This study presents an adaptive biarticular thigh exosuit (BATEX) that enhances walking efficiency and speed through bioinspired actuation and a self-tuned, ground reaction force (GRF)-driven controller. The BATEX employs Force-Modulated Compliance (FMC) to coordinate hip and knee assistance via artificial biarticular muscles, with users tuning a single control gain to personalize assistance. Twelve healthy adults participated in experiments evaluating preferred walking speed (PWS), preferred transition speed (PTS), metabolic cost, and muscle activation under no-exosuit (NE), zero-torque (ZT), and assisted (EXO) conditions. Compared with NE, BATEX increased PWS by 14.3% and PTS by 9.0% ( ${p} \lt {0}.{01}$ ). Furthermore, BATEX assistance can also reduce net metabolic cost at PWS of unassisted walking by 9.5% relative to NE and 17.9% relative to ZT ( ${p} \lt {0}.{01}$ ). Electromyography revealed significant reductions in lower-limb muscle activation, particularly in the biceps femoris, gluteus maximus, and gastrocnemius. Moreover, positive exosuit mechanical power was correlated with changes in metabolic cost, indicating that mechanical assistance effectively offset biological energy expenditure. These results demonstrate that GRF-driven, user-tuned biarticular assistance effectively improves walking economy and gait speed in healthy adults, providing a framework that could be adapted in future studies to enhance mobility in broader populations.

Human locomotion involves complex coordination of over-actuated muscle systems and joints, making simulation and control design highly challenging. While recent reinforcement and imitation learning methods can replicate human-like kinematics, they often fail to produce physiologically realistic force patterns, largely due to the limited availability and consideration of reference force plate or electromyography (EMG) data. This paper presents a hybrid imitation learning framework that integrates muscle-driven simulations with reinforcement learning to address over-actuation and to account for the fact that the available data reflects closed-loop actions involving muscle control. A key contribution is the incorporation of metabolic cost into the reward function, shaping energetically efficient and physiologically plausible controllers. Simulation results demonstrate that the learned policies generate muscle activations and ground reaction forces that align more closely with OpenSim references and experimental data than standard imitation learning. The method provides a scalable tool for developing and validating closed-loop control strategies for assistive systems such as exoskeletons and prostheses, while highlighting broader implications for learning-based control of over-actuated biomechanical systems.

Effective coordination between the human neuromuscular system and wearable assistive devices remains a key challenge in enhancing gait performance. We propose a concerted control strategy synchronizing biological and artificial actuators using shared feedback. Positioned between centralized (e.g., CPG) and distributed (e.g., reflex-based) control, this approach avoids a central controller by relying on a coordinating signal. Ground reaction force (GRF) emerged as a strong candidate for this role. To implement this concept, we use Force Modulated Compliance (FMC) - a control mechanism that adjusts joint stiffness based on real-time GRF input. FMC has been validated in simulations and robotic platforms, confirming its ability to synchronize joint actuation. We applied this strategy in an active soft biarticular thigh exosuit (BATEX) and tested it in human walking experiments. The GRF-informed controller increased preferred walking speed, advanced the walk-to-run transition, and reduced metabolic cost. These results highlight the effectiveness of GRF-based control in enhancing human-exosuit coordination and aligning assistance with natural gait dynamics. This bioinspired approach offers a scalable framework for real-world locomotion support by harmonizing human and robotic contributions.

Postural balance is essential for both humans and robots, as failures increase fall risk and limit robotic performance in real-world settings. Although humans and robots share fundamental balancing mechanics, biological complexity limits the isolation of individual muscle functions and direct principle transfer to robots. In this study, we use EPA-Walker, a bio-inspired robot actuated by electric motors and pneumatic artificial muscles (PAMs), as a physical platform to investigate perturbed standing. Here, we focus on the PAM-driven actuation to systematically examine the roles of muscle morphology and control, and to validate biomechanical findings in a robotic setting. To enable a clear upper body perturbation, a Control Moment Gyroscope (CMG) was integrated. We evaluated two stabilization paradigms: passive standing, in which joint compliance was tuned through static PAM pressurization, and active balancing, in which a bio-inspired ground reaction force (GRF) feedback controller generated muscle reflexes. The results showed that biarticular thigh muscles, particularly the hamstrings and rectus femoris, played the most prominent role in enhancing robustness through both morphology in the passive experiments and reflex control in the active experiments, consistent with findings from human perturbation studies. While ankle muscles, such as the soleus, were essential for stable standing mainly through their passive morphological contribution, their reflex-based action could also improve robustness in a more specific manner through center-of-pressure regulation. Activating a single muscle could significantly improve robustness beyond morphology, enabling recovery from 3 Nm perturbations. Furthermore, synergistic reflex of biarticular muscles, especially hamstrings and gastrocnemius, extends the robustness to larger perturbations (5 Nm). Our contribution highlights the synchronization of control strategies with the underlying morphological design through a universal sensory feedback signal, namely, GRF. These findings demonstrate the value of bio-inspired robots as testbeds to understand the potential principles underlying human motor control and support the transfer of such principles to legged robots and assistive systems.

Abstract Optimizing assistive wearable devices is crucial for their efficacy and user adoption, yet state-of-the-art methods like Human-in-the-Loop Optimization (HILO) and biomechanical modeling face limitations. HILO is time-consuming and often restricted to optimizing control parameters, while inverse dynamics assumes invariant kinematics, which is unreliable for adaptive human-device interaction. Predictive simulation offers a powerful alternative, enabling computational exploration of design spaces. However, existing approaches often lack systematic optimization frameworks and rigorous validation against experimental data. To address this, we developed a Design Optimization Platform that integrates predictive simulations within a two-level optimization structure for personalizing assistive device design. This paper primarily validates the platform’s predictive simulations against a publicly available dataset of the passive Biarticular Thigh Exosuit (BATEX), assessing its reliability. Our findings show that the model can sufficiently predict the kinematics and major muscle activations, except for the pelvis tilt and some biarticular muscles. The key finding is that successful identification of personalized optimal BATEX stiffness parameters needs acceptable prediction of metabolic cost trends, not their precise values. Our analysis further reveals that the model’s accuracy in predicting Vasti muscle activation in the baseline condition is a significant indicator of its success in predicting metabolic cost trends. This demonstrates that accurate prediction of performance trends is more important for effective simulation-based design optimization than perfect biomechanical accuracy, advancing targeted and efficient assistive device development.

Human walking is inherently three-dimensional, requiring coordinated control in multiple planes of motion. Among these, controls in the sagittal and frontal planes play central roles in propulsion and balance. While sagittal plane control has been extensively modelled, the control mechanisms underlying frontal plane stability remain less understood. In this study, we hypothesize that frontal plane control can be achieved through a simple yet biologically inspired strategy: modulating hip compliance in response to ground reaction forces (GRFs). To validate this, we extend our previously proposed bipedal walking control framework, termed Concerted Control, to three dimensions and implement this control law at both the joint and muscle levels in bipedal models. Across a wide range of walking speeds (0.7-1.8 m s-1), both implementations yield dynamically stable three-dimensional gaits without any movement constraints, and their hip abduction-adduction torque profiles closely match human data, achieving cross-correlation values above 0.86 in most conditions. These results demonstrate that GRF-based hip compliance modulation is a parsimonious and robust control principle for frontal plane stability. More broadly, our findings highlight the potential of GRF-driven compliance control as a unifying strategy for neuromechanical modelling and bio-inspired bipedal locomotion.

Abstract Evaluating bioinspired design principles in wearable assistive devices provides a unique opportunity to interrogate our understanding of the critical factors that enable agile, stable, and economical human movement. We introduce the BiArticular Thigh EXosuit (BATEX), a wearable device integrating two morphological features found in biological legged systems: biarticular muscles and elastic tissues. BATEX employs two biarticular springs spanning the hip and knee to emulate the human rectus femoris and hamstring muscles, creating beneficial synergy to enhance walking economy. This design enables two energy-shuffling mechanisms: temporal (spring-like storage/return at a joint) and spatial (strut-like transfer across joints). In walking experiments at 1.3 m/s with N = 9 participants, a single compliant biarticular spring yielded a 7% metabolic cost reduction compared to walking without BATEX. Individually optimized configurations further improved metabolic reduction to 9%. BATEX morphology allowed users not only to off-load biological joint power (Assist) but also to increase total power (Augment). Across all exosuit configurations, the mechanical impact of the exosuit was reflected by a significant correlation between changes in users’ biarticular muscles’ activity and changes in net metabolic rate. In sum, compliant-biarticular exosuit architectures can concurrently assist and augment human lower-limb joint function, providing significant metabolic savings during walking.

Personalized exoskeleton assistance has substantial potential to enhance human locomotion performance. However, current human-in-the-loop optimization methods for generating personalized assistance are cumbersome and time-consuming. Since humans can perceive locomotion through internal sensory feedback, user preference-based self-tuning may facilitate the individualization of exoskeleton assistance to meet individual needs. Here, we explore a user-driven human-in-the-loop tuning approach for walking assistance, hypothesizing that individuals can quickly find their preferred personalized assistance through subjective perception. We conducted experiments with 11 healthy participants, who were instructed to tune four control parameters while wearing a hip exoskeleton and walking on a treadmill. The tuning procedure concluded when participants indicated that they had found their preferred assistance. Then we surveyed the sense of agency to assess the user experience. We evaluated the effort of walking with the preferred setting and explored the metabolic cost landscape around it. Participants identified their preference in 10.9 ± 0.9 min, while testing 30.5 settings and spending 18.7 s per setting on average. Preferred assistance profiles varied widely between participants, with timing differences of up to 22.5% of the stride time. The metabolic cost of walking with the preferred assistance was reduced by 16.6 ± 1.1% compared to walking with the exoskeleton in a zero-torque condition. Timing deviations of up to ±8% of the stride time did not significantly affect metabolic cost reduction, indicating the robustness of the preferred assistance profiles. Significant changes in the sense of agency between unassisted and assisted walking demonstrate its sensitivity to partial exoskeleton assistance. The results highlight the potential of preference-based user-tuning while suggesting that additional guidance throughout the user-tuning procedure may support a systematic exploration, thereby advancing the preference-based individualization of exoskeleton assistance.

Walking on sloped terrain requires substantial mechanical and control adaptations for effective energy management compared to level ground locomotion. The Virtual Pivot Point (VPP) hypothesis explains sagittal plane angular momentum regulation during level walking, but its validity in slope walking remains unexplored. This study combines human experiments with template-model simulations to investigate how the VPP strategy is modulated during slope walking. Participants walked on an instrumented ramp at various inclinations (0°, ± 7.5°, ± 10°), while a 2D spring-loaded inverted-pendulum model with a trunk segment simulated the task. Experimental results confirmed that the VPP is a robust feature of slope walking ( R 2 > 0.975). The simulation reproduced the change in hip torque and trunk adaptations by modulating VPP position. Results of this study indicate that VPP position and trunk dynamics could afford stability and energy management on gentle slopes, but to robustly navigate steeper ramps, humans recruit a multi-joint strategy where the knee and ankle joints play a crucial role in managing the energetic demands of sloped terrain. Beyond advancing our understanding of locomotor control, these insights have practical implications for the design of exoskeletons that adapt to uneven terrain.

Passive exosuits offer potential for reducing the metabolic cost of walking, but evaluating their effectiveness is challenging due to difficulties in measuring internal biomechanics and muscle-level energetics. This study investigates the metabolic effects of the BATEX, a passive biarticular thigh exosuit assisting hip and knee flexion/extension. The exosuit features a spring parallel to the rectus femoris (RF) and a spring parallel to the hamstring (HAM). We combined experimental data from ten healthy adults walking on a treadmill under various exosuit conditions (different stiffness and configurations) with EMG-informed musculoskeletal simulations to estimate muscle-level behavior. The simulations aimed to replicate experimental metabolic measures, analyze cost fluctuations across the gait cycle, and elucidate muscle-tendon energetic responses, comparing individuals experiencing reductions (positive responders, PR) versus increases (negative responders, NR) in metabolic cost. Simulation results successfully replicated experimental whole-body metabolic cost trends for most participants (8/10, ${r} \gt {0}.{7}$ , ${p}\lt {0}.{05}$ ), showing comparable overall net reductions primarily driven by significant savings during the swing phase. Combining RF and HAM springs resulted in complex, non-additive interactions. Muscle-level analysis identified both targeted effects on hip/knee flexors/extensors and non-targeted effects on ankle and hip ab/adductor muscles, with responses differing markedly between PR and NR groups. Distinct energy-saving mechanisms were associated with each spring type (RF: reduced activation; HAM: reduced fiber velocity). The differing responses of the PR and NR groups to the same device setup motivate personalization, and our muscle-level insights offer a foundation for future device refinement.

Background/Objectives: This study investigated the biomechanics of a hip exoskeleton during sit-to-stand transitions. Methods: Eleven participants performed the task under three conditions: without the exoskeleton (No Exo), wearing the exoskeleton without assistance (Exo Off), and wearing it with hip extension assistance (Exo On). Results: The analyses revealed that joint angles (hip, knee, and ankle) and vertical ground reaction forces were comparable across all conditions. However, Exo Off significantly increased transition time, whereas Exo On did not differ significantly from No Exo. Additionally, both exoskeleton conditions led to increased integrated EMG (iEMG) activity in the rectus femoris, vastus medialis, and gluteus maximus-likely due to the added device mass. Notably, iEMG analysis revealed a significant reduction in gluteus maximus activity in Exo On compared to Exo Off. Conclusions: Despite providing only moderate torque assistance (0.12 Nm/kg), the results suggest that well-timed exoskeleton support can partially reduce the physical demands of sit-to-stand transitions. However, the observed reduction in gluteus maximus activity was limited, likely reflecting the combined effects of the assistance strategy, including its magnitude and timing, user adaptation and training, postural demands due to device weight and external torques, and mechanical constraints such as potential joint misalignment. Further research is needed to optimize hip exoskeleton support for daily activities.

Design codes for pedestrian-induced vibration on footbridges treat each pedestrian as a prescribed vertical force on the deck. This convention ignores the feedback by which deck motion modifies the gait, and at resonance overpredicts the measured field amplitudes.This paper introduces a load model for vertical human-structure interaction that closes this feedback loop. Each pedestrian is modelled as a single-degree-of-freedom oscillator whose leg stiffness varies over the gait cycle. On a vibrating deck this periodic stiffness couples the pedestrian to the deck motion. The per-step force on the bridge is then the classical rigid-ground walking force plus an interaction component proportional to the deck velocity, which acts on the deck like an amplitude-dependent damper. At resonance this damping is negative and the gait pumps energy into the deck. The response self-limits at a finite equilibrium amplitude, because the negative damping weakens as the deck amplitude rises. Away from resonance the damping stays positive.A main advantage of the proposed model is that its stability criterion and equilibrium amplitude require no prescribed driving force, unlike classical moving-force models, and that its parameters transfer to any bridge through the first-mode frequency, with the amplitude scales of the coupling tied to the comfort thresholds of design codes. A second advantage is that for groups the per-pedestrian damping contributions add linearly across the deck.Without recalibration, the model reproduces the reported field amplitudes of four independently documented footbridges within ratios 0.93 to 1.43, while classical moving-force models overestimate them by factors 3.2 to 33.The framework yields a closed-form stability criterion and a self-limiting equilibrium amplitude, both obtained without coupled crowd simulation and suited to footbridge serviceability assessment.

This study addresses the vertical human-structure interaction (vHSI) and its effect on vertical ground reaction forces (vGRFs) as well as on gait adaptation in terms of cadence adjustment by analyzing the gait of individual subjects crossing an experimental pedestrian bridge. A comprehensive study was performed on 26 subjects to compare GRFs of two structural configurations. In the single-span configuration, the structure is susceptible to human-induced vibrations, while in the second system configuration, the bridge can be regarded as rigid ground, making it ideal for studying the influence of vHSI. The results show a positive correlation between increasing bridge acceleration and increasing peak vGRF values. Furthermore, subjects with a preferred cadence in the range of the structure's natural frequency tend to synchronize with it. Although the effect of vHSI on gait adaptation and the change in vGRFs is clearly shown, this work also demonstrates a high uncertainty of vHSI effects due to inter- and intravariability. The results emphasize the need for further studies on active HSI effects on human gait to base structural load modeling on more valid assumptions.

Personalized and effective control is essential for the acceptance of wearable robotic systems such as lower-limb exoskeletons. This paper introduces the adaptation and control problem in these systems, outlines key challenges, and presents a hierarchical control framework considering the BiArticular Thigh EXosuit (BATEX) as an example. Lower-limb exoskeletons are wearable robots that assist walking by applying joint-level torques in coordination with the user. The proposed architecture includes a low-level hybrid controller for velocity tracking, a mid-level neuromechanical controller for stiffness modulation based on biomechanical feedback, and a high-level user-in-the-loop gain adaptation. To explore the role of predictive methods, a model predictive controller is implemented at the actuator level, improving force tracking, disturbance rejection, and constraint handling compared to conventional control, outlining the potential of a unified predictive control framework exploited at the different levels. Experiments with human subjects indicate enhanced gait performance, demonstrating the promise of hierarchical and predictive control for adaptive, user-centered assistance in wearable robotics.

This letter proposes a bio-inspired, simple, and easy-to-implement walking controller, termed Concerted Control, which leverages a shared common signal to coordinate movements across multiple joints without requiring predefined trajectories. The backbone of this controller is our previously developed Force Modulated Compliance (FMC) control concept, which modulates joint stiffness using ground reaction force (GRF). In Concerted Control, FMC is implemented across multiple joints, allowing implicit coordination through the shared GRF signal in absence of any centralized controller. We tested the performance of this controller on a simulated bipedal walker model and demonstrated that Concerted Control can generate human-like walking gaits across a wide range of speeds, from 0.7 to $1.8\,\text {m/s}$. Additionally, we assessed the robustness of these gaits against external angular momentum perturbations, and the results showed a high level of robustness. Concerted Control offers a promising approach for enhancing the control of bipedal robots and assistive systems.

Human locomotion exhibits extraordinary adaptability and robustness, yet the mechanisms by which lower limbs adjust to sudden environmental disruptions remain poorly understood. To address this, we employed the bioinspired human-sized EPA-Hopper II robot to examine how lower-limb joints recover from an abrupt drop in ground height, mimicking unexpected perturbations encountered in natural settings. Our study investigates the roles of the monoarticular soleus (SOL) and biarticular gastrocnemius (GAS) muscle configurations, focusing on how their compliance influences the robot’s hopping stability. Experiments reveal that a coordinated interplay between SOL and GAS markedly improves recovery from disturbances, enhancing energy distribution and joint synchronization. Detailed kinematic and power analyses show that GAS facilitates energy transfer across joints, while SOL’s spring-like properties support rapid recovery. These results highlight how bioinspired muscle arrangements enable robust locomotion through intrinsic mechanical interactions. By leveraging a robotic platform to probe these dynamics, this work deepens our understanding of biological locomotion and informs the design of bioinspired bipedal robots and prosthetics capable of thriving in unpredictable environments.

Legged robots, particularly quadrupeds, excel at navigating rough terrains, yet their performance under vertical ground perturbations, such as those from oscillating surfaces, remains underexplored. This study introduces a novel approach to enhance quadruped locomotion robustness by training the Unitree Go2 robot on an oscillating bridge - a 13.24-meter steel-and-concrete structure with a 2.0 Hz eigenfrequency designed to perturb locomotion. Using Reinforcement Learning (RL) with the Proximal Policy Optimization (PPO) algorithm in a MuJoCo simulation, we trained 15 distinct locomotion policies, combining five gaits (trot, pace, bound, free, default) with three training conditions: rigid bridge and two oscillating bridge setups with differing height regulation strategies (relative to bridge surface or ground). Domain randomization ensured zero-shot transfer to the real-world bridge. Our results demonstrate that policies trained on the oscillating bridge exhibit superior stability and adaptability compared to those trained on rigid surfaces. Our framework enables robust gait patterns even without prior bridge exposure. These findings highlight the potential of simulation-based RL to improve quadruped locomotion during dynamic ground perturbations, offering insights for designing robots capable of traversing vibrating environments.

Precise and portable ground reaction force (GRF) measurement is critical for advancing biomechanical gait analysis and enabling more effective control of robots and assistive devices. This study investigates vertical GRF estimation during walking using a soft, lightweight, and cost-effective 3D-printed ferroelectret insole. The insole design incorporates four monolithically 3D-printed piezoelectric sensors positioned under key foot contact areas, which generate nonlinear voltage in response to applied forces. A 1-D convolutional neural network (CNN), featuring two convolutional and two fully connected layers, was trained to predict vertical GRF across five different walking speeds (50-150% of normal walking speed). The CNN was validated using K-fold cross-validation, enhancing model generalization. Results showed an average root-mean-squared error of 9.24% and $R<^>{2}$ values exceeding 0.99 across different speeds, demonstrating the potential of 3D-printed ferroelectret sensors for portable GRF measurement in gait analysis and robotics applications.

Lower limb exoskeletons and exosuits have shown promise in augmenting human physical capabilities, with applications ranging from rehabilitation to performance enhancement. Accurate evaluation of their impact on metabolic energy expenditure is crucial for optimizing design and control strategies. While experimental measurement of metabolic cost via indirect calorimetry provides direct assessment, it is often impractical outside laboratory settings. Computational models offer an alternative, but their effectiveness in predicting metabolic cost changes induced by assistive devices remains underexplored. This study investigates the impact of incorporating different levels of complexity and sensory information, as well as various metabolic cost models, on estimating muscle metabolic cost during walking with a passive biarticular thigh exosuit. We compare three modeling approaches: joint-space dynamics, musculoskeletal simulation with effort minimization, and EMG-informed musculoskeletal simulation, each employing several metabolic models. Results show that EMG-informed musculoskeletal simulation, particularly using the Uchida (2016) metabolic model, provides the highest accuracy in predicting metabolic cost changes. Musculoskeletal simulation with effort minimization also shows promise, offering a viable alternative without the need for EMG data. These findings highlight the potential of computational models in evaluating and optimizing assistive devices.

Designing optimal assistive wearable devices is a complex task, often addressed using human-in-the-loop optimization and biomechanical modeling approaches. However, as the number of design parameters increases, the growing complexity and dimensionality of the design space make identifying optimal solutions more challenging. Predictive simulation, which models movement without relying on experimental data, provides a powerful tool for anticipating the effects of assistive devices on the human body and guiding the design process. This study aims to introduce a design optimization platform that leverages predictive simulation of movement to identify the optimal parameters for assistive wearable devices. The proposed approach is specifically capable of dealing with the challenges posed by high-dimensional design spaces. The proposed framework employs a two-layered optimization approach, with the inner loop solving the predictive simulation of movement and the outer loop identifying the optimal design parameters of the device. It is utilized for designing a knee exoskeleton with a damper to assist level-ground and downhill gait, achieving a significant reduction in normalized knee load peak value by $ 37\% $ for level-ground and by $ 53\% $ for downhill walking, along with a decrease in the cost of transport. The results indicate that the optimal device applies damping torques to the knee joint during the Stance phase of both movement scenarios, with different optimal damping coefficients. The optimization framework also demonstrates its capability to reliably and efficiently identify the optimal solution. It offers valuable insight for the initial design of assistive wearable devices and supports designers in efficiently determining the optimal parameter set.