Using joint actuators to drive the skeletal movements is a common practice in character animation, but the resultant torque patterns are often unnatural or infeasible for real humans to achieve. On the other hand, physiologicallybased models explicitly simulate muscles and tendons and thus produce more human-like movements and torque patterns. This paper introduces a technique to transform an optimal control problem formulated in the muscleactuation space to an equivalent problem in the joint-actuation space, such that the solutions to both problems have the same optimal value. By solving the equivalent problem in the joint-actuation space, we can generate humanlike motions comparable to those generated by musculotendon models, while retaining the benefit of simple modeling and fast computation offered by joint-actuation models. Our method transforms constant bounds on muscle activations to nonlinear, state-dependent torque limits in the joint-actuation space. In addition, the metabolic energy function on muscle activations is transformed to a nonlinear function of joint torques, joint configuration and joint velocity. Our technique can also benefit policy optimization using deep reinforcement learning approach, by providing a more anatomically realistic action space for the agent to explore during the learning process. We take the advantage of the physiologically-based simulator, OpenSim, to provide training data for learning the torque limits and the metabolic energy function. Once trained, the same torque limits and the energy function can be applied to drastically different motor tasks formulated as either trajectory optimization or policy learning.

Human-like balance controllers are desired for wearable exoskeletons in order to enhance human-robot interaction. Momentum-based controllers (MBC) have been successfully applied in bipeds, however, it is unknown to what degree they are able to mimic human balance responses. In this paper, we investigated the ability of an MBC to generate humanlike balance recovery strategies during stance, and compared the results to those obtained with a linear full-state feedback (FSF) law. We used experimental data consisting of balance recovery responses of nine healthy subjects to anteroposterior platform translations of three different amplitudes. The MBC was not able to mimic the combination of trunk, thigh and shank angle trajectories that humans generated to recover from a perturbation. Compared to the FSF, the MBC was better at tracking thigh angles and worse at tracking trunk angles, whereas both controllers performed similarly in tracking shank angles. Although the MBC predicted stable balance responses, the human-likeness of the simulated responses generally decreased with an increased perturbation magnitude. Specifically, the shifts from ankle to hip strategy generated by the MBC were not similar to the ones observed in the human data. Although the MBC was not superior to the FSF in predicting human-like balance, we consider the MBC to be more suitable for implementation in exoskeletons, because of its ability to handle constraints (e.g. ankle torque limits). Additionally, more research into the control of angular momentum and the implementation of constraints could eventually result in the generation of more human-like balance recovery strategies by the MBC.

Postural responses to similar perturbations of standing balance vary widely within and across subjects. Here, we identified two sources of variability and their interactions by combining experimental observations with computational modeling: differences in posture at perturbation onset across trials, and differences in task-level goals across subjects. We first collected postural responses to unpredictable backward support-surface translations during standing in 10 young adults. We found that maximal trunk lean in postural responses to backward translations were highly variable both within (mean of ranges 28.3°) and across subjects (range of means 39.9°). Initial center of mass (COM) position was correlated with maximal trunk lean during the response but this relation was subject specific (R² = 0.29 - 0.82). We then used predictive simulations to assess causal relations and interactions with task-level goal. Our simulations showed that initial posture explains the experimentally observed intra-subject variability with a more anterior initial COM position increasing the use of the hip strategy. Differences in task-level goal explain observed inter-subject variability with prioritizing effort minimization leading to ankle strategies and prioritizing stability leading to hip strategies. Interactions between initial posture and task-level goal explain observed differences in intra-subject variability across subjects. Our findings suggest that variability in initial posture due to increased sway as observed in older adults might increase the occurrence of less stable postural responses to perturbations. Insight in factors causing movement variability will advance our ability to study the origin of differences between groups and conditions.

Optimal control simulations have shown that both musculoskeletal dynamics and physiological noise are important determinants of movement. However, due to the limited efficiency of available computational tools, deterministic simulations of movement focus on accurately modelling the musculoskeletal system while neglecting physiological noise, and stochastic simulations account for noise while simplifying the dynamics. We took advantage of recent approaches where stochastic optimal control problems are approximated using deterministic optimal control problems, which can be solved efficiently using direct collocation. We were thus able to extend predictions of stochastic optimal control as a theory of motor coordination to include muscle coordination and movement patterns emerging from non-linear musculoskeletal dynamics. In stochastic optimal control simulations of human standing balance, we demonstrated that the inclusion of muscle dynamics can predict muscle co-contraction as minimal effort strategy that complements sensorimotor feedback control in the presence of sensory noise. In simulations of reaching, we demonstrated that nonlinear multi-segment musculoskeletal dynamics enables complex perturbed and unperturbed reach trajectories under a variety of task conditions to be predicted. In both behaviors, we demonstrated how interactions between task constraint, sensory noise, and the intrinsic properties of muscle influence optimal muscle coordination patterns, including muscle co-contraction, and the resulting movement trajectories. Our approach enables a true minimum effort solution to be identified as task constraints, such as movement accuracy, can be explicitly imposed, rather than being approximated using penalty terms in the cost function. Our approximate stochastic optimal control framework predicts complex features, not captured by previous simulation approaches, providing a generalizable and valuable tool to study how musculoskeletal dynamics and physiological noise may alter neural control of movement in both healthy and pathological movements.

Both resistance training (RT) and perturbation-based training (PBT) have been proposed and applied as interventions to improve reactive balance performance in older adults. PBT is a promising approach but the adaptations in underlying balance-correcting mechanisms through which PBT improves reactive balance performance are not well-understood. Besides it is unclear whether PBT induces adaptations that generalize to movement tasks that were not part of the training and whether those potential improvements would be larger than improvements induced by RT. We performed two training interventions with two groups of healthy older adults: a traditional 12-week RT program and a 3-week PBT program consisting of support-surface perturbations of standing balance. Reactive balance performance during standing and walking as well as a set of neuro-muscular properties to quantify muscle strength, sensory and motor acuity, were assessed pre- and post-intervention. We found that both PBT and RT induced training specific improvements, i.e., standing PBT improved reactive balance during perturbed standing and RT increased strength, but neither intervention affected reactive balance performance during perturbed treadmill walking. Analysis of the reliance on different balance-correcting strategies indicated that specific improvements in the PBT group during reactive standing balance were due to adaptations in the stepping threshold. Our findings indicate that the strong specificity of PBT can present a challenge to transfer improvements to fall prevention and should be considered in the design of an intervention. Next, we found that lack of improvement in muscle strength did not limit improving reactive balance in healthy older adults. For improving our understanding of generalizability of specific PBT in future research, we suggest performing an analysis of the reliance on the different balance-correcting strategies during both the training and assessment tasks.