People adapt a consistent center-of-mass trajectory in a novel force field

Abstract

During human walking the whole body center-of-mass (COM) trajectory may be a control objective, a goal the central nervous system uses to plan and regulate movement. Our previous observation, that after practice walking in a novel laterally directed force field people adapt a COM trajectory similar to their normal trajectory, supports this idea. However, our prior work only presented data demonstrating changes in COM trajectory in response to a single force field. To evaluate whether this phenomena is robust, in the present study we present new data demonstrating that people adapt their COM trajectory in a similar manner when the direction of the external force field is changed resulting in drastically different lower limb joint dynamics. Specifically, we applied a continuous, left-directed force field (in the previous experiment the force field was applied to the right) to the COM as participants performed repeated trials of a discrete walking task. We again hypothesized that with practice walking in the force field people would adapt a COM trajectory that was similar to their baseline performance and exhibit aftereffects, deviation of their COM trajectory in the opposite direction of force field, when the field was unexpectedly removed. These hypotheses were supported and suggest that participants formed an internal model to control their COM trajectory. Collectively these findings demonstrate that people adapt their gait patterns to anticipate consistent aspects of the external environment. These findings suggest that this response is robust to force fields applied in multiple directions that may require substantially different neural control.

NEW & NOTEWORTHY With experience people adapted a predictive internal model to control their whole body center-of-mass walking trajectory that anticipated the disruptive laterally directed forces of a novel and consistent external environment. Collectively these findings demonstrate that adaptation of gait to anticipate consistent aspects of the external environment is a response that is robust to force fields in multiple directions that require substantially different lower limb dynamics and neural control.

INTRODUCTION

How does the nervous system plan walking trajectories to move from one location to another? Although there are an infinite number of potential solutions, research examining the expressed behavioral choices during walking provides insight (1–3). During walking, the trajectory of the whole body center of mass (COM) is smooth and stereotypical while other features of walking, such as joint kinematics and foot placements, are variable (1–3). In addition, in response to perturbations occurring while walking, several studies have found a preservation of COM dynamics (4–8). Collectively, this research suggests that during walking the COM trajectory may be a control objective, a goal that the central nervous system uses to plan and regulate movement (9). However, how the nervous system adapts the COM trajectory to interact with changing dynamics is not well understood (10).

To investigate how COM trajectory is controlled when walking in a changing external environment, we recently investigated how people adapt to a novel and consistent laterally directed force field during repeated trials of a discrete goal-directed walking task (11). Initially, the force field resulted in large lateral deviations of the COM trajectory in the same direction as the applied force. With experience, participants resumed a COM trajectory resembling walking in an unperturbed environment. This finding suggests that COM trajectory was a control objective of the task. Then when the force field was unexpectedly removed, participants exhibited large lateral COM deviations in the opposite direction of the anticipated force field. The aftereffects observed suggest that participants formed a predictive internal model, a neural representation of the relationship between a motor command and the resulting movement (12–14) that accounted for the anticipated external environment, to control their COM trajectory.

Although these results are valuable, because we only examined changes in COM trajectory in response to a single force field it is difficult to conclude that our findings are robust. Observing how COM trajectory is controlled in a variety of contexts would provide more robust evidence about how the nervous system is controlling COM trajectory during walking. Thus, here we present new data that have not been previously published but were collected at the same time as the data that were presented previously (11). Participants repeated the prior experimental protocol with a single important change: the direction of the laterally directed force field was changed from directed toward the participants’ right to toward their left. This modification will drastically change joint dynamics, and threat to maintaining upright walking, because participants will be accelerated in a direction where there is no support limb readily available to offset the abnormal lateral deviation created by the force field. If with practice walking in this novel force field participants again adapt a COM trajectory that resembles walking in a natural (unperturbed) environment and show aftereffects when the force field is unexpectedly removed, this would require substantially different neuromuscular control strategies than used in the prior experiment. Such a finding would strengthen the conclusions of the prior study, namely that people use a predictive internal model to control the COM trajectory when walking in a changing external environment.

Therefore, the purpose of the present study was to investigate how the nervous system controls the COM trajectory during a goal-directed walking task performed in a novel and consistent environment that would require neuromuscular control strategies substantially different from our prior study. We created this environment by applying a continuous, laterally directed force field to the COM that was proportional in magnitude to forward walking velocity and directed toward the participant’s left [in the prior study (11), the force field was directed to the participant’s right]. We hypothesized that, similar to previous studies of predictable changes in environment (10, 11), people would form an internal model of the COM trajectory, as shown by adaptation of COM signed deviation toward baseline in the force field, and aftereffects, deviation in the opposite direction of initial COM signed deviation when the force field is unexpectedly removed.

Additionally, we investigated two potential strategies contributing to COM trajectory control, COM lateral offset to measure anticipatory postural adjustments and foot placement (step width) to measure foot placement. COM lateral offset is a measure of the anticipatory postural adjustment made during gait initiation. By changing the lateral position of their COM before taking a step forward people can bias their posture in anticipation of the external environment. In our prior study, we found that with experience the COM lateral offset changed in the opposite direction of applied field, indicating that this mechanism contributed to predictive control of the COM (11). Thus, we hypothesize that in the present study when people experience a consistent force field directed toward their left, they will bias their COM lateral offset toward their right to offset some of the effects of this force field. Additionally, in this task changing the lateral foot placement location of the first step is a method to change the lateral acceleration of the COM (15) after the force field is applied. In our prior study, we found that when the force field was first applied people immediately changed their lateral foot placement to be more biased in the same direction as the force field, suggesting that this mechanism contributed to the reactive control of the COM (11). Thus, we hypothesized that in the present study when people experience a consistent force field directed toward their left, they will immediately change the lateral foot placement location of their first step to be more biased to their left to offset some of the effect of the force field.

METHODS

The methods presented here are near identical to those published in a previous article (11). The data presented here and those presented earlier were collected from each participant on the same day. Only half of the data have been published (11). The other half are presented in this study. Raw data from this study are available at https://doi.org/10.21985/n2-pnrm-kj07.

Participants

Thirteen healthy young adults (7 females; 22.8 ± 2.1 yr and 65.9 ± 8.5 kg, mean ± SD) participated in this study. The Northwestern University Institutional Review Board approved the protocol, and all participants provided written informed consent. Participants were able to walk for 30 min without fatigue or other health risks. They also were free of any musculoskeletal and/or vestibular pathologies affecting gait or balance.

Experimental Setup

Participants performed a series of discrete goal-directed walking trials, walking from a start target to an end target (Fig. 1A). The targets (0.3 × 0.3 m square) were projected (Hitachi America, Ltd) on the floor. The distance between targets was adjusted for each participant to be 1.5× leg length, approximately a two-step task. For safety, participants wore a trunk harness attached to a passive overhead safety device (Aretech, Ashburn, VA) that did not restrict fore-aft motion. The harness was adjusted so it did not provide body weight support or restrict lateral movement.

Figure 1.

Methods: experimental setup (A), applied force field (B), and protocol (C). A: schematic top view of a participant shown performing the goal-directed walking task, walking from the start to the end target. The cable robot consisted of a pair of actuated cables (dashed black lines), routed through a trolley system (black circles) and attached bilaterally (gray circles) on the medial aspect of a snug pelvic harness worn by the participant. Force on each cable was controlled by a series-elastic linear motor (gray boxes). F_y (dark gray vector) represents the lateral force applied to the center of mass (COM), and V_x (light gray vector) represents forward walking velocity. B: representation of the force applied to the participant’s COM during a force field trial. Force applied (dark gray trace) is proportional in magnitude to forward walking velocity (light gray trace) with a gain of 80 N/(m/s). C: participants performed 110 consecutive trials, each represented as a rectangular block ordered left to right. The force field was applied (black squares) during Force Field trials (except for Catch trials) and absent (white squares) during Baseline and Washout trials. Data were analyzed during 5 distinct experimental 4-trial periods highlighted in their respective colors. Reprinted from Ref. 11.

A cable-driven robot applied a lateral force field to the COM during walking to create a novel walking environment (Fig. 1A) (16). The force field was proportional in magnitude to forward walking velocity and was directed toward the participant’s right or left side (Fig. 1B). The force field was proportional in magnitude to the forward walking velocity because we wanted to create a novel walking task in which the application of the applied external forces was dependent on the participant’s own motions. Creating applied forces that are under the control of the person’s own motions should facilitate learning (17). Forward walking velocity was calculated in real time with the derivative of position measured by a string potentiometer. Position data were sampled at 100 Hz and low-pass filtered to reject spikes in the derivative (velocity) due to noise. Participants could not detect whether the field would be applied until they began each trial and their forward velocity exceeded 0.2 m/s. The applied force field gain [80 N/(m/s)] was consistent across participants and selected to provide a challenging walking environment but not strong enough to evoke a fall or injury (Supplemental Video; available at https://doi.org/10.21985/n2-0zrq-rq72).

To ensure similar forward walking velocities across all trials, participants received feedback at the end of each trial. After being given feedback, they were instructed to modify their next trial accordingly. A monitor positioned at the end of the walking path provided visual feedback stating either “too slow,” “too fast,” or “success,” depending on how the peak forward velocity compared to a desired value of 1.2 ± 0.1 m/s.

Before participants began the experiment, they practiced the goal-directed walking task without any velocity feedback and force fields for two trials. Next, we had participants practice the task until they successfully reached the speed target so that they would know what the task felt like. Finally, they practiced two trials with no velocity feedback information but with the force field applied to the right, followed by two trials with the force field applied to the left.

We placed 13 active markers on the pelvis (3 markers) and bilaterally on the greater trochanter, lateral malleolus, calcaneus, and 2nd and 5th metatarsals to measure kinematics. A 12-camera motion capture system (Qualisys, Gothenburg, Sweden) recorded three-dimensional (3-D) marker coordinates at 200 Hz.

Protocol

Participants performed two blocks of 110 consecutive goal-directed walking trials (Fig. 1C). Post hoc testing revealed no effect of order (11). Each block consisted of 20 Baseline trials (no applied forces), followed by 70 Force Field trials, and concluding with 20 Washout trials (no applied forces). Additionally, three Catch trials (no applied forces) were interspersed within the Force Field trials occurring during trials 45, 60, and 75. Catch trials were used to evaluate control strategies. All participants performed one block during which all the Force Field trials were directed toward their right and one block during which all the Force Field trials were directed toward their left. Block order (right- or left-directed Force Field trials) was randomized across participants. Post hoc testing revealed no effect of order (11). Participants were not aware of the trial order (Baseline, Force Field, or Washout) or the total number of trials they would perform; they only knew that perturbations may or not may occur each trial. After completing the first block of 110 trials participants rested and then repeated the sequence. Results of the block of trials when the force field was directed to the right have been previously analyzed and published (11). Here we analyze the block of trials when the force field was directed to the left. This data set has not been previously published.

Each walking trial began with the participant standing with both feet located in the start target (Supplemental Video). Before moving, participants heard an auditory “3-2-1” countdown followed by a “GO” cue. Once they heard the cue, participants walked to the end target. Participants were instructed to always step with their right foot first. The trial concluded once both feet were located within the end target. An audible “beep” signaled that the trial was over and that the participant should return to the start target (no force fields were applied during this time). To minimize upper limb movements participants crossed their arms, but besides these provisions participants were able to complete the task in the manner they felt most comfortable.

Data Processing and Calculations

Kinematic marker data were processed with Visual3D (C-Motion, Germantown, MD) and a custom MATLAB (MathWorks, Natick, MA) program. Marker data were gap-filled and low-pass filtered (Butterworth, 6-Hz cutoff frequency). Gait events, time of initial foot contact and toe-off, were identified by the inferior-superior positions of the calcaneus and 2nd metatarsal markers for each step. Initial contact was identified as the local minimum of the calcaneal marker per step and toe-off as the local minimum of the 2nd metatarsal per step. To ensure correct marking, all steps were visually inspected to verify accurate event detection. COM position was calculated in Visual3D as the center of the pelvis model, determined by three pelvic and two greater trochanter markers.

We analyzed kinematic data of COM trajectory between the start and end targets to characterize control of COM trajectory. To do this, we calculated COM signed deviation, the signed area of COM trajectory relative to a straight-line path originating from the lateral COM position at first toe-off (Fig. 2). By taking the difference between areas on either side of a straight-line path, COM signed deviation reflected directional biases in COM trajectory.

Figure 2.

Data analysis: center of mass (COM; A) and foot placement (B). A: COM lateral offset, shown in purple, was defined as the lateral distance between the COM location before the “GO” cue (1) and COM location at first toe-off (2). This value represents lateral excursion of the COM prior to forward movement. The gray trajectory represents the COM motion during the trial. COM signed deviation was defined as the blue shaded area between the gray trajectory and a straight-line path originating from 2 and ending at 3. Deviations to the right of the straight-line path were given negative values, and deviations to the left were positive. B: step width, shown in light blue, was defined as the lateral distance between steps. For example, step width for step 1 was defined as the lateral distance between step 0 and step 1.

To gain insight into the strategies that participants used to create the observed COM trajectory, we evaluated anticipatory postural adjustments (APAs) (18, 19). We did this by quantifying the lateral movement of the COM before forward movement. Our calculation for COM lateral offset was the lateral distance between the COM position at the “GO” cue and the COM position at first toe-off (Fig. 2). Additionally, we evaluated foot placement using step width, which we calculated as the medio-lateral distance between the left and right 5th metatarsal markers at initial foot contact. Step width was evaluated for the first two steps of each trial.

Statistical Analysis

We hypothesized that, similar to previous studies of predictable changes in environment (10, 11), people would form an internal model of the COM trajectory, as shown by adaptation of COM signed deviation toward baseline in the force field, and see aftereffects, deviation in the opposite direction of initial COM signed deviation, when the force field is unexpectedly removed. We tested this hypothesis by evaluating whether the magnitude of the COM signed deviation changed from the initial to final trials in the force field. Additionally, we examined COM trajectories during Catch trials when the force field was unexpectedly removed. Finally, to gain insight into the strategies that participants used to create the observed COM trajectory, we examined whether anticipatory postural adjustments (COM lateral offset) and foot placement (step width) were adapted in the force field.

We averaged four trials at five different experimental periods defined within the 110 consecutive stepping trials (Fig. 1C). Experimental periods were defined as follows: Baseline performance was estimated as an average of trials 17–20 (last 4 trials of Baseline), Early Field was an average of trials 21–24 (first 4 trials in the Force Field), Late Field was an average of trials 87–90 (last 4 trials in the Force Field), Catch was an average of trials 45, 60, 75, and 91 (3 Catch trials and first trial of Washout), and finally Washout was an average of trials 107–110 (last 4 trials of Washout). We calculated these average performance estimates for all five experimental periods (Baseline, Early Field, Late Field, Catch, and Washout) for each metric (COM signed deviation, COM lateral offset, and step width) and for each participant.

Our statistical analyses were conducted with SPSS (IBM, Armonk, NY). Our analyses included a one-way repeated-measures ANOVAs with a within-subject factor of experimental period (Baseline, Early Field, Late Field, Catch, and Washout) to evaluate all metrics (COM signed deviation, COM lateral offset, and step width). When we found a significant main effect, Bonferroni-corrected pairwise comparisons were made between experimental periods, which included six pairwise comparisons (Baseline-Early Field, Baseline-Late Field, Baseline-Catch, Baseline-Washout, Early Field-Late Field, and Late Field-Catch) to evaluate adaptation and control strategies. We set significance at the P < 0.05 level for the ANOVAs and pairwise comparisons.

RESULTS

Center-of-Mass Signed Deviation

During the initial walking trials in the force field, participants demonstrated a large lateral deviation of COM trajectory in the direction of the applied force field (Fig. 3A; Supplemental Video). Positive COM signed deviation indicates that COM trajectory was biased toward the left. As participants continued to perform the walking trials in the force field, COM signed deviation returned to values that were not different from Baseline (Fig. 3, A and B). During Catch trials, participants displayed large lateral COM deviations in the opposite direction of the applied force field (Fig. 3, A and B). Finally, when the force field was removed during washout, participants’ COM signed deviation was initially offset in the opposite direction of the applied force field and trended toward Baseline values with practice (Fig. 3, A and B). Please note that Fig. 3B is averaged across participants for each trial, for visualization of the trials over time; this is different from the averaging used for our statistical results. COM signed deviation changed from −0.033 ± 0.01 m² (mean ± SD) at Baseline to 0.028 ± 0.03 m² at Early Field (more biased to the left) (Fig. 3C). Outcomes from the one-way repeated-measures ANOVA identified a significant main effect of experimental period on COM signed deviation (P < 0.001), indicating that differences exist between comparisons of Baseline, Early Field, Late Field, Catch, and Washout. Pairwise comparisons identified significant differences in COM signed deviation between Baseline and Early Field (P < 0.001), indicating that Early Field was directed more to the left than Baseline (Fig. 3C). Pairwise comparisons also found no significant differences between Baseline and Late Field (P = 0.147), with a Late Field value of −0.027 ± 0.03 m² (Fig. 3C). Furthermore, there was a significant difference between Early Field and Late Field (P < 0.001), indicating that Early Field was directed more to the left than Late Field (Fig. 3C). During Catch trials, participants displayed large lateral COM deviations in the opposite direction of the applied force field: −0.091 ± 0.04 m² (Fig. 3C). Pairwise comparisons identified significant differences in COM signed deviation between Baseline and Catch (P < 0.001) and between Late Field and Catch (P < 0.001), indicating that Catch was directed more toward the right compared with both Baseline and Late Field (Fig. 3C). However, Washout, −0.042 ± 0.01 m², was significantly different from Baseline (P < 0.001), indicating that Washout was more negative, directed more to the right, compared with Baseline (Fig. 3C). It is important to note that COM signed deviation was negative on average because participants started the task by stepping with their right foot first. As a result of the COM shifting over the right foot initially during the trial, the data are biased toward the right, which is indicated by a negative COM signed deviation value. Additionally, these data were collected as a first study using this protocol, and therefore we did not have preliminary data and did not perform a power analysis. However, we did evaluate effect size by measuring partial eta squared and found the value to be 0.862. We conclude from this finding that there were large differences between experimental periods.

Figure 3.

Center-of-mass (COM) signed deviation: representative participant (A), group averages (B), and statistical results (C). A: COM trajectory during select trials from analysis periods for a single representative participant. B: group averages of COM signed deviation are shown across experimental trial (black line with gray SE shading). The background distinguishes trials used for data analysis periods, shaded in their respective colors. C: mean $\pm \mathrm{\hspace{0.17em}}$SD for COM signed deviation across experimental period. x-Axis labels: B, EF, C, LF, and W represent periods listed in order in A. *Significance (P < 0.05) from Baseline; +significance (P < 0.05) from Late Field.

Center-of-Mass Lateral Offset

During Early Field there was no change in COM lateral offset from Baseline (Fig. 4A). As participants continued to perform walking trials in the field, COM lateral offset increased in the opposite direction of the applied force field (Fig. 4A). When force field was removed during Catch trials, COM lateral offset remained increased in the opposite direction of the applied force field (Fig. 4A). Outcomes from the one-way repeated-measures ANOVA identified a main effect of experimental period on COM lateral offset (P < 0.001), indicating differences between Baseline, Early Field, Late Field, Catch, and Washout. Negative COM lateral offset values indicate movement to the right, relative to COM location at the beginning of the trial. COM lateral offset was not different: 0.014 ± 0.01 m at Baseline and 0.010 ± 0.006 m at Early Field (Fig. 4B). Pairwise comparisons found that COM lateral offset was not significantly different between Baseline and Early Field (P = 0.238) (Fig. 4B; Supplemental Video). COM lateral offset changed to −0.002 ± 0.01 m at Late Field (Fig. 4B). This change in COM lateral offset was reflected in pairwise comparisons revealing a significant difference between Baseline and Late Field (P < 0.001), indicating that Late Field was more negative, directed more to the right, than Baseline (Fig. 4B). Additionally, there was a significant difference between Early Field and Late Field (P = 0.008), indicating that Late Field was more negative, directed more to the right, than Baseline (Fig. 4B). No significant difference was found between Late Field and Catch (P = 0.055) (Fig. 4B). Additionally, a significant difference was found between Baseline and Catch (P < 0.001), with the Catch values deviating in the same direction as the Late Field trials, relative to Baseline (Fig. 4B). Finally, no significant difference was found between Baseline and Washout (P = 0.60) (Fig. 4B).

Figure 4.

Center-of-mass (COM) lateral offset: group averages (A) and statistical results (B). A: group averages of COM lateral offset are shown across experimental trials (black line with gray SE shading). The background distinguishes trials used for data analysis periods, shaded in their respective colors. COM lateral offset did not immediately change when the field was first turned on but adapted to an increased value with practice in the field. B: mean $\pm \mathrm{\hspace{0.17em}}$SD for COM lateral offset across experimental period. x-Axis labels: B, EF, C, LF, and W represent periods listed in order in A. *Significance (P < 0.05) from Baseline; +significance (P < 0.05) from Late Field.

Step Width

During step 1, participants immediately modified foot placement in the direction of the applied force field during Early Field, continued this strategy throughout the Force Field trials, and displayed aftereffects, a change in step width in the opposite direction as initial error (Fig. 5A). During step 2, we observed a lot of variability (Fig. 5B). Statistical analysis revealed a significant main effect of experimental condition for step width for both steps 1 (P < 0.001) and 2 (P = 0.031), indicating differences between Baseline, Early Field, Late Field, Catch, and Washout. During Early Field trials participants took a narrow step (a step toward the left, in the direction of the applied force field) for step 1 and a step similar in magnitude to Baseline for step 2 (Fig. 5B; Supplemental Video). Step width decreased from 0.133 ± 0.03 m at Baseline to 0.047 ± 0.05 m at Early Field for step 1 and maintained a value from 0.125 ± 0.03 m to 0.171 ± 0.08 m for step 2 (Fig. 5B). These adjustments were supported by pairwise comparisons revealing a significant decrease in step width between Baseline and Early Field for step 1 (P < 0.001) and no significant change for step 2 (P = 0.548) (Fig. 5B). Across trials in the field, participants continued to take narrow steps for step 1 and maintain Baseline values for step 2 (Fig. 5B). This is supported by a significant difference between Baseline and Late Field for step 1 (P = 0.001), indicating that step width was more negative, narrower, than Baseline and no significant difference for step 2 (P = 0.600), as well as no significant difference between Early Field and Late Field for step 1 (P = 0.600) and step 2 (P = 0.097) (Fig. 5B). When the force field was unexpectedly removed during Catch trials, step width values increased in the opposite direction as Early Field trials for step 1 and remained at Baseline values for step 2 (Fig. 5B). There was a significant difference between Baseline and Catch for step 1 (P = 0.002), indicating that Catch was more positive, wider, than Baseline, but no significant difference for step 2 (P = 0.128) (Fig. 5B). Additionally, there was a significant difference between Late Field and Catch for step 1 (P < 0.001), indicating that Catch was more positive, wider, than Baseline, and no significant difference for step 2 (P = 0.600) (Fig. 5). There was no significant difference between Baseline and Washout trials for step 1 (P = 0.6) or step 2 (P = 0.6) (Fig. 5B).

Figure 5.

Step width: group averages (A) and statistical results (B). A: group-averaged step width across experimental trials (black line with gray SE shading) shown for step 1 (left) and step 2 (right). The background distinguishes trials used for data analysis periods, shaded in their respective colors. Step width for step 1 decreased within the first trial in the field and did not adapt back to Baseline after practicing in the field. Step width for step 2 remained the same throughout all experimental periods. B: mean $\pm \mathrm{\hspace{0.17em}}$SD for step width across experimental period. x-Axis labels: B, EF, C, LF, and W represent periods listed in order in A. *Significance (P < 0.05) from Baseline.

DISCUSSION

This study investigated whether the nervous system controls COM trajectory during a goal-directed walking task in a novel and consistent environment. In support of our hypothesis, our results suggest that COM trajectory was controlled by forming a predictive internal model. Evidence that participants formed an internal model to specifically offset the predicted external force field includes the following: 1) the magnitude of the COM signed deviation during Late Field trials adapted to be similar to Baseline values, and 2) participants exhibited aftereffects during Catch trials when the force field was unexpectedly removed (significant COM signed deviations during Catch trials in the direction opposite of the anticipated force field).

Center-of-Mass Signed Deviation

When the force field was first unexpectedly applied (Early Field) all participants demonstrated a change in COM signed deviation in the same direction as the field, indicating that the applied forces were strong enough to alter normal COM movement (Fig. 3, A and B). With practice, participants adapted their COM trajectory, resulting in no difference in COM signed deviation between Baseline and Late Field. During Catch trials, there were substantial COM signed deviations in the opposite direction of the force field (Fig. 3, A and B). This aftereffect supports our hypothesis that participants would form an internal model that accommodates the external environment.

Our findings align with previous studies that have observed stereotypical COM trajectories during goal-directed walking (1–3, 11) and the preservation of COM dynamics in response to perturbations (4–8). Furthermore, our findings also suggest that the nervous system formed a predictive internal model that was used to control the COM trajectory during walking. These results are consistent with our previous study in which the force field was applied in the opposite direction (11). We found consistent results despite there being differences in lower limb dynamics, thus suggesting that these results are robust.

Center-of-Mass Lateral Offset

COM lateral offset, a measure of an anticipatory postural adjustments, is thought to assist with balance control during step initiation (20). COM lateral offset may also aid in control of COM trajectory during short forward walking movements by biasing the participant’s posture in anticipation of the external environment. In our prior study (11) we found that with practice participants increased their COM lateral offset in the opposite direction of the force field. We found similar results in the present study: the force field was directed toward the left and participants adapted by increasing their COM lateral offset to the right (Fig. 4). Specifically, as can be seen in Fig. 4A, the very first trial in the force field appears similar in magnitude to Baseline because the predictive feedforward control anticipates that the force field (or lack thereof) will be consistent with that experienced during the previous 20 trials. Then, in the second trial in the force field, we see a shift of the COM trajectory toward the right because the participant now has experience to suggest that the force field may be directed toward the left. Our statistical results suggest that this anticipatory COM lateral offset was an important strategy in controlling the COM trajectory during forward walking. This result provides insight that a predictive feedforward control was used to control the COM trajectory because the COM lateral offset occurs before receiving sensory information about the force field characteristics that will be present during the upcoming walking trial.

Step Width

To further understand the strategies underlying the resulting COM trajectory, we evaluated foot placement by measuring step width. We found that participants immediately took a step toward the direction of the force field when it was initially applied (Fig. 5). When the force field was unexpectedly removed, foot placement deviated more toward the right, opposite of the applied field (Fig. 5). This immediate change in foot placement suggests that step width may have been a result of passive dynamics and/or feedback control of the swing limb (21, 22) rather than a feedforward strategy. Interestingly, these results are similar to our previous experiment, which found immediate changes in foot placement when the force field was turned on and a return to Baseline when the force field was turned off. In both studies, foot placement may be assisting in immediate modulation of COM trajectory rather than adapting a predictive control strategy, as seen for COM signed deviation and COM lateral offset.

We found no difference in step width between any experimental periods in step 2 (Fig. 5). Because there were changes in the first step, we anticipated that there would also need to be adjustments on step 2 to adequately reach the target. However, the variability in step 2 was very large. A post hoc analysis found that participants were using two separate strategies to accomplish the goal-directed walking task. Six of the participants used two steps to reach the target, whereas seven of the participants used three steps. We were not able to find any other reasonable explanation within the collected demographics or descriptive factors between the groups (including sex, exercise regimen, injury history, and limb dominance) that could explain the groups. It is important to note that although there were differences in these strategies this did not impact the major findings we observed for COM signed deviation and COM lateral offset. This is confirmed by the significance of our statistical results as well as visualization of each COM trajectory pattern for participants in each group.

Conclusions

We found that when people practiced a goal-directed walking task in a novel and consistent external force field that initially disrupted their normal walking trajectory, they adapted a predictive internal model to control their COM motion. This control mechanism anticipated the external environment and enabled people to create a COM walking trajectory in the force field that resembled their trajectory during baseline walking conditions. This outcome supports the conclusions of an earlier experiment (11) that observed this phenomenon in response to an external environment applied in the opposite direction of that used in the present study. Collectively these findings demonstrate that adaptation of gait to anticipate consistent aspects of the external environment is a response that is robust to force fields in multiple directions that require substantially different lower limb dynamics and neural control.

DATA AVAILABILITY

Raw data from this study are available at https://doi.org/10.21985/n2-pnrm-kj07.

SUPPLEMENTAL MATERIAL

Supplemental Video: https://doi.org/10.21985/n2-0zrq-rq72.

GRANTS

This project was supported in part by a training grant from the NIH (T32EB009406), NIH Grant 5T32AR073157, Terminal Year Fellowship from Northwestern’s McCormick School of Engineering and Applied Science, and Merit Review Award I01RX001979 from the United States Department of Veterans Affairs, Rehabilitation Research and Development Service.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.A.B., G.B., and K.E.G. conceived and designed research; M.A.B., G.B., and K.E.G. performed experiments; M.A.B., G.B., and K.E.G. analyzed data; M.A.B., G.B., and K.E.G. interpreted results of experiments; M.A.B. prepared figures; M.A.B. drafted manuscript; M.A.B., G.B., and K.E.G. edited and revised manuscript; M.A.B., G.B., and K.E.G. approved final version of manuscript.