a. Prototype

Version 11.1 by Vladimir Rullens on 2025/11/07 01:16

Scope and purpose

Our prototype is focused on assessing whether our claims CL001: Regular physical activity improves PwD physical well-being and CL002: Dance session improves PwD mental well-being hold. As such, its functionality is focused primarily on UC02.1: Dancing Session Start, incorporating RQ01.2: Human-Robot conversation and RQ01.4: Hold dance session. While UC02.2: Dancing Session Clean-Up and UC03.1: Companion Mode are incorporated, they are only done so on a superficial level, as by this point, the experiment will end. Further details on our experiment procedure are explained further in Evaluation: Test.

Procedure

Our prototype procedure is displayed in the following Figure:

Prototype_expl.png

An overview of the prototype, showcasing different functionalities.

In terms of functionality, our prototype utilizes a visual version of the Robot dog Miro that is operated through an LLM. An actual robot was not used. During the interaction, both the user and Miro interact with each other through speech-to-text and text-to-speech translations respectively, in order to allow for voice communication. To play a song for the dance session, our LLM makes use of the Spotify API, starting/pausing/stopping the song at will.

The exact procedure is as follows, based on UC02.1: Dancing Session Start:

1. Miro introduces itself and invites the user to a Dance Session using the previously mentioned communication method.

2. Once the user has accepted, a short discussion is held on the song/genre. Namely, the LLM asks the user if they would be interested in certain genres, such as jazz, rock, pop, or classical. We note the user is not forced to pick one of the given options.

3. Based on the user's response, a song is played using the Spotify API.

4. The song will play until the user interrupts the song by talking to the LLM, allowing them to modify the state of the activity. Here, by communicating with the LLM, the user can i.e. modify the song or choose to stop the activity entirely, which would send the LLM to its Companion Mode state.

Companion Mode. If the user is not in the process of entering the 'Dancing Session' or currently in one, the LLM will be in 'Companion Mode' as a fallback. We note this state was not explored during the experiments.