Lower limb exoskeletons have been extensively developed over the last several decades for people with and without movement disorders. Although lower limb exoskeletons have been shown to provide effective assistance to improve gait and reduce metabolic cost during movements, they are often heavy, bulky and uncomfortable.  Many studies with exoskeletons are limited to indoor environments and to overground or treadmill walking at a constant speed, whereas one's activities in daily life include several types of locomotion over various terrains. In order to provide adequate control in many locomotion types and in the transitions between them, an exoskeleton requires sensors to accurately detect the user's movement capacity and intentions, which may require a great number of wearable sensors. For these reasons, feasible exoskeleton use in daily life remains a challenge. The studies in this thesis are aimed at addressing some of these limitations.

The overall objectives of this thesis are to study movement biomechanics in different locomotion modes, to develop useful methods to study the interaction between a wearable exoskeleton and its user, and to develop methods that detect a person's movement ability and intentions with minimal sensor requirements. The aims of the first two studies were to create a simulation of an exoskeleton and its user and to study how different exoskeleton parameters affect the user; specifically, to study the influence of a knee exoskeleton's different weight distributions and assistive strategies on the user's required muscular effort and on the interaction forces. The aim of the third and fourth studies was to study the biomechanics and biosignals during different locomotion modes and the transitions between them, such as walking and stair climbing, and to use these signals to detect as early as possible a person's movement intentions to transition from one mode to another. The aim of the fifth study was to accurately predict, with as few wearable sensors as possible, a person's generated knee joint moment during walking.

The methods used in this thesis include musculoskeletal modeling and simulation, experimental motion capture of able-bodied participants, physical prototyping of a knee exoskeleton, and off-line prediction algorithms based on captured motion data, using fundamental concepts from muscle synergy and from recurrent neural networks.

The main findings in the first two studies are that the influence of a knee exoskeleton's weight distribution on muscle activities was movement-dependent; the external load in various exoskeleton configurations led to an additional required effort in some movements but not in all, suggesting that an exoskeleton's physical design should be aligned with the intended user's movement goals. Further main findings were that simulations of an exoskeleton's assistive strategies and the resulting muscular efforts of the user can assist in and possibly speed up the prototyping process. 

The focus in the third and fourth studies is on movement biomechanics and biosignals in various modes of locomotion and in the transitions between them. The main findings in these studies are that the computational methods we propose based on wearable sensor signals could accurately detect a person's movement intentions to transition between locomotion modes during the step preceding the transition. This finding has important potential in the design and execution of exoskeleton control. Finally, the main findings in the fifth study are that an accurate prediction of a person's knee joint moments could be performed with as few as four electromyography sensors.  

Application of these findings can have important potential in facilitating more feasibility and compliance in exoskeleton use in realistic contexts in the future.