An optimal-control–based methodology for identifying the motor control policy and predicting internal body forces in human locomotion

In the last two decades, in silico approaches based on neuromusculoskeletal modeling and simulation have become a pillar in the biomechanics as they allow to noninvasively estimate the internal body forces generated by musculotendon actuators and articular contact during motion. Such quantities are de facto unmeasurable in vivo, but their knowledge has profound implications in several health-related fields, including the design and control of assistive and rehabilitation devices, ergonomics, as well as planning of rehabilitation treatments and surgical interventions. Unfortunately, current estimates of internal body forces exerted during motion are not sufficiently accurate to be used in the clinical practice. Beside the realism of the musculoskeletal model on which simulations are based, the control policy selected for resolving muscle redundancy plays a crucial role. To this end, model-based predictive simulations of human movement constitute a precious resource for a deeper understanding of the neuromotor control policies encoded by the central nervous system. However, their potential is not fully realized yet, making it difficult to draw convincing conclusions about the actual optimality principles underlying human movement. The present research deals with the development of a principled and robust tool to test candidate physiologically-inspired motor control objectives for investigating the motor control policy during human locomotion and for estimating the internal body forces. Although unimpaired human walking was analyzed here, the same methods can be equally applied to investigate other motor tasks as well as impaired movement. The proposed strategy was devised as a bilevel, inverse optimal control framework based on a full-body three-dimensional neuromusculoskeletal model. In the lower level, prediction of walking is formulated as a principled multi-objective optimal control problem based on a weighted Chebyshev metric, whereas the contributions of candidate control objectives are systematically and efficiently identified in the upper level. Thanks to the attention placed on mitigating the issues related to local minima, the obtained results were analyzed with a good level of confidence. With respect to current predictive approaches, the proposed framework has proved to be effective in determining the contributions of the selected objectives and in reproducing salient features of human locomotion. The trajectories of the estimated lower-limb joint loads resemble those of their experimental measurements, especially when modeling shock-absorption mechanisms. Promising results were obtained from a preliminary sensitivity analysis, suggesting that the strategy developed is quite stable against uncertainties in some model’s parameters. Nonetheless, some deviations between predicted and experimental trajectories have emerged, indicating potential directions for future research. The proposed optimal control framework is general enough for investigating other motor tasks, as well as individuals with disabilities or conditions affecting the musculoskeletal system, with the ultimate goal of learning the motor control policy that best explain observed human motion.