Force Myography-Based Human Robot Interactions via Deep Domain Adaptation and Generalization

Estimating applied force using force myography (FMG) technique can be effective in human-robot interactions (HRI) using data-driven models. A model predicts well when adequate training and evaluation are observed in same session, which is sometimes time consuming and impractical. In real scenarios, a pretrained transfer learning model predicting forces quickly once fine-tuned to target distribution would be a favorable choice and hence needs to be examined. Therefore, in this study a unified supervised FMG-based deep transfer learner (SFMG-DTL) model using CNN architecture was pretrained with multiple sessions FMG source data (D-s, T-s) and evaluated in estimating forces in separate target domains (D-t, T-t) via supervised domain adaptation (SDA) and supervised domain generalization (SDG). For SDA, case (i) intra-subject evaluation (D-s not equal Dt-SDA, T-s approximate to Tt-SDA) was examined, while for SDG, case (ii) cross-subject evaluation (D-s not equal Dt-SDG, T-s not equal Tt-SDG) was examined. Fine tuning with few "target training data" calibrated the model effectively towards target adaptation. The proposed SFMG-DTL model performed better with higher estimation accuracies and lower errors (R-2 >= 88%, NRMSE <= 0.6) in both cases. These results reveal that interactive force estimations via transfer learning will improve daily HRI experiences where "target training data" is limited, or faster adaptation is required.