action near a given range around the dispenser. My dogs bark quite often, hence it is important to set a range so that the dispenser knows it is a call for water and no other reason.

I am imagining training the dog to bark or perform an action when he needs water, and then have some amount of fresh water poured into their bowl each time they are thirsty. The model could be trained to rotate a motor, that would allow the water to pour through a funnel from a small storage tank that could be refilled. The water would then be poured into the bowl for a few seconds and stop by rotating the motor back, blocking the funnel after a given set of time. I am also not sure how realistic or effective this approach could be, but it would be a fun project to think about.