Personalized Autocomplete Teleoperation: Real-Time User Adaptation using Transfer Learning with Partial Feedback

Autocomplete enhances teleoperation by recognizing operator intentions and automating motions when instructed. This has shown to improve performance and reduce operator load. Traditionally, recognizing operator intended motions is accomplished using pre-trained deep learning (DL) models. In this paper, we introduce personalization to the Autocomplete Teleoperation Framework. By customize the Autocom-plete DL model using transfer learning with partial feedback. This approach allows the model to adapt to the specific operator in a relatively short period. Several scenarios have been evaluated to balance the speed of learning with the accuracy of the model. Results demonstrate the effectiveness of the proposed method in adapting to the specific operator.