MuJoCo VR environments for assessing myocontrol performance and subjective parameters

How can we reliably measure myocontrol performance and subjective parameters such as agency, ownership and trust in a prosthetic system? Previous research suggests that a high level of interaction realism is required to effectively test myocontrol paradigms and transfer skill to the real world. In addition, the use of eye tracking during these dynamic interactions may be a viable option to assess subjective parameters. To investigate this further, FAU has developed virtual reality scenarios (such as the one shown here) based on, e.g., the Pasta Box task, using MuJoCo for robotic simulation and physical interaction, and incorporating eye-tracking measurements.

(The Pasta Box task test first appeared at https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0199549.)  

Current study to investigate agency, ownership, and trust 

In the current study, this environment is used to assess agency, ownership, and trust in a virtual prosthetic system. Different levels of artificial control failures and delays are introduced into the control system during dynamic interactions and the participant’s response is measured. Measurements included eye-tracking parameters, movement and muscle activity. Stay tuned for the results, which will be published soon.  

Future work 

We believe that highly realistic virtual interactions are a step towards reliable human-in-the-loop (HITL) testing of myocontrol, including both objective and subjective assessment. Therefore, a future extension of this work towards a virtual assessment suite for testing myocontrol and shared autonomy paradigms would be a highly interesting contribution to the scientific community. Furthermore, the usage of such environments for the HITL myocontrol training and usage with other robotic systems should be investigated in future work.