Hartmut Geyer

People with vestibular hypofunction (PwVH) have poor walking stability, characterized by increased trunk sway, increased spatiotemporal variability, and reduced margins of stability. PwVH may have poor walking stability because of the vestibular loss itself or because they adopt cautious gait patterns to compensate for the vestibular loss (e.g., slower walking with wider and shorter steps). This study used a recently developed model of human walking stability to explore the cause of poor walking stability in PwVH. Results showed that incorporating decreased vestibular sensitivity in the model resulted in increased lateral trunk sway and increased spatiotemporal variability. Reduced margins of stability emerged when the model walked slowly. We performed additional analyses to investigate what controller changes could restore stability metrics to gain insight for potential training mechanisms. In a model with decreased vestibular sensitivity, trunk sway was restored by increasing controller gains dependent on vestibular feedback, reminiscent of vestibular adaptation. Spatiotemporal variability was reduced to normal levels in the presence of decreased vestibular sensitivity by increasing controller gains dependent on somatosensory feedback, reminiscent of sensory reweighting. In contrast, taking wider steps increased margins of stability when walking slowly. Overall, these findings demonstrate that poor walking stability in PwVH may be attributed to both decreased vestibular sensitivity as well as adapted gait mechanics. Furthermore, the model supports current clinical practice of vestibular adaptation and sensory reweighting exercises, as well as modifications to increase gait speed and step width to improve different aspects of walking stability in PwVH.

Abstract Cervical spinal cord stimulation (SCS) improves upper-limb function in individuals with chronic post-stroke hemiparesis, yet how it shapes motor control of arm movement remains unclear. During goal-directed reaching in healthy individuals, movements consist of a coordinated acceleration phase toward the target followed by a deceleration phase that stabilizes the limb near the endpoint. Disruptions in neuromotor control post-stroke can be partially restored by SCS, with prominent improvements occurring in the deceleration phase. To quantitatively characterize these effects, we used a proportional–derivative (PD) control model to fit planar reaching data from 12 healthy and 5 stroke participants. Movements were well described by the model, with proportional gain terms capturing the acceleration phase and derivative gain terms capturing velocity-dependent deceleration. In healthy individuals, model fits revealed a consistent balance between position and velocity-dependent torques that closely matched the optimal solution for smooth and stable reaching predicted by optimal feedback control simulations. In stroke, this balance was altered and partially normalized by SCS, with the most consistent changes observed in the velocity-dependent term. While the prevailing hypothesis is that SCS boosts motor drive to weak agonistic muscles, these results indicate a synergistic and potentially dominant effect in suppressing hyperexcitability of antagonistic muscles controlling the deceleration phase of movement. Finally, the PD controller model revealed frequency-dependent effects of SCS, suggesting that the model parameters may serve as biomarkers for guiding the selection of stimulation parameters. This work serves as a framework for characterizing how neuromodulatory therapies influence arm control .

Common activities of daily living naturally blend walking on level ground, stairs, and slopes. Predicting human locomotion behavior across these environments is critical for control design in lower-limb assistive devices. One approach to achieving this goal is to use feedforward simulation of conceptual gait models that capture the whole body dynamics of human locomotion. So far, however, these models have mainly been developed to predict human walking and running dynamics on level ground. OBJECTIVE: To develop a conceptual gait model that can walk on level ground as well as negotiate stairs and ramps of different inclinations (up to $\pm 42^\circ$) while producing biomimetic ground reaction forces (GRFs). METHODS: We extended the passive bipedal spring mass model by incorporating active elements, introducing seven control parameters. These parameters were then tuned to match the human GRFs for stairs. RESULTS: By optimizing these control parameters, we found that the resultant model's GRF correlates strongly with human data with median Pearson correlation coefficients $R_{x}>0.89$ and $R_{y}>0.96$ for stair walking. Furthermore, we observe its predictive capabilities are not limited to stairs, as it can also mimic the GRFs of human slope walking. Finally, we conducted a parameter analysis to assess how model parameters affect GRFs and identified a reduced, five parameter model that can reproduce human GRFs with a fidelity similar to the seven parameter model. SIGNIFICANCE: The model may help to predict human locomotion behavior in more complex terrain, for instance, when planning foot placements in the interactive control of powered lower-limb exoskeletons and prostheses.

Current control strategies for powered lower limb prostheses often lack awareness of the environment and the user's intended interactions with it. This limitation becomes particularly apparent in complex terrains. Obstacle negotiation, a critical scenario exemplifying such challenges, requires both real-time perception of obstacle geometry and responsiveness to user intention about when and where to step over or onto, to dynamically adjust swing trajectories. We propose a novel control strategy that fuses environmental awareness and human cooperativeness: an on-board depth camera detects obstacles ahead of swing phase, prompting an elevated early-swing trajectory to ensure clearance, while late-swing control defers to natural biomechanical cues from the user. This approach enables intuitive stepping strategies without requiring unnatural movement patterns. Experiments with three non-amputee participants demonstrated 100 percent success across more than 150 step-overs and 30 step-ons with randomly placed obstacles of varying heights (4-16 cm) and distances (15-70 cm). By effectively addressing obstacle navigation -- a gateway challenge for complex terrain mobility -- our system demonstrates adaptability to both environmental constraints and user intentions, with promising applications across diverse locomotion scenarios.

This work uses computational modeling to investigate the role of spinal reflex tuning during locomotor adaptation. We show, in simulation, that tuning spinal reflex gains leads to gait asymmetry adaptation, not vice versa, and that patterns of gait adaptation on a split-belt treadmill are mostly driven by tuning of spinal reflexes, and not by biomechanical disturbances triggered by belt changes. The model further hints at the cerebellum as the source of spinal reflex modulation.

Behaviors of animal bipedal locomotion can be described, in a simplified form, by the bipedal spring-mass model. The model provides predictive power, and helps us understand this complex dynamical behavior. In this paper, we analyzed a range of gaits generated by the bipedal spring-mass model during walking, and proposed a stabilizing touch-down condition for the swing leg. This policy is stabilizing against disturbances inside and outside the same energy level and requires only internal state information. In order to generalize the results to be independent of size and dimension of the system, we nondimensionalized the equations of motion for the bipedal spring-mass model. We presented the equilibrium gaits (a.k.a fixed point gaits) as a continuum on the walking state space showing how the different types of these gaits evolve and where they are located in the state space. Then, we showed the stability analysis of the proposed touch-down control policy for different energy levels and leg stiffness values. The results showed that the proposed touch-down control policy can stabilize towards all types of the symmetric equilibrium gaits. Moreover, we presented how the peak leg force changes within an energy level and as it varies due to the type of the gait since peak force is important as a measurement of injury or damage risk on a robot or animal. Finally, we presented simulations of the bipedal spring-mass model walking on level ground and rough terrain transitioning between different equilibrium gaits as the energy level of the system changes with respect to the ground height. The analysis in this paper is theoretical, and thus applicable to further our understanding of animal bipedal locomotion and the design and control of robotic systems like ATRIAS, Cassie, and Digit.

Policy decomposition is a novel framework for approximating optimal control policies of complex dynamical systems with a hierarchy of policies derived from smaller but tractable subsystems. It stands out amongst the class of hierarchical control methods by estimating <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a priori</i> how well the closed-loop behavior of different control hierarchies matches the optimal policy. However, the number of possible hierarchies grows prohibitively with the number of inputs and the dimension of the state-space of the system making it unrealistic to estimate the closed-loop performance for all hierarchies. Here, we present the development of two search methods based on Genetic Algorithm and Monte-Carlo Tree Search to tackle this combinatorial challenge, and demonstrate that it is indeed surmountable. We showcase the efficacy of our search methods and the generality of the framework by applying it towards finding hierarchies for control of three distinct robotic systems: a simplified biped, a planar manipulator, and a quadcopter. The discovered hierarchies, in comparison to heuristically designed ones, provide improved closed-loop performance or can be computed in minimal time with marginally worse control performance, and also exceed the control performance of policies obtained with popular deep reinforcement learning methods.

Abstract In a recent human study, we found that adaptive changes in step length asymmetry (SLA) are correlated with similar changes in the H-reflex gains of the leg muscles during split-belt treadmill locomotion. While this observation indicated a closer link between gait asymmetry and spinal reflex adaptation, it did not reveal their causal relationship. To better understand this relationship, here we use a neuromuscular model of human walking whose control relies primarily on spinal reflexes. Subjecting the model to split-belt treadmill locomotion with different combinations of belt speed and reflex gain adaptation patterns, we find that belt speed changes increase the variability in SLA but do not result in consistent SLA patterns as observed in human experiments, whereas reflex gain adaptations do. Furthermore, we find that the model produces SLA patterns similar to healthy adults when its reflex gains are adapted in a way similar to the H-reflex changes we observed in our previous human study. The model also predicts SLA patterns similar to the ones observed for cerebellar degeneration patients when the reflexes do not adapt beyond a sudden dip at the time the ipsilateral belt speed is lowered. Our results suggest that SLA does not arise from imposing belt speed changes but requires the adaptation of the reflex gains, and that the dynamic adaptation of these gains may be an essential part of human gait control when encountering unexpected environment changes such as the uneven speed changes in split-belt treadmill locomotion.

This work presents direct evidence for spinal reflex modulation during locomotor adaptation. In particular, we show that reflexes can be modulated on-demand unilaterally during split-belt locomotor adaptation and speculate about reflex modulation as an underlying mechanism for adaptation of gait asymmetry in healthy adults.