The high metabolic cost of walking in individuals with a transtibial amputation is often attributed to reduced ankle propulsion. However, the cost of stabilizing walking may also increase after amputation, potentially explaining why restoring ankle propulsion does not always normalize metabolic cost. Because the contributions of propulsion and stabilization are difficult to dissociate experimentally, we used simulations to probe the effects of transtibial amputation and different prostheses on walking cost. We simulated gait using a conceptual planar torque-driven model. Leg joint torques were controlled by feedforward and linear, time-varying full state feedback. We computed feedforward torques associated with propulsion and feedback gains for stabilizing walking that minimized expected effort (sum of torques squared) in the presence of sensorimotor noise while imposing stability. We simulated walking for an intact model and three walkers with a prosthesis: a passive, a feedforward-controlled active, and a feedforward and local feedback-controlled active prosthesis. We solved a deterministic approximation of the resultant stochastic optimal control problems for different levels of sensorimotor noise. Feedforward and expected feedback torques were higher for the passive prosthesis walker than for the intact walker. An active prosthesis restored total biological effort (i.e., effort of all but the prosthesis joint) to the level of the intact walker, but this effort was generated by fewer biological joints. Biological feedback effort was higher in prosthesis walkers than in the intact walker. Adding local feedback to an active prosthesis reduced biological feedback effort, but only at low noise levels. Our results suggest that walking with a passive prosthesis increases effort related to both propulsion and stabilization. While active prostheses can support propulsion, local feedback control only marginally reduced stabilization-related effort, suggesting that controllers require sensory information from the rest of the body to fully restore the metabolic cost of walking.

Walking energy consumption is higher in people with versus without a transtibial amputation, but the underlying reasons are poorly understood. Active prostheses that restore ankle power do not necessarily decrease walking energy consumption, suggesting other reasons than a lack of ankle power for the increased energy consumption. Transtibial amputation impacts walking stability, as evidenced by the increased fall risk, and there is an energetic cost associated with stabilizing walking. It is, however, unclear how transtibial amputation affects the energetic cost of stabilizing walking. We assessed the metabolic cost of stabilizing walking against treadmill belt speed perturbations (SD = 0.13 m/s) in 16 subjects with and 23 subjects without a transtibial amputation at three walking speeds between 0.6 and 1.6 m/s. We focused on sagittal plane stability, as a transtibial amputation disables muscle-driven modulation of the ankle torque, which plays an important role in controlling walking in the sagittal plane at slow speeds. Perturbations induced 0.24 W/kg larger increases in energy consumption across walking speeds in subjects with than without a transtibial amputation. Whereas mean reductions in step length in response to perturbations were similar between groups, individuals with an amputation increased step length variability of their intact leg more than individuals without an amputation, especially at low speeds. As continuous control is required for unperturbed walking, an increased metabolic cost of stabilizing walking might explain-at least partially-the higher energetic cost of walking. These insights are important when seeking to reduce the metabolic cost of walking after transtibial amputation.NEW & NOTEWORTHY The higher metabolic cost of walking in individuals with versus without a transtibial amputation is poorly understood, hindering the design of interventions. We demonstrated that transtibial amputation considerably increases the energy consumption for stabilizing walking in the sagittal plane and increases the reliance on step length adjustments to stabilize walking. This opens perspectives for restoring walking energetics after amputation through prostheses that support sagittal plane stability.

Amputations and other impairments heavily impact human locomotion. To compensate for these disabilities, advanced devices such as active orthoses and prostheses are being developed with the aim of restoring locomotor function. To achieve this goal, the control of such technologies has been identified as a key research challenge. A reference kinematic profile for the affected or missing joints is often useful for device control. This paper introduces a novel method for estimating reference kinematics of leg joints from kinematic signals of the contralateral leg. This method leverages fusion of two different estimators - i.e. Complementary Limb Motion Estimation and Adaptive Oscillators - which are combined through the use of a logistic function. The method is then applied to estimate the kinematics of a hip joint, and the performance is compared with that of each individual estimator during walking. The correlation between the estimated signal and the ground truth is assessed on a cycle-by-cycle basis. During transitions, our new method behaves similarly to Complementary Limb Motion Estimation. After a few cycles, the method shifts towards Adaptive Oscillators to achieve a better correlation during quasi-steady-state locomotion.

There is a metabolic cost associated with stabilizing walking, but it remains unclear to what extent stabilizing walking in the sagittal plane contributes to this cost. Furthermore, strategies for stabilizing walking in the sagittal plane vary with speed, but it is unclear whether this also leads to a speed-dependent metabolic cost of stabilizing walking. Here, we explored the metabolic cost of stabilizing walking in the sagittal plane across speeds and its relationship with control strategies. To this aim, we applied continuous treadmill belt speed perturbations (a standard deviation of 0.13 ms-1) to 22 healthy individuals walking at 0.8, 1.2, and 1.6 ms-1. We evaluated changes in metabolic energy consumption and control strategies between perturbed and unperturbed walking and explored relationships between energy consumption and control strategies. Perturbations induced larger increases in metabolic rate and changes in control strategies at slower than faster walking speeds, suggesting that walking is more robust against perturbations at faster speeds. Perturbations increased the metabolic rate by 16.7% at the slowest vs. 4.6% at the fastest walking speed. When perturbed, subjects took shorter, wider, and more variable steps and variability in ankle muscle activation increased, but most changes were larger at slower speeds. Metabolic rate increased more due to perturbations in individuals who reduced step length more, that is, relied more on anticipatory adjustments of the walking pattern. Our findings are especially relevant to explain the increased metabolic cost of individuals with mobility impairments, who often walk slower and have altered walking control.NEW & NOTEWORTHY Stabilizing walking requires more active control in the frontal than in the sagittal plane. Nevertheless, we demonstrated that there is a considerable energetic cost associated with stabilizing walking in the sagittal plane, especially at slower speeds. This cost is higher in individuals who adjusted their average walking pattern more when walking was perturbed. Diseases that affect both walking speed and control might therefore have a disproportionately large effect on the metabolic cost of walking.

As robots increasingly integrate into the workplace, Human-Robot Collaboration (HRC) has become increasingly important. However, most HRC solutions are based on pre-programmed tasks and use fixed safety parameters, which keeps humans out of the loop. To overcome this, HRC solutions that can easily adapt to human preferences during the operation as well as their safety precautions considering the familiarity with robots are necessary. In this paper, we introduce GPTAlly, a novel safety-oriented system for HRC that leverages the emerging capabilities of Large Language Models (LLMs). GPTAlly uses LLMs to 1) infer users’ subjective safety perceptions to modify the parameters of a Safety Index algorithm; 2) decide on subsequent actions when the robot stops to prevent unwanted collisions; and 3) re-shape the robot arm trajectories based on user instructions. We subjectively evaluate the robot’s behavior by comparing the safety perception of GPT-4 to the participants. We also evaluate the accuracy of natural language-based robot programming of decision-making requests. The results show that GPTAlly infers safety perception similarly to humans, and achieves an average of 80% of accuracy in decision-making, with few instances under 50%. Code available at: https://axtiop.github.io/GPTAlly

Powered ankle-foot prostheses offer the potential to emulate natural locomotion dynamics, thereby addressing the issues related to uneven gait and insufficient propulsion typically experienced by individuals with lower limb amputation wearing a passive prosthetic device. Despite significant progress, existing powered prostheses are often hindered by their substantial build height, bulky design, excessive weight, and noise level, limiting their widespread adoption. This work presents efficient and lightweight spring ankle, a lightweight (1.15 kg) and compact (11 cm high) powered ankle-foot prosthesis fitting within the volume of a shoe and capable of providing a net positive mechanical energy over the gait cycle. This level of integration is achieved through an innovative arrangement of a spring and actuator mechanisms operating in synergy. This hybrid architecture offers users the choice to walk actively, with propulsive energy assistance; regeneratively, potentially allowing for energy harvesting to recharge the device battery; or completely turned off (passive). This prototype has been validated during benchtop experiments and through trials involving four amputated participants. These tests encompassed various scenarios, including treadmill walking and everyday ambulation tasks. In addition, a sensitivity analysis was conducted to assess how different control parameters impacted the provided mechanical energy and resulting gait performance.

Gait symmetry between both legs is a typical hallmark of healthy walking. In contrast, several pathologies induce asymmetry in the gait pattern, regarding both spatial and temporal features. This can be due to either an asymmetrical change of the body morphology—e.g., after an amputation or an injury—or a damage in the brain—such as stroke or cerebral palsy. This deficit in gait symmetry usually induces higher metabolic effort in locomotion and might further accelerate severe comorbidities such as osteoarthritis and low back pain. Consequently, several assistive devices—such as active exoskeletons or prostheses—are currently developed to mitigate gait asymmetry and restore a healthier gait pattern. Typically, the development of such devices requires extensive tests and validations, and it is practically and ethically not always desirable to recruit disabled patients to run these tests in the preliminary stages of development. In this review paper, we collect and analyse the different reversible interventions described in the literature that can induce asymmetry in the gait pattern of healthy walkers. We perform a systematic literature research by exploring five databases, i.e., Pubmed, Embase, Web of Science, Google Scholar, and Scopus. This narrative review identifies more than 150 articles reporting 16 different interventional methods used to induce asymmetric gait pattern in healthy walkers or with the potential to do so. These interventions are categorized according to their mode of action, and their effects on spatiotemporal parameters, joint kinematics and kinetics are summarized adopting a macroscopic viewpoint. Interventions are compared in terms of efficacy, maturity of the results, and applicability. Recommendations are provided for guiding researchers in the field in using each of the identified manipulations in its most relevant research contexts.

In this document, we develop a method to dimension antagonistic artificial muscles to actuate a joint while modulating its stiffness through the biological principle of co-contraction. Our method relies on a linear model of the force-length characteristic of biological muscles. This model is then used to derive relationships between the activation levels of the artificial muscles and joint torque, equilibrium angle, and stiffness. In particular, we provide a graphical representation of these relationships that illustrates the principle of co-contraction and underlies our dimensioning method. Then, we illustrate our method with a numerical example by showing how to find the main characteristics of artificial muscles for a given operating point and stiffness range.

There is a metabolic cost associated with controlling walking balance but it remains unclear to what extent sagittal plane balance control contributes to this cost. Furthermore, sagittal plane balance control strategies vary with speed but it is unclear whether this also leads to a speed-dependent balance-related metabolic cost. Here, we explored the metabolic cost of stabilizing walking in the sagittal plane across speeds and its relationship with balance control strategies. To this aim, we applied continuous treadmill belt speed perturbations (standard deviation of 0.13 ms-1) to 22 healthy individuals walking at 0.8, 1.2, and 1.6 m/s. We evaluated changes in metabolic energy consumption and balance control strategies between perturbed and unperturbed walking and explored relationships between both. Perturbations induced larger increases in metabolic rate and changes in balance control strategies at slower than faster walking speeds, suggesting that walking is more robust against perturbations at faster speeds. Perturbations increased the metabolic rate by 16.7% at the slowest versus 4.6% at the fastest walking speed. When perturbed, subjects took shorter, wider, and more variable steps, and variability in ankle muscle activation increased but most changes were larger at slower speeds. Metabolic rate increased more due to perturbations in individuals who reduced step length more, i.e. relied more on anticipatory adjustments of the walking pattern. Our findings are especially relevant for explaining the increased metabolic cost of individuals with mobility impairments, who often walk slower and have altered balance control. ### Competing Interest Statement The authors have declared no competing interest.

Foundation models provide the adaptability needed in robotics but often require explicit tasks or human verification due to potential unreliability in their responses, complicating human-robot collaboration (HRC). To enhance the reliability of such task-planning systems, we propose 1) an adaptive task-planning system for HRC that reliably performs non-predefined tasks implicitly instructed through HRC, and 2) an integrated system combining multimodal large language model (LLM)-based task planning with multimodal communication of human intention to increase the HRC success rate and comfort. The proposed system integrates GPT-4V for adaptive task planning and comprehension evaluation during HRC with multimodal communication of human intention through speech and deictic gestures. Four pick-and-place tasks of gradually increasing difficulty were used in three experiments, each evaluating a key aspect of the proposed system: task planning, comprehension evaluation, and multimodal communication. The quantitative results show that the proposed system can interpret implicitly instructed tabletop pick-and-place tasks through HRC, providing the next object to pick and the correct position to place it, achieving a mean success rate of 0.80. Additionally, the system can evaluate its comprehension of three of the four tasks with an average precision of 0.87. The qualitative results show that multimodal communication not only significantly enhances the success rate but also the feelings of trust and control, willingness to use again, and sense of collaboration during HRC.

Constraining gait rhythm with a metronome has been shown to influence gait pattern in many different ways. Although rhythmic cues can improve several parameters in some clinical populations, they do alter the long-range autocorrelations naturally exhibited in series of stride durations. However, transitions between walking with and without a metronome (and vice versa) have not been measured; it is therefore unclear how people adapt to such a change in task. To address this gap, a total of 21 healthy volunteers were asked to walk overground under three conditions: one unconstrained control condition, followed by two conditions in which a metronome was activated during either the first or second half of the trial to test both transitions. The long-range autocorrelations were assessed over a sliding window on the stride series to measure their evolution. Our observations were reproduced with a computational model, allowing us to relate sudden changes in movement parameters to the long-range autocorrelations, which are typically measured over longer timescales. The results showed a clear transition in both conditions involving a metronome, with long-range autocorrelations of the series of stride durations gradually reduced when the metronome was turned on and recovered when it was turned off. In these two conditions, the change in long-range autocorrelations could be reproduced in the model by an instantaneous switching of the control policy associated with the presence or not of the metronome, suggesting that long-range autocorrelations emerge from a flexible control strategy that rapidly regulates timing and amplitude parameters according to task requirements.NEW & NOTEWORTHY Through an experiment involving transitions between walking with and without a metronome, we studied how people adapt to such a change of task by measuring the evolution of long-range autocorrelations (LRA) in the stride series. The results were reproduced in a model by an instantaneous change in the control policy, which validates the hypothesis that LRA emerge from a flexible control that rapidly regulates timing and amplitude parameters according to task requirements.

Hip disarticulation and hemipelvectomy are the most severe forms of lower limb amputation, posing significant challenges to prosthetic solutions in terms of size, biomechanical functionality, and user compatibility. While active ankle and knee prostheses have made spectacular progresses recently in restoring a physiological gait, these advancements did not percolate yet to hip prosthesis design. This article introduces an innovative design of an active hip prosthesis displaying remote center of motion, and range of motion and torque compatible with the most ubiquitous locomotion tasks, i.e., walking and stand-to-sit-to-stand transitions. The designed structure incorporates a tilted double parallelogram mechanism, in order to optimize compactness and minimize internal constraints. The proposed hip prosthesis design features minimal encumbrance, with a horizontal size of 140mm and a frontal width of 136 mm. Its range of motion spans from -30 degrees to 90 degrees, providing a comfortable sitting position with existing shell design. Remarkably, the mass of this hip joint is a mere 3.30 kg, excluding battery and power electronics.

This paper presents the design and the experimental characterization of a Cyber-Physical System tailored for research in fundamental and applied fluid mechanics (fluid-structure interaction problems), biomechanics (biolocomotion), and civil engineering (wind- or flow-structure interactions). The design is aimed at ideally controlling the six degrees of freedom of the manipulated object, being versatile to different experimental scenarios, and usable in real-time and closed-loop if manipulating an active object. Mechanical design robustness is examined through an experiment emphasizing the crucial constraint of robot rigidity. Subsequently, the robotic Cyber-Physical System kinematic and dynamic capabilities are validated, demonstrating compliance to specifications, along with considerations regarding acceleration saturation. A third experiment analyzes the robot end-effector natural frequencies, yielding frequency ranges that should not be excited in future experiments. Findings contribute to providing directions for refining the mechanical design, synthesizing control strategies, and enhancing the device robustness and performance in various flow-device interaction scenarios.

Over the last decades, active knee prostheses demonstrated their ability to support most daily-life locomotion tasks for transfemoral amputees. However, passive devices are still used by a majority of patients due to their simplicity of use, low cost, independence with respect to an external power source, and adaptability across various gait profiles in daily tasks. A key challenge in the development of passive prostheses is to have them supporting the largest possible amount of locomotion tasks in a strictly passive manner. Following a recent trend promoting layered prostheses with active as-needed assistance, this paper focuses on the development of a passive knee prosthesis enabling walking on level ground and providing passive assistance for sit-to-stand-to-sit transitions. The fundamental feature on our innovative concept consists in relying on four passive mechanisms in order to render multiple stiffnesses adapted to the selected task or its specific phases (stance/swing). This paper reports the conceptual design of the prosthesis, its mechanical dimensioning, and a first embodiment fitting into the volume of existing devices.

In the recent past, considerable research efforts have focused on the development of powered lower-limb prostheses. As these devices strive to emulate biological joint behavior, their embedded controller plays a crucial role in adapting to the currently performed task. The challenge when developing such an algorithm lies in achieving high efficiency, characterized by accuracy and reactivity, while depending on a minimal number of sensors, preferably embedded inside the device. This paper introduces a high-level controller, in charge of detecting the user's intention, specifically designed for the Efficient and Lightweight Spring Ankle (ELSA) prosthesis. A heuristic method was preferred over machine learning classifiers for its simplicity. Embedded IMU and position sensor within the device provide direct inputs to the algorithm, which operates following straightforward rules. The efficacy of the algorithm is demonstrated through the identification of seven distinct locomotion tasks. Besides the intention detection, we also suggest what would be the most appropriate ambulation mode of ELSA for each task, implemented with smooth transitions. The algorithm was tested through a preclinical validation, involving a single transtibial amputee. Out of the 6 trials, 5 were considered as being consistent and allowed to identify design improvements. Importantly, the user provided positive feedback about the intuitiveness of the algorithm.

Lower-limb exoskeletons are wearable devices whose main purposes are human rehabilitation and bilateral locomotion assistance. In particular, there is a growing interest for their use to symmetrize the gait of hemiparetic patients. This often consists in using the kinematics of the less affected side as a reference for the most affected one. In this work, we followed this approach to design a symmetrization algorithm using the formalism of motor primitives, i.e. a low-dimensional set of signals that provide the desired assistance through their combination. The amount of variables to be stored in memory is thus intrinsically limited, and this framework is particularly adapted to include other modes of assistance and/or transitions between locomotion tasks. In this paper, we report the preliminary validation of this newly developed algorithm with a hip exoskeleton and a single participant replicating hemiparetic walking. Results show that the algorithm effectively managed to reduce both temporal and spatial gait asymmetry.

In the recent past, the development of lower-limb prostheses has taken a new turn with the emergence of active systems. However, their intrinsic wearable nature induces strict requirements regarding weight and encumbrance. In order to reduce the load — and thus the bulkiness — of their active part, several prototypes leverage the concept of compliant actuation, consisting in including an elastic element in parallel and/or in series with the actuator. In this paper, we explore the usability of polymer compliant ropes placed in parallel with the actuator of an ankle prosthesis. Ropes are intrinsically light and compact, and thus offer several advantages as compared to more traditional coil or leaf springs. Polymer materials were selected for their high energy density and yield strength. We conducted a set of experimental tests with several ropes, pretension levels, and periodic loading profiles. Results show that polymer-based ropes have a high potential for ankle assistance devices, since they can store the required energy in a low volume. However, further research should be conducted to improve their efficiency, since we estimated that only about 50% of the stored energy can be released, with few variations as a function of the rope preconditioning and loading profile.

This study is the first that investigates the effects of a wearable active pelvis orthosis using an oscillator-based adaptive assistance on the level of long-range autocorrelations in series of stride durations during overground walking. It is also the first to compare the effects of different assistance settings on spatiotemporal gait metrics.

Background In the recent past, wearable devices have been used for gait rehabilitation in patients with Parkinson’s disease. The objective of this paper is to analyze the outcome of a wearable hip orthosis whose assistance adapts in real time to the patient’s gait kinematics via adaptive oscillators. In particular, this study focuses on a metric characterizing natural gait variability, i.e., the level of long-range autocorrelations (LRA) in series of stride durations. Methods Eight patients with Parkinson’s disease (Hoehn and Yahr stages 1 - 2.5) performed overground gait training three times per week for four consecutive weeks, assisted by a wearable hip orthosis. Gait was assessed based on performance metrics such as the hip range of motion, speed, stride length and duration, and the level of LRA in inter-stride time series assessed using the Adaptive Fractal Analysis. These metrics were measured before, directly after, and 1 month after training. Results After training, patients increased their hip range of motion, their gait speed and stride length, and decreased their stride duration. These improvements were maintained 1 month after training. Regarding long-range autocorrelations, the population’s behavior was standardized towards a metric closer to the one of healthy individuals after training, but with no retention after 1 month. Conclusion This study showed that an overground gait training with adaptive robotic assistance has the potential to improve key gait metrics that are typically affected by Parkinson’s disease and that lead to higher prevalence of fall. Trial registration : ClinicalTrials.gov Identifer NCT04314973. Registered on 11 April 2020.

Experimental characterization of bird flight without instrumenting the animal requires measuring the flow behind the bird in a wind tunnel. Models are used to link the measured velocities to the corresponding aerodynamic forces. Widely-used models can, however, prove inconsistent when evaluating the instantaneous lift. Yet, accurately estimating variations of lift is critical in order to reverse-engineer flapping flight. In this work, we revisit mathematical models of lift based on the conservation of momentum in a control volume around a bird. Using a numerical framework to represent a flapping bird wing and compute the flow around it, we mimic the conditions of a wind tunnel and produce realistic wakes, which we compare to experimental data. Providing ground truth measurements of the flow everywhere around the simulated bird, we assess the validity of several lift estimation techniques. We observe that the circulation-based component of the instantaneous lift can be retrieved from measurements of velocity in a single plane behind a bird, with a latency that is found to depend directly on the free-stream velocity. We further show that the lift contribution of the added-mass effect cannot be retrieved from such measurements and quantify the level of approximation due to ignoring this contribution in instantaneous lift estimation.