earable exoskeletons might reduce human effort during walking. However, many of the current exoskeletons rely on heavy actuators and/or external power supplies; this has a negative impact on their efficiency and operation range. As an alternative, (quasi)passive exoskeletons have been developed. One of the proposed passive exoskeleton concepts is the exotendon concept of van den Bogert [1]. In this concept, long elastic cables span multiple joints. The cables can temporarily store and transfer energy between joints. In simulation, the average absolute joint torque can be reduced by 71%. The simulations are based on the hypotheses 1) that the exoskeleton does not influence the joint angles and 2) the total joint torques and a reduction in the human joint torques results in a reduction in the metabolic cost of walking. The goal of this article is to experimentally evaluate the exotendon concept and test the hypotheses underlying it. We implemented the exotendon concept in a lightweight exoskeleton. Experimental results show that the exotendons indeed reduced the average absolute joint torques. However, the exotendons also influenced the joint kinematics, and the metabolic cost of walking did not decrease. Therefore, the underlying assumptions of the exotendon concept are invalid. We also found that, in practice, the amount of support given by the exotendons is limited to about 35% of the theoretical optimal support. For higher levels of support, the motion is hindered and the support is experienced as uncomfortable by the users of the exoskeleton. HypothesisIt has recently been shown that walking with an exoskeleton can reduce the metabolic cost of walking [2]. For exoskeletons to become useful in daily life, their power consumption is a key factor. The required power is often provided by batteries, which limits the operating time of the exoskeletons. For the HAL exoskeleton (Cyberdyne, Tsukuba, Japan) and Ekso (Ekso Bionics, Richmond, California), the operating time is approximately 3 h [3], [4].The high power requirements of exoskeletons contrast with the efficient locomotion found in nature. Human and animal legs possess mechanisms that save energy while in motion. Elastic tendons can temporarily store energy, and multiarticular tendons and muscles can transfer energy between joints [5]- [7].The model optimizations of van den Bogert suggest that human joint torque and power can be reduced by placing elastic structures, called exotendons, parallel to the leg [1]. These exotendons have a similar function to biological uniand multiarticular tendons. Simulations suggest that the human joint torque, the torque provided by the leg muscles, can be reduced by 21% with uniarticular exotendons at the ankle. This reduction increases to 46% if triarticular exotendons are used that span the hip, knee, and ankle. For more complex configurations with multiple exotendons per leg, the predicted reduction increases to 71%. The hypotheses underlying the exotendon concept are: 1) the exotendons do not influence the joint angles and...

This paper evaluates the Achilles exoskeleton. The exoskeleton is intended to provide push-off assistance for healthy subjects during walking. The assistance is provided by a series elastic actuator that has been optimized to provide maximal push-off power. The paper presents the control method of the exoskeleton and the evaluation of the exoskeleton.

We developed a passive exoskeleton that was designed to minimize joint work during walking. The exoskeleton makes use of passive structures, called artificial tendons, acting in parallel with the leg. Artificial tendons are elastic elements that are able to store and redistribute energy over the human leg joints. The elastic characteristics of the tendons have been optimized to minimize the mechanical work of the human leg joints. In simulation the maximal reduction was 40 percent. The performance of the exoskeleton was evaluated in an experiment in which nine subjects participated. Energy expenditure and muscle activation were measured during three conditions: Normal walking, walking with the exoskeleton without artificial tendons, and walking with the exoskeleton with the artificial tendons. Normal walking was the most energy efficient. While walking with the exoskeleton, the artificial tendons only resulted in a negligibly small decrease in energy expenditure.

To promote active participation of neurological patients during robotic gait training, controllers, such as "assist as needed" or "cooperative control", are suggested. Apart from providing support, these controllers also require that the robot should be capable of resembling natural, unsupported, walking. This means that they should have a transparent mode, where the interaction forces between the human and the robot are minimal. Traditional feedback-control algorithms do not exploit the cyclic nature of walking to improve the transparency of the robot. The purpose of this study was to improve the transparent mode of robotic devices, by developing two controllers that use the rhythmic behavior of gait. Both controllers use adaptive frequency oscillators and kernel-based non-linear filters. Kernelbased non-linear filters can be used to estimate signals and their time derivatives, as a function of the gait phase. The first controller learns the motor angle, associated with a certain joint angle pattern, and acts as a feed-forward controller to improve the torque tracking (including the zero-torque mode). The second controller learns the state of the mechanical system and compensates for the dynamical effects (e.g. the acceleration of robot masses). Both controllers have been tested separately and in combination on a small subject population. Using the feedforward controller resulted in an improved torque tracking of at least 52 percent at the hip joint, and 61 percent at the knee joint. When both controllers were active simultaneously, the interaction power between the robot and the human leg was reduced by at least 40 percent at the thigh, and 43 percent at the shank. These results indicate that: if a robotic task is cyclic, the torque tracking and transparency can be improved by exploiting the predictions of adaptive frequency oscillator and kernel-based nonlinear filters.

This paper presents the Achilles exoskeleton, an autonomous ankle exoskeleton that can generate 52% of the positive plantarflexion power around the ankle of a 80 kg individual with only 1.5 kg of mass added around the ankle joint. The mass of the exoskeleton is lower and the power density is higher than that of existing autonomous exoskeletons. This high power density was achieved by designing a series elastic actuator that consists of an electric motor and ball-screw gear with a carbon fiber reinforced leaf-spring as lever-arm. A dynamic model that includes the motor and gear properties, spring stiffness, and exoskeleton geometry was used to optimize the design parameters for positive power injection. Doing this for multiple combinations of preselected motors and gears and comparing their support to weight ratio, revealed the best drive combination. The performance of the realized exoskeleton was assessed in several tests. The actuator can track the optimized actuator stroke trajectory with a following error that has a RMS of 2.3 mm, it can track force reference signals with amplitudes of 1 N to 100 N with a bandwidth between 8.1 Hz and 20.6 Hz, and it outputs a maximum mechanical power of 80.2 W. These results show that the device is suitable for fulfilling its purpose: reducing the metabolic cost of walking with an autonomous device.