Individuals employed in transportation (such as train, airplane, and boat operators), agriculture, construction, and manufacturing industries face an elevated risk of falling when regularly exposed to Foot Transmitted Vibration (FTV); Work-related slips, trips, and falls attributed to vibration represent a significant issue, contributing to substantial injuries and economic repercussions globally and it is important to study the effect of walking on a vibrating floor. A walking humanoid model was built in a previous PhD thesis at Politecnico di Milano with real-time physical engine (Newton Dynamics 3.14) based on real anthropometric data is employed to analyse how the vibration characteristics, amplitude, frequency and direction affect the human gait, including spatiotemporal (step length, width, cadence) metrics, variability and stability-related measures. Mediolateral (ML) and pitch FTV directions are integrated to the existing model by introducing actuator to the floor. Simulations under pitch vibration at frequencies ranging from 0.1 to 0.25 Hz and a fixed angular amplitude of 4°are performed. More epochs of iterations are required to train the controller parameters to obtain the optimal results for the higher frequency pitch vibration. Gait performance is assessed based on the number of consecutive steps completed, and gait quality is quantified by using cadence, step length, step width, their standard deviations. The steps the model can generate decreases from 9 to 2 with the increase of the frequency after the pitch vibration introduced, the step length decreases and step width widens to compensate for the pitch vibration, the stability of the model to maintain balance decreases 56% approximately. Eighteen simulations at frequencies ranging from 0.25 Hz to 2.0 Hz and amplitudes of 0.01 m, 0.05 m, and 0.1 m in medio-lateral direction vibration are conducted by launching the walking humanoid model. Extra 500 iterations of optimization are enough to obtain stable optimal results of ML vibration when the initial guess is set as the output of no vibration condition. The results demonstrate that locomotor performance of the model is frequency- and amplitude-dependent, the step length decreases slightly with the increase of frequency, while the step width increases to compensate for the medio-lateral vibration compared with no vibration. The model only falls down under 2 Hz, 0.1 m ML floor vibration. Multiple simulations are conducted under 2 Hz, 0.1 m ML vibration to examine how weight of cost function influence gait metrics while varying the weight in the control logic applied to the walking humanoid model. The optimal configuration of the weights was established to achieve the maximum stable steps the model can perform coupled with lowest variability with this configuration. The stability and consistency of the simulation results based on the model is evaluated with repeated experiments on humans conducted in previous studies in literature. The model is capable to predict the effect of medio-lateral floor vibration to the gait metrics with high accuracy by evidence that the step length of simulation is exactly the value of experiment data in the literature.
Le persone impiegate nei settori dei trasporti (come operatori ferroviari, aeronautici e navali), dell’agricoltura, dell’edilizia e dell’industria manifatturiera sono esposte a un rischio elevato di cadute quando soggette regolarmente a vibrazioni trasmesse ai piedi (Foot-Transmitted Vibration, FTV). Scivolamenti, inciampi e cadute legati alle vibrazioni rappresentano un problema rilevante, contribuendo in modo significativo a infortuni e impatti economici a livello globale; risulta quindi importante studiare l’effetto della deambulazione su superfici vibranti. Un modello umanoide di camminata è stato sviluppato in una precedente tesi di dottorato presso il Politecnico di Milano, utilizzando un motore fisico in tempo reale (Newton Dynamics 3.14) basato su dati antropometrici reali. Tale modello è impiegato per analizzare come le caratteristiche della vibrazione, ampiezza, frequenza e direzione, influiscano sull’andatura umana, includendo metriche spazio-temporali (lunghezza del passo, larghezza del passo, cadenza), variabilità e misure legate alla stabilità. Le direzioni di vibrazione medio-laterale (ML) e in beccheggio (pitch) sono state integrate nel modello esistente introducendo un attuatore a slitta nel pavimento. Sono state eseguite simulazioni con vibrazione in beccheggio a frequenze comprese tra 0.1 e 0.25 Hz e ampiezza angolare fissa di 4°. Per le vibrazioni a frequenza più elevata è stato necessario un maggior numero di iterazioni per addestrare i parametri del controllore e ottenere risultati ottimali. Le prestazioni del cammino sono state valutate in base al numero di passi consecutivi completati, mentre la qualità del cammino è quantificata tramite cadenza, lunghezza del passo, larghezza del passo e relative deviazioni standard. Con l’aumento della frequenza, dopo l’introduzione della vibrazione in beccheggio, il numero di passi generati dal modello diminuisce da 9 a 2; la lunghezza del passo si riduce e la larghezza aumenta per compensare la perturbazione, mentre la stabilità del modello nel mantenere l’equilibrio si riduce di circa il 56%. Sono state inoltre condotte diciotto simulazioni con vibrazione medio-laterale a frequenze comprese tra 0.25 Hz e 2 Hz e ampiezze di 0.01 m, 0.05 m e 0.1 m, utilizzando il modello umanoide. Ulteriori 500 iterazioni di ottimizzazione risultano sufficienti per ottenere risultati stabili e ottimali nel caso di vibrazione ML quando la stima iniziale è impostata come l’output della condizione senza vibrazione. I risultati mostrano che le prestazioni locomotorie del modello dipendono da frequenza e ampiezza: la lunghezza del passo diminuisce leggermente con l’aumento della frequenza, mentre la larghezza del passo aumenta per compensare la vibrazione medio-laterale rispetto alla condizione senza vibrazione. Il modello cade solo nel caso di vibrazione del pavimento ML a 2 Hz e 0.1 m. Sono state eseguite ulteriori simulazioni nelle condizioni di 2 Hz e 0.1 m per analizzare l’influenza dei pesi della funzione di costo sulle metriche del passo, variando tali pesi nella logica di controllo applicata al modello umanoide. È stata individuata una configurazione ottimale dei pesi che consente di massimizzare il numero di passi stabili e minimizzare la variabilità. La stabilità e la coerenza dei risultati di simulazione sono state valutate confrontandole con esperimenti ripetuti su soggetti umani presenti in letteratura. Il modello è in grado di prevedere con elevata accuratezza l’effetto della vibrazione medio-laterale sulle metriche del passo, come dimostrato dalla corrispondenza esatta della lunghezza del passo tra simulazioni e dati sperimentali riportati in letteratura.
Numerical simulations of the effect of Foot-Transmitted Vibration on the human gait
BAI, YARUI
2026
Abstract
Individuals employed in transportation (such as train, airplane, and boat operators), agriculture, construction, and manufacturing industries face an elevated risk of falling when regularly exposed to Foot Transmitted Vibration (FTV); Work-related slips, trips, and falls attributed to vibration represent a significant issue, contributing to substantial injuries and economic repercussions globally and it is important to study the effect of walking on a vibrating floor. A walking humanoid model was built in a previous PhD thesis at Politecnico di Milano with real-time physical engine (Newton Dynamics 3.14) based on real anthropometric data is employed to analyse how the vibration characteristics, amplitude, frequency and direction affect the human gait, including spatiotemporal (step length, width, cadence) metrics, variability and stability-related measures. Mediolateral (ML) and pitch FTV directions are integrated to the existing model by introducing actuator to the floor. Simulations under pitch vibration at frequencies ranging from 0.1 to 0.25 Hz and a fixed angular amplitude of 4°are performed. More epochs of iterations are required to train the controller parameters to obtain the optimal results for the higher frequency pitch vibration. Gait performance is assessed based on the number of consecutive steps completed, and gait quality is quantified by using cadence, step length, step width, their standard deviations. The steps the model can generate decreases from 9 to 2 with the increase of the frequency after the pitch vibration introduced, the step length decreases and step width widens to compensate for the pitch vibration, the stability of the model to maintain balance decreases 56% approximately. Eighteen simulations at frequencies ranging from 0.25 Hz to 2.0 Hz and amplitudes of 0.01 m, 0.05 m, and 0.1 m in medio-lateral direction vibration are conducted by launching the walking humanoid model. Extra 500 iterations of optimization are enough to obtain stable optimal results of ML vibration when the initial guess is set as the output of no vibration condition. The results demonstrate that locomotor performance of the model is frequency- and amplitude-dependent, the step length decreases slightly with the increase of frequency, while the step width increases to compensate for the medio-lateral vibration compared with no vibration. The model only falls down under 2 Hz, 0.1 m ML floor vibration. Multiple simulations are conducted under 2 Hz, 0.1 m ML vibration to examine how weight of cost function influence gait metrics while varying the weight in the control logic applied to the walking humanoid model. The optimal configuration of the weights was established to achieve the maximum stable steps the model can perform coupled with lowest variability with this configuration. The stability and consistency of the simulation results based on the model is evaluated with repeated experiments on humans conducted in previous studies in literature. The model is capable to predict the effect of medio-lateral floor vibration to the gait metrics with high accuracy by evidence that the step length of simulation is exactly the value of experiment data in the literature.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_07_Bai.pdf
non disponibili
Licenza:
Tutti i diritti riservati
Dimensione
3.68 MB
Formato
Adobe PDF
|
3.68 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/384790
URN:NBN:IT:POLIMI-384790