Waitress quadcopter explores how to serve drinks by reinforcement learning

Versatility of unmanned aerial vehicles has paved the way for drones to be employed in a broad class of tasks from aerial photography to search and rescue, or even emergency medical service such as delivering defibrillator for heart attack victims. Such multi-functionality has sparked an enormous research area which is full of eminently innovative ideas for the smart nations. In this work, a quadcopter drone is used as a waitress to deliver drinks from the bar top, right to customers' tables in a pub. As every inexperienced apprentice, it goes through an educational process in order to learn how to serve drinks to fulfill the expectations of the customers. In the proposed reinforcement learning algorithm, the agent uses the evaluative feedback from the customers as reward and adapts its actions accordingly. Simulation studies demonstrate the comparative results for different action-selection policies under fairly subjective feedback conditions for different types of customers.