Colloquia

Samenvatting

Walking exoskeletons are wearable robots that are strapped to the legs of (partially) paralyzed patients. The exoskeleton has actuated motors that control joint torque or motion in the legs. The long term goal of exoskeleton research is to enable patients to walk with the same ease as an able-bodied person. However, the state-of-the-art exoskeletons still have a long way to go to achieve similar performances as able-bodied people.

The main problems that the current exoskeletons have is that they are unable to maintain balance and their versatility is limited. Balance is currently maintained by the user of the exoskeleton with the use of crutches. The versatility of the gait, i.e. the ability to perform a range of different gaits on different terrains, is restricted by the fact that only preprogrammed steps can be executed.  The preprogrammed steps are selected by the user and are the execution is triggered by either pressing a button or by upper body movement.

The goal of this thesis was to improve exoskeleton control in terms of their ability to maintain balance and versatility of generated gait. Specifically, a framework that is able to plan and execute steps was designed in simulation. The steps are planned based on the known immediate surrounding terrain using a step planning procedure. The planned steps are executed taking into account simplified walking dynamics, using the Divergent Component of Motion framework. Balance is maintained using the ankle and stepping strategy, which are basic balancing strategies for bipeds. Given a push perturbation, capture points on uneven ground are used to determine when and where to step or only actuate the ankle.  The framework is tested on five obstacles of the Cybathlon exoskeleton race and manages to generate stable gait and recover from perturbations on all of those.