Abstract - 

The long-term goal of this research is to help design rehabilitation platforms that are robust and reliable and tailored to subjects’ needs. Many chronic stroke survivors often have poor volitional control of the impaired limbs. Current rehabilitation platforms have drawbacks when assisting humans in Cartesian coordinate space, such as approaching and hitting robot singularities. This work seeks to create a controller that can optimize for dexterity while using null space control to avoid singularities. Using an artificial potential field in the null space of the robot’s Jacobian matrix, we can optimize the dexterity of a 7- degree-of-freedom robotic manipulator without ever forcing the end-effector position to move. Our simulation-based evaluation of a dexterity-optimized motion planning framework demonstrated that both offline and online planners were able to improve the Yoshikawa Dexterity Index, with the online optimizer making localized adjustments when the trajectory deviated beyond a threshold at a total of n = 19 waypoints. While the artificial potential field method enhanced local dexterity, especially when all neighbors were considered, large force magnitudes occasion- ally caused instability due to changes in the Jacobian. Although no statistically significant advantage was observed over the offline approach, the framework shows promise for increasing end- effector flexibility and supporting more natural human-in-the-loop rehabilitation tasks. This work and the results produced will be invaluable to rehabilitation robotics; with the improved dexterity achieved in this work, these devices are much more reliable for general rehabilitative use. Additionally, this work benefits the biomechanics and robotics communities by helping to understand the relationships between the range and possibility of robot-assisted human movements, bringing to light some of these movements’ limits.

Other Work