UCLA

Physical interaction between humans and robots is becoming more widespread in society due to major research efforts in developing intelligent robotics in numerous industries: health care, manufacturing, transportation, energy, etc. The coupled human-robot system allows the strengths of one part to overcome the limitations of the other; the high-level task planning and problem-solving abilities of the human are complemented by the accuracy, strength, reliability, and repeatability of the robot.

One technology to benefit from these advancements is the exoskeleton, which has applications primarily in robot-assisted rehabilitation and human augmentation. Admittance control enables motion of the exoskeleton by generating reference trajectories corresponding to a lightweight virtual system that responds to human-applied forces. The exoskeleton's controller then tracks the reference, giving the illusion that the exoskeleton's dynamics are those of the lightweight system. The ease at which the exoskeleton moves with the operator is known as transparency. The virtual system can emulate many desired dynamics, such as those of point masses or rigid bodies. This flexibility is foundational for more advanced applications, such as having the exoskeleton also provide assistive forces during rehabilitation or restricting motions to be within a safe workspace.

Naturally, when coupling a human to an exoskeleton, there are important safety, reliability, and performance considerations. Safety is often addressed at all levels of the system: mechanics, controller, and reference generation, and is further explored in this work. The feedback connection between the human and the exoskeleton also creates possible stability issues, which may not exist in either system independently. Delay-induced instability and a mechanism for mitigation are also explored.

Furthermore, improving transparency can conflict with wearability. However, for robot-assisted rehabilitation, such a trade-off may be necessary. Thus, achieving comparable behavior for a system that has a simpler human-robot interface is of importance and is also explored.

Finally, the flexibility of virtual dynamics can enable behavior that is not otherwise possible. For instance, creating virtually constrained motion for path-guided rehabilitation and virtual reality-based object manipulation can greatly improve robot-assisted rehabilitation. These examples showcase the significant potential of robotics for the field of rehabilitation.