Drink-O-Mender

Description

This application was developed by Hannes Ritschel, Andreas Seiderer, and me for the "Lange Nacht der Wissenschaft" ("Long Night of Science")  on May 5th, 2018. This event took place in the city hall of Augsburg and aimed to present regional research projects to the general public.

The social robot Reeti (built by the French company Robopec) acted as a nutrition advisor and informed visitors about the calories contained in the offered drinks. Using reinforcement learning, the robot was meant to discover whether pointing out the high or low amounts of calories would lead to healthier choices.

Custom-built drink holders and an electronic scale informed the system how much of which juice had been consumed. Based on the calories in the cup, a reward value was calculated for the last chosen strategy and used to adjust the probability for selecting it next time.

During the winter term of 2018/2019, I supervised a student group that presented this demonstrator in several residential homes. This example application served as a conversation starter for discussing potential application areas and requirements in this domain with inhabitants and caregivers.

Publications