REPOSITORY > RESULTS

Abstract

Research and development of exoskeletons for injury prevention and assistance with everyday
living has steadily increased in the last decade. With some products already available
on the market, a widespread adoption of exoskeletons appears to be within close reach.
However, many challenges still remain to be overcome, one of which is the development
of high level control strategies that would allow for a seamless human-robot integration.
The purpose of this thesis is to explore novel control architectures suitable for high-level
exoskeleton control and expand the current state of the art with an emphasis on the effects
of human-robot interactions on human movements.
First, we explored the effectiveness of using probabilistic models to facilitate the decision
making in a high-level controller for use with a quasi-passive spinal exoskeleton. The
controller consisted of Gaussian mixture models in combination with a state machine that
identified and classified the movement of the user in real time and provided a timely control
output for the quasi-passive spinal exoskeleton. The results of this study showed that our
approach is a useful tool for the control of quasi-passive exoskeletons. Second, we evaluated
the use of probabilistic movement primitives (ProMP) for predicting user movement
intentions during walking and arm reaching. In the walking scenario we showed that we
can accurately predict some crucial gait parameters in real-time during the stride cycle.
In combination with other systems monitoring the environment our method can be used
to predict collisions of the feet with the surrounding environment. For the arm reaching
scenario, we present an exoskeleton control method that utilizes ProMP to generate predictions
of user movements in real-time which are then combined with a velocity-field-based
controller to provide assistance to the user in a predictive way. We evaluated our approach
with a haptic robot, where participants had to perform movements towards different target
locations. We showed that we accurately predicted user movement intentions while at the
same time significantly decreasing the overall physical effort exerted by the participants to
achieve the task, without affecting their movement kinematics.
For any device assisting healthy individuals, it is crucial that it does not significantly
modify their movements, as this could have negative health implications. However, sometimes
the modifications of user kinematics with exoskeletons is desired. For example when
we want to recreate artificial environments such as microgravity. We investigated whether
local gravity simulation of altered gravitational conditions on the arm would lead to changes
in kinematic parameters comparable to a full-body experience of microgravity and hypergravity.
We developed a robotic device that applied forces at the wrist to simulate microor
hypergravity conditions for the arm while subjects performed pointing movements on
a touch screen. We analysed and compared the results of several kinematic parameters
and muscle activity using this system with data of the same subjects being fully exposed
to altered gravitational conditions onboard a parabolic flight. Our results suggest, that
simulated altered gravity can elicit similar changes in some movement characteristics for
arm reaching movements. This forms the basis for developing devices such as exoskeletons
to aid in training individuals prior to undertaking tasks in varying gravitational conditions.