During neurologic exams, clinicians perform manual assessments of various joints and rely on haptic experiential knowledge to diagnose underlying neurologic conditions. It is imperative to afford clinical learners more exposure to the haptic feeling of common abnormal behaviors. Traditionally, training is carried out on practice patients or even classmates, but outcomes are not always consistent and reliable. Alternatively, medical education training simulators could render accessible, safe, consistent, and scalable training environments.

Although simulators have been widely adopted for surgical and anatomical procedures, there are no commercial task trainers for practicing neurologic examination techniques. Task trainers that can simulate multiple behaviors at various severity levels should integrate highfidelity force control capability into human-size limb mannequins. Considering the device shares many similar technical challenges to powered prosthetics and exoskeletons, we refer to this category of device as robotic task trainers. A few research prototypes have been proposed previously but none have been adopted publicly, possibly due to cost, maintenance, portability, or mechanical complexity issues.

In this dissertation research, we used a series elastic actuator (SEA) design strategy to develop robotic task trainers that achieve a balance across cost, size, and performance. Two prototype task trainers were developed. The lower-extremity trainer mimics ankle clonus and deep tendon reflex. The upper-extremity trainer replicates upper-arm spasticity, lead-pipe rigidity, and cogwheel rigidity. We discussed their design, sensing, modeling, and control aspects. Validation tests (benchtop performance and clinical expert assessments) highlight that these devices can be viable training solutions for learners in healthcare professions.