INSIGHTS INTO HUMAN LOCOMOTION STRATEGIES AND MOTOR LEARNING FOR AN AGEING POPULATION USING TRANSFER TESTING AND COUPLED SIMULATIONS
Abstract
The elucidation of human locomotion strategies has potential applications in the prevention of sarcopenia and in the reduction of falls. Given the diverse biochemical, mechanical and functional age-related changes seen in the neuro-musculoskeletal system, the decline in motor function is difficult to study experimentally. In this study, we use transfer testing and coupled simulation strategies within a deep reinforcement learning environment to better understand the complex problem of motor control adaptation to age-related changes. Using transfer testing, a 3D musculoskeletal model is separately trained on parameters of the young adult model (Y) for either forward or backward falls after completing two steps forward, and tested using a 30% age-related reduction for all parameters (M_all). This strategy produces a backward fall for a forwardly trained simulation, showing potential sensitivity of these parameters to a given fall direction. Second, a coupled simulation solution is used to simulate recovery from falls by considering the center-of-mass position relative to the base of support. Results for the M_all trained model showed a longer simulation time and a greater vertical pelvis velocity with a maximal value of 4.26m/s. In particular, the results of the coupled simulations clearly show that both the young and M_all condition models respond with a step back and stronger leg extensor activations to propel the model forward to recover from the simulated fall. We developed a novel coupling between transfer testing and coupled simulation strategies to improve upon muscle models for characterizing muscle function, and also to begin testing different hypotheses, such as the strategy and force required to avoid a fall at different limits. This opens new avenues for precision rehabilitation with patient-specific muscle-driven recovery exercises.
References
- 1. United Nations, World population prospects: The 2017 revision, key findings & advance tables, Working Paper No. ESA/P/WP/248, 2017. Google Scholar
- 2. , The epidemiology of elderly falls attended by emergency medical services in Victoria, Australia, Injury 49 :1712–1719, 2018. Crossref, Web of Science, Google Scholar
- 3. , Resistance training as treatment for sarcopenia: Examining sex-related differences in physiology and response, Clin Ther 144 :33–40, 2022. Crossref, Web of Science, Google Scholar
- 4. , Aging, obesity, sarcopenia and the effect of diet and exercise intervention, Exp Gerontol 155 :1111561, 2021. Crossref, Web of Science, Google Scholar
- 5. , Effects of exercise on muscle mass, strength, and physical performance in older adults with sarcopenia: A systematic review and meta-analysis according to the EWGSOP criteria, Exp Gerontol 151 :111420, 2021. Crossref, Web of Science, Google Scholar
- 6. , Progressive resistance and balance training for falls prevention in long-term residential aged care: A cluster randomized trial of the Sunbeam program, J Am Med Dir Assoc 19(14) :361–369, 2018. Crossref, Web of Science, Google Scholar
- 7. , Assessments of trunk postural control within a fall-prevention training program for service members with lower limb trauma and loss, Gait Posture 92 :493–497, 2022. Crossref, Web of Science, Google Scholar
- 8. , A pilot study of reactive balance training using trips and slips with increasing unpredictability in young and older adults: Biomechanical mechanisms, falls and clinical feasibility, Clin Biomech 67 :171–179, 2019. Crossref, Web of Science, Google Scholar
- 9. , Multicomponent exercises including muscle power training enhance muscle mass, power output, and functional outcomes in institutionalized frail nonagenarians, Age (Dordr) 36(12) :773–785, 2014. Crossref, Google Scholar
- 10. , The effects of plantar perception training on balance and falls efficacy of the elderly with a history of falls: A single-blind, randomized controlled trial, Arch Gerontol Geriatr 77 :19–23, 2018. Crossref, Web of Science, Google Scholar
- 11. , Targeting age-dependent functional and metabolic decline of human skeletal muscle: The geroprotective role of exercise, myokine IL-6, and vitamin D, Int J Mol Sci 21(3) :1010, 2020. Crossref, Web of Science, Google Scholar
- 12. , Theories and control models and motor learning: Clinical applications in neurorehabilitation, Neurologia 30(11) :32–41, 2015. Crossref, Web of Science, Google Scholar
- 13. , Age-related decline in cortical inhibitory ton strengthens motor memory, NeuroImage 245 :118681, 2021. Crossref, Web of Science, Google Scholar
- 14. , Neural plasticity in the ageing brain, Nat Rev Neurosci 7 :30–40, 2006. Crossref, Web of Science, Google Scholar
- 15. , Learning to Run challenge: Synthesizing physiologically accurate motion using deep reinforcement learning, in The NIPS ’17 Competition: Building Intelligent Systems,
The Springer Series on Challenges in Machine Learning , Springer, Cham, pp. 101–120, 2018. Crossref, Google Scholar - 16. , Deep reinforcement learning for modeling human locomotion control in neuromechanical simulation, J NeuroEng Rehabil 18 :126, 2021. Crossref, Web of Science, Google Scholar
- 17. , Human locomotion with reinforcement learning using bioinspired reward reshaping strategies, Med Biol Eng Comput 59(11) :243–256, 2021. Crossref, Web of Science, Google Scholar
- 18. , Deep reinforcement learning coupled with musculoskeletal modelling for a better understanding of elderly falls, Med Biol Eng Comput 60(16) :1745–1761, 2022. Crossref, Web of Science, Google Scholar
- 19. , Multi-level transfer learning for improving the performance of deep neural networks: Theory and practice from the tasks of facial emotion recognition and named entity recognition, Appl Soft Comput 109 :107491, 2021. Crossref, Web of Science, Google Scholar
- 20. , Ensemble transfer learning-based multimodal sentiment analysis using weighted convolutional neural networks, Inf Process Manag 59(13) :102929, 2022. Crossref, Web of Science, Google Scholar
- 21. , Predictive neuromechanical simulatins indicate why walking performance declines with ageing, J Physiol 596(7) :1199–1210, 2018. Crossref, Web of Science, Google Scholar
- 22. Lillicrap TP, Hunt JJ, Pritzel A, Heess NM, Erez T, Tassa Y, Silver D, Wierstra D, Continuous control with deep reinforcement learning, preprint, arXiv:1509.02971 [cs.LG], 2016. Google Scholar
- 23. , Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people an International Academy on Nutrition and Aging (IANA) Task Force, J Nutr Health Aging 13(110) :881–889, 2009. Crossref, Google Scholar
- 24. , Kinematic analysis of video-captured falls experienced by older adults in long-term care, J Biomech 48 :911–920, 2015. Crossref, Web of Science, Google Scholar
- 25. , The kinematics and strategies of recovery steps during lateral losses of balance in standing at different perturbation magnitudes in older adults with varying history of falls, BMC Geriatr 20 :1249, 2020. Crossref, Web of Science, Google Scholar
- 26. , Coupled artificial neural networks to estimate 3D whole-body posture, lumbosacral moments, and spinal loads during load-handling activities, J Biomech 102 :109332, 2020. Crossref, Web of Science, Google Scholar
- 27. , Micro-FEM orthogonal cutting model for bone using microscope images enhanced via artificial intelligence, Procedia CIRP 8 :385–390, 2013. Crossref, Google Scholar
- 28. , Real-time human pose estimation on a smart walker using convolutional neural networks, Expert Syst Appl 184 :115498, 2021. Crossref, Web of Science, Google Scholar
