A novel wearable device to deliver unconstrained, unpredictable slip perturbations during gait

Task-specific training has gained popularity in rehabilitation as an effective intervention to regain [1, 2] and reinforce [3] motor skills. This methodology closely mimics the sensorimotor and environmental interactions of the target task, thereby forcing the patient or subject to repetitively execute the movements necessary to accomplish the goal of the task [4]. This paradigm is widely utilized in studies of gait stability and fall prevention by administering repeated perturbations to an individual in the form of simulated trips and slips [5‐7].

Many successful task-specific techniques and apparatuses have been reported in the literature to study slipping perturbations in particular, the cause of 22–25% of falls in the community [8, 9]. Of these, the most widely used are sudden, transient accelerations on a treadmill during otherwise steady-state walking [10‐14], one or more unlockable sliding platforms embedded in a level walkway [15‐17], and oiled areas encountered along a path [18, 19]. Important advances in our understanding of reactions to “slip-and-fall” events across age groups, as well as supporting evidence for the potency of perturbation training on stability, have stemmed from the use of these perturbation methods. For example, arm swing in the sagittal plane mitigates deviations in trunk posture following a slip, an action that older adults are less able to utilize [19]. Individuals appear to transition from a reactive to anticipatory mechanism during perturbation training, characterized by an anteriorly-shifted center of mass (CoM), altered lower limb kinematics, and reduced heel contact velocity [17, 20]. Increases in compensatory step lengths [14], improved measures of stability immediately after slip onset and recovery step touchdown, and spatial-temporal adaptations [11] have also been reported. The adjustments obtained from treadmill slip training in particular can be scaled to match the intensity of perturbation [14], and those from sliding platforms can be generalized to other tasks of varying context similarity [21‐23]. While these results have only been shown for their respective perturbation methods to date, the findings illustrate the potential plasticity of learned slip recovery strategies. After participating in a perturbation protocol, subjects have retained their acquired stabilization skills for months [22], and quickly readapt after periods over a year [23, 24].

Clearly, available methods for administering task-specific slip perturbations are capable of eliciting valuable insight on slip attributes, response movements, and stability adaptations. However, they have practical limitations associated with producing slips that are unexpected [6, 25], kinematically unconstrained [26], and that represent the variety of slip conditions encountered in daily life. Foot velocity, sliding distance, slip location, and/or slip direction are restricted with these procedures, narrowing the range of unique disturbances that can be applied. Specifically, emulated perturbations from sliding platforms and treadmills are limited entirely to the anterior-posterior direction, while actual slips possess a notable mediolateral component [26]. Slippery floor surfaces provide the most realistic disturbances when encountered, but they suffer from a high degree of predictability with repeated exposure [6, 25]. These constraints are inherent to the devices and methods used to administer the perturbations.

In order to reduce or remove these restrictions and allow for more realistic slips to be imposed, a new type of apparatus is required. Our aim was to develop such a device. Namely, this apparatus would deliver repeated slipping perturbations that were unconstrained in direction and magnitude, yet be spatially and temporally unpredictable to the wearer and controlled by an attending researcher or clinician. In addition, the slips delivered by this device could be delivered at any point during stance phase. We elected to pursue a wearable design, as this would remove the spatial predictability limitation while still allowing slips to be delivered during over-ground walking. Indeed, a previous exploration of the design space for variable-friction devices concluded that, compared to various environment-based approaches, a wearable, shoe-based option was the most promising [27]. Following development of the initial prototype, subject and non-subject tests were performed to quantify the functional attributes of the device, as well as assess and validate its ability to deliver diverse perturbations and elicit recovery reactions. Based on the results of these tests, we outline improvements to be made for future device iterations. Finally, we also highlight a number of previously inaccessible gait modalities and research topics that are unlocked with the use of a wearable slipping device.

Through the experiments described above, we assessed the functional characteristics of our novel slipping device and its utility to deliver unrestricted slipping perturbations in a laboratory setting. Our results demonstrate WASP’s ability to do this at any point during stance phase, which was a primary intention of our device. This is achieved through rapid reduction of friction underfoot when triggered, thereby perturbing the base of support (BoS) of the wearer. Administered disturbances were highly variable, and appeared to be influenced at least in direction by onset phase. In its attached state, the device had a minimal effect on the wearers’ lower limb gait kinematics and caused small yet significant changes to a number of spatiotemporal parameters.

Also through our observations, we identified a number of improvements to be addressed in future iterations of wearable slip perturbation devices. First and perhaps the most important of these is the uncontrollable variability in outsole release, which renders the precise targeting of specific phases of gait by the experimenter difficult. This stems from two particular aspects of our device. The first aspect is the inconsistent delay between reception of the trigger command and outsole release. Because this release is caused by rapid melting of the monofilament, the delay is dependent on factors not easily controlled with our current mechanism. To address this issue, we developed an alternative release mechanism that uses a cam and follower system to detach the outsole after performing the experiments outlined previously. This mechanism still receives a wireless trigger signal from the experimenter, but instead of supplying power to a strand of nichrome wire, the Bluetooth-enabled relay delivers electricity from a battery to a small DC motor. This motor is outfitted with a 1000:1 gear ratio transmission in order to produce a high degree of torque. The driveshaft of the motor connects to a drop cam, which depresses a spring-loaded pin that protrudes from the housing when in contact with the peak of the cam (see illustration in Fig. 7, Additional file 3). To attach the outsole, a loop of monofilament is threaded through the buckles of the nylon webbing straps and hooked to the protruding pin of the trigger mechanism. When activated, the motor turns the cam, which allows the spring loaded pin to rapidly retract into the housing. This retraction releases the monofilament loops and thereby the outsole from the foot [see Additional file 3]. Reattaching the outsole requires the cam and pin to be reset to their “armed” positions by turning the motor, followed by hooking the monofilament back onto the protruding pin. Because outsole release does not rely on severing a loop of filament, there is no need to replace the filament after every trial. Using the same release delay testing method performed on the nichrome mechanism, we found that the cam and follower mechanism is both faster and more consistent (0.056 ± 0.02 s, Fig. 7). Based on this result, we plan to utilize the cam and follower release component in future studies using this wearable slip device, as improved control over the instant that friction is reduced will allow more targeted, consistent perturbations to be administered. Second, because the current form of WASP relies on an experimenter to manually release the outsole through a Bluetooth-connected device, one must be very accurate and consistent to target specific gait phases or events repeatedly. Nevertheless, there will undoubtedly be errors between the intended and actual times of release. This potential for error can be remediated by developing an automatically released system. Such a system could rely on signals from footswitches or an inertial measurement unit to estimate gait phases and trigger outsole release. This would require the user to simply “arm” the device before a trial to activate at a specified gait phase during the next step, after a number of strides, or after a predetermined length of time. The uncontrollable variability created by the two aspects above should not be confused with the desired “experimenter-controlled” variability in slip onset phase, which is determined primarily by the phase of stance at which the device is activated. If the goal was to deliver all slips at the same phase, a user would simply trigger the device at the same phase for every trial (for the present device, it should be triggered roughly 500 ms before the intended phase of release). Further, while the difference in supplied friction before and after triggering was substantial, slip severity could be increased through greater reduction of traction underfoot. Increased severity would increase the likelihood of eliciting falls, therefore providing greater challenges to the wearer’s balance recovery abilities. The ability to modulate the available friction after release would also be advantageous to actively change the severity of administered slips, analogous to controlling distance or velocity with treadmills or sliding platforms [13]. This feature in a wearable device could prove valuable to a progressive perturbation training program. Finally, from a clinical perspective, our device suffers from the need for a moveable overhead harness in order to be used safely, as well as a relatively large area for the wearer to walk through. These limitations are shared by all over-ground slip perturbation methods, and they may preclude settings that lack such space or equipment from using the device. Despite these limitations, we believe the unconstrained and unpredictable nature of wearable slipping devices may complement the knowledge already gained from existing methods by facilitating the study of unique slip features, which will be described in detail to follow. Successfully addressing our device’s drawbacks through alternative release mechanisms and materials will streamline this research and allow for more targeted, consistent perturbations to be delivered.

Our wearable device demonstrated the capability to produce slips with variable onset times. This capability enables the investigation of the role biomechanical context plays in slip severity, specific reactive movements, and fall rates. For example, differences in slip attributes that follow slips initiated at distinct phases may be due to the natural course of kinematic and dynamic context over stance. In the case of ES slips, the anteriorly-directed shear forces may carry the slipping foot forward and away from the rest of the body [33]. Combined with the loss of shear ground reaction forces (GRFs), this creates an exaggerated moment arm that pitches the body backward. Recovering from this backward pitch is more challenging than respective forward rotations [34, 35]. MS and LS slip recoveries may benefit from less compromising body positions and passive dynamics. For MS disturbances specifically, the affected foot appears to slide laterally from the CoM, likely causing a momentary rotation of the body in the opposite direction. The unaffected foot at this time is in swing, and may be able to oppose the motion of the CoM by redirecting its trajectory and planting lateral to the CoM. LS slips might require even less intervention, as the generally posterior-directed slip may set the CoM in motion anteriorly toward the unperturbed foot that is already positioned ahead of the body. Each phase may offer a distinct set of initial conditions from which an individual must find a unique, successful recovery strategy. Although our preliminary results on reactive movements are not sufficient to infer such effects of slip onset phase, they indicate differences that may be detectable in a future study using WASP.

The wearable apparatus delivered slipping perturbations diverse in distance, direction, and velocity in response to a range of onset times. Likewise, diverse slips that occur when navigating the environment depend only on the available friction between the support surface and the outsole as well as the distribution of forces acting over that interface [33], allowing an extremely complex and variable space of possible slip attributes. Compared to previous work, the slips provoked by WASP were slightly shorter and generally slower. In studies where these were controlled, displacements ranged from 1 cm up to 38 cm with velocities between 10 cm/s and 900 cm/s [12‐14]. When not constrained, average displacements of 78 cm and 61 cm, as well as peak velocities of 200 cm/s and 184 cm/s, were observed in fallers and non-fallers, respectively [16]. These protocols only produced translations in a single plane [12‐14, 16], however, imposing unnatural constraints on the administered perturbations [26]. As suggested by Troy and Grabiner, greater experimental control is gained at the expense of realism of the task [26]. Indeed, context similarity seems to be an important factor to the success of task-specific perturbation training, as transfer of learned stabilization skills from sliding platform training to an over-ground unconstrained slip was significantly greater compared to the same training done on a treadmill [13]. A wearable perturbation device may offer even greater task specificity of lab-induced slips to those encountered in the environment than both methods, therefore future research should investigate the importance of task specificity by comparing improvements in stability gained from training with a wearable device to those from sliding platform and treadmill protocols.

Because slips caused by WASP are highly variable and unconstrained, the compensatory steps and foot orientations in response may have to be equally diverse in order to successfully recover. Recovery steps, seemingly by being placed in the direction a fall would otherwise occur [36], reconfigure the BoS to arrest the destabilizing angular momentum of the body through an opposing angular impulse [37] and maintain body weight support. Multiple studies have aimed to characterize reactive stepping ability [10, 38‐40] and report improvements following a perturbation training protocol [12, 14]. It may be possible to supplement these advances through the use of more ecologically valid means, as a broader range of effective recovery steps could be practiced.

A wearable device such as WASP enables the application of perturbation training to other locomotory modes and environments that have not been fully examined in the existing literature, yet describe a significant portion of the steps taken day-to-day. The vast majority of gait perturbation studies have focused on linear, forward gait across a level surface, yet walking in the community is far more complex. Turning comprises approximately 35–45% of all steps taken [41], and a study by Crenshaw and colleagues indicated that 20% of reported falls in their sample occurred while changing direction [9]. The required coefficient of friction needed to safely execute a turn is 0.45 [42], more than twice the value for straight walking [33]. Slopes present a wide range of potential conditions that demand unique biomechanical adaptations to traverse. Walking directly up or down an inclined surface generates significantly larger anterior-posterior shear GRFs [43], requiring greater friction as the magnitude of the incline increases [44]. These forces peak at different times during stance as well [43], which may lead to different “critical phases” where the risk of a slip is greatest. Crossing slopes in any other direction creates asymmetrical functional leg lengths [45], resulting in unique body positions that may influence the repertoire of effective recovery strategies available in the event of a slip. In addition, WASP allows for other mechanical consequences of the perturbation to be examined that are inhibited by the restricted nature of other techniques. One crucial example may be the freedom of foot orientation to change during a slip. Perturbed foot yaw may alter the area of the BoS and therefore the possible locations of the CoP from which to generate countering GRFs. To our knowledge, foot orientation during or after a slip has yet to be investigated.

Our study indicates that wearable apparatuses like WASP are capable of instigating variable slipping perturbations during stance phase, in turn eliciting unconstrained slip mechanics and recovery reactions. Slips encountered in nature are intrinsically complex due to the limited number of constraining factors present. By closely mimicking this complexity, future work using this device will investigate the possibility of developing a more comprehensive repertoire of reactive stabilization strategies that is effective against a broader range of slipping conditions and environments.