Integration of kinematic and muscular null spaces for motor augmentation

Human Motor Augmentation (HMA) is an engineering field aimed at enhancing human capabilities beyond the able-bodied spectrum via technology, focusing on manipulation and locomotion. This is theoretically feasible due to task null spaces — variable spaces permitting motor variations without impacting concurrent and related physiological functions. These spaces span neural, kinematic, and muscular variables. This thesis explores if two distinct task null spaces — a kinematic one defined by diaphragmatic modulation, and a muscular one defined by electromyographic (EMG) signals from auricular muscles (AMs) — can be integrated to create a two degrees of freedom (2DoFs) human-machine interface (HMI) preserving essential null space properties. The main objective was to assess participants’ ability to independently control an extra robotic arm (XRA) over several training sessions. Diaphragmatic modulation controlled the robot's translational DoF, while AM EMG activity managed the rotatory DoF. The experimental protocol consisted of five sessions with ten participants using a planar XRA prototype for reaching tasks. Each session included three phases involving the robot control over: translational DoF, rotational DoF and both. Within these phases, single and dual-tasks were implemented to assess if users could maintain XRA control while performing secondary tasks related to the control strategies used, such as counting or facial expressions (FEs). Results showed progressive improvement in control accuracy and efficiency across sessions. In 1DoF tasks, training enhanced success rates and reduced execution times in both single and dual-task conditions. A significant difference in success rate emerged when comparing simple rotatory control to its dual-task counterpart, highlighting a higher difficulty in maintaining FEs during AM control. No differences appeared in translational tasks, showing preserved speech ability during diaphragm modulation. In the 2DoFs simple control task, success rate rose from a median of 81.2% in session one to 97.9% in session five, with execution time reduced by approximately 23%, from 7.3 to 5.6 seconds. While performance in the 2DoFs dual-task conditions was lower than in single task one, both success rates and execution times improved across sessions, showing motor learning. Generally, in dual-task scenarios, users achieved high performance in both primary reaching task and secondary task concurrently, suggesting that they could learn to isolate XRA control from additional tasks, even when integrating both control strategies. Motor learning was also analyzed in terms of reaching trajectory optimization, finding that the paths became more efficient, direct, and stable over time, especially in the 2DoFs tasks, indicating effective coordination of diaphragmatic and AM inputs. Finally, the NASA Task Load Index showed a significant reduction in cognitive load across all six questionnaire dimensions, underscoring progress toward more intuitive and effortless XRA motor control. In conclusion, our findings show that, despite initial challenges in integrating kinematic and muscular null spaces, users adapt and learn efficiently over a few sessions. Physiological functions were generally preserved, enabling XRA control to operate independently while coordinating with concurrent functions. This proof of concept for combining different control strategies across distinct null spaces may inspire innovative approaches in HMA, translating it into practical HMI applications.