Supporting dataset for Demystifying Reward Design in Reinforcement Learning for Upper Extremity Interaction: Practical Guidelines for Biomechanical Simulations in HCI

This dataset contains simulator folders compatible with the User-in-the-Box framework for simulating human-computer-interaction. It includes final training checkpoints of reinforcement learning agents trained with distinct reward functions across four interaction tasks: pointing, tracking, choice reaction, and remote control. The RL agents were trained within the User-in-the-Box simulation environment, using MuJoCo (version 3.3.0 required). These checkpoints enable researchers to reproduce experiments, analyze learned policies, and investigate the influence of reward function design on agent behavior. To evaluate a checkpoint, run the following command: python uitb/test/evaluator.py simulators/mobl_arms_index_pointing --record --logging Note: The evaluation script evaluator.py is part of the User-in-the-Box repository, which must be cloned beforehand. For detailed information on implementation and experimental procedures, please refer to the associated publication.