Studies involving human mobility often isolate a particular task of interest, such as level walking or inclined walking, and attempt to capture the steady-state biomechanics driving that task. In contrast, general human ambulation involves continuous adjustment to the varying terrain one may encounter while moving from one place to the next. In order to better understand human movement, it is necessary to investigate the mechanics that enable successful transitions between these terrains and account for the severity of a given terrain.
The tasks of ascending and descending the steps of a staircase show the importance of adapting to varied terrain. Falls occurring on stairs present a significant public health risk to people of all ages [1]–[3], are responsible for over 1,000,000 visits to United States emergency departments annually [4], and represent a major cause of traumatic brain injury in adults [2]. Stairs can be particularly hazardous for the elderly, who must contend with age-related changes in mobility [5]–[8] and are more likely to experience potentially-fatal bone fractures following a fall [1], [3]. Furthermore, the range of physical dimensions staircases take on can be a confounder in studies about stair falls [9] and furthers the difficulty of assessing fall hazards in a standardized manner [10]. With many falls occur during the transitions between level ground and stair walking [11], the mechanics employed during these transitionary phases are relevant to understanding these incidents.
Further motivation can be found in the area of powered mobility-assistive devices, where considerable research has focused on of how best to handle the transitions between level ground and stairs. Strategies range from the simple methods of user-operated switching [12], to volitional switching with EMG [13]–[15], anticipatory strategies based on input from biomechanical sensors [16], [17], and more recently computer vision [18], [19]. To successfully emulate the fluency of able-bodied ambulation, an assistive device would ideally facilitate not only the timely transition between level walking and stair walking, but also the mechanics of the transition between the two terrains [20]. This was demonstrated in a recent study [21] that modeled transition kinematics by interpolating between steady-state modes and incorporating an additional term to account for kinematics only seen in transitions, operating on the assumption that transition kinematics require special treatment from a modeling perspective. Thus, identifying whether transition kinematics are unique and how they deviate from those of steady-state walking will supplement and motivate the continued refinement of assistive technologies.
Prior studies about stair biomechanics have noted that individuals share some basic mechanical patterns [22] when traversing stairs, and that stair inclination angle affects kinematics and kinetics in the lower limbs [23], [24]. Others have observed that the transitions between level walking and stair walking are accompanied by anticipatory mechanics and muscle activations [25], [26]. While these studies demonstrate that there are biomechanical responses elicited by transitions and terrain severity in stair walking, there are presently no studies that statistically characterize how and when the lower limb mechanics change under these circumstances.
In this study, we highlight the differences in sagittal-plane lower-limb joint trajectories during the locomotor transitions between level walking and stair walking, as well as the effects of inclination angle on stair walking mechanics. This includes the transitions into and out of stair walking for both ascending and descending tasks, with these tasks repeated over four different inclination angles. Kinematic data collected from ten able-bodied subjects [27] was separated into level walking (LW), transition (TR), and stair walking (SW) strides. Using statistical parametric mapping (SPM), we compare full-stride joint trajectories against each other rather than specific points of interest, for a more holistic picture of how joint kinematics vary due to these contextual and environmental factors. We hypothesize that transition strides will present unique joint kinematics when compared to those of level walking and stair walking. Furthermore, we hypothesize that the angle of staircase inclination will affect kinematics in transition strides, despite the modest difference in stair height between inclinations.
Finally, we perform a Gaussian process regression (GPR) on the same dataset to predict joint angles based on continuous gait phase and stair grade as well as discrete ambulation context (e.g., ascending/descending, transition type). GPR is a “black-box” statistical method that has been used in gait-related contexts [28]–[30] for its ability to handle the highly nonlinear joint trajectories and to incorporate the variability of the underlying dataset [31]. Using cross-validation to assess the model error for each joint over each task, we hypothesize that there will be lower error in predictions of steady-state kinematics than in transition kinematics.