UC Berkeley

For millions of individuals, a spinal cord injury has taken away their ability to walk. While wheelchairs and leg braces offer mobility options, none offer a means to stand up and walk. For these individuals, secondary injuries can be prevalent, and special care must be taken to avoid the pain and cost of pressure sores, urinary tract infections, and other such ailments. Furthermore, there is an emotional benefit to being able to stand and walk. Events such as choosing your own seat at the theater or sports game, walking your daughter down the aisle at her wedding, reaching the pasta on the top shelf at the grocery store, or checking out of a hotel at the main counter, are taken for granted by those who can walk, but for those who use a wheelchair for mobility, these are stark reminders of the limitations of the chair.

Exoskeletons provide a means by which these individuals can get up again and walk. They offer power joints and a support for the body so that a user with a spinal cord injury can rely on the robot's power to replace what their body no longer provides. While the architecture and design of such an exoskeleton is complex, the control of the exoskeleton offers numerous challenges.

This thesis presents the development and testing of a method to allow the user to communicate his desired motion to the robot. For an exoskeleton to truly provide freedom for the user, the user must be able to operate the exoskeleton independently. To do this, the exoskeleton must know what the user wants to do and when and then decide if that maneuver is safe. The user communicates his desired action to the exoskeleton using the Human Machine Interface (HMI).

This thesis describes development of the hardware and software for the HMI beginning with the conception of the structure of the HMI based on end-user surveys and observations of users. The hardware was then developed to determine the state transitions and the software was written to determine desired state changes. The Human Machine Interface was then verified using a mockup to test and then was tested on the eLEGS exoskeleton. The software was verified through experiments and theoretically using classifiers. The Human Machine Interface was tested by subjects with a wide range of injuries and abilities to ensure that it performed safely for all users. Based on experience with the Human Machine Interface, improvements in robustness and usability were made.

This thesis also presents the development of some of the continuous controllers used to achieve the sitting and standing motions. While traditional control strategies rely on models, control of exoskeletons includes a human in the loop, which can be a sizeable disturbance. Therefore, the controller development must be robust to this disturbance and also take into account the comfort and safety of the user.

The results presented here show numerous spinal cord injury patients of varying levels and completeness able to ambulate independently using the HMI developed for eLEGS. They are able to walk, sit, and stand naturally, thus providing wheelchair users a viable means of walking again.