Adding backlash to the connection elements can improve the performance of a robotic exoskeleton

Abstract

Kinematic mismatch between exoskeletons and human body results in excess internal forces/torques and hence discomfort as well as increase the power consumption. The connection stiffness has been shown to have potentials for minimizing the kinematic mismatch effects. However, realization of a desired stiffness in the connection element seems difficult if not impractical.

In this work, adding controlled backlash to the exoskeleton-body connection is investigated as a possible solution for the kinematic mismatch challenge. A stiffness model which includes backlash parameters was formulated and identified experimentally using three male subjects on a typical lower extremity exoskeleton. A performance index which includes tracking error and user's comfort was calculated on a normal gait cycle. The impact of the backlash parameters including: type, direction and range was analyzed in detail.

The impact of adding a tuned backlash to the shank is significant while it is negligible for the thigh connection. The highest performance improvement was obtained by cylindrical backlash added to the shank axial direction (97%) with a range of 15^∘ and 13 mm. The proposed method and results are believed to be directly applicable to the optimal design of robotic exoskeletons.

Introduction

Numerous exoskeletons have been developed for lower [1], [2], [3] or upper [4] limbs to enhance healthy users’ strength and endurance [5], [6], [7] or assist/rehabilitate patients with neuromuscular defects or movement disorders [8], [9], [10]. All exoskeletons, especially rehabilitative ones, should comfortably interact with the human body and precisely move the limbs along the desired trajectories. However, the comfort and precise tracking are not completely achievable in practice due to the differences between the kinematics of the exoskeleton and the human body, especially for the lower extremities. The kinematic mismatch results in an over constraint mechanism when the exoskeleton and the body are connected and supposed to move together. As a result, depending on the flexibility of the connection elements, it may cause large internal forces/torques (i.e. discomfort or even injuries for the user) or large tracking error (between the exoskeleton and the body) [11]. The kinematic mismatch has different sources such as non-identical human and exoskeleton joints [12,13], dimensional differences between the robot and the wearer [14], error in the human joints rotation axis determination [15,16], or incorrect wearing of the exoskeleton [17,18].

The mainstream approach of handling the kinematic mismatch is based on increasing the exoskeleton compatibility with the human body by modifying the conventional mechanisms of the exoskeletons. Self-aligning mechanisms employ extra active or passive joints to adapt the kinematics of the exoskeleton to the human body [19,20]. Although exoskeletons with more degrees of freedom (DOFs), resulting from adaptive joints, have less kinematic mismatch, the interaction forces between the exoskeleton and the body may rise due to the extra weight and inertia of the added joints [17]. However, instead of extra conventional joints, compliant joints, designed for maximum compatibility to human body joints (e.g. the knee joint [21], [22], [23], [24], [25]), can be another solution for the kinematic mismatches [26].

Despite the effectiveness of the above mentioned methods, the kinematic mismatch is usually present to some extent. Two kinematic chains, the exoskeleton and human body, with some kinematic incompatibilities would be theoretically locked with infinitely large interaction forces/torques at the connections if the exoskeleton was rigidly connected to the body [27]. The elastic deformability of the connections (e.g. the connecting cuffs, skin, soft tissues, etc.) allows relative motions between the robot and the body avoiding excess interacting forces. In other words, the elastic stiffness of the connection is an important parameter that determines the exoskeleton performance measures such as the user discomfort and tracking error [28], [29], [30]. In a previous work [28], the authors provided an analysis of how the stiffness of the shank and thigh connections impact the discomfort and tracking error for an active exoskeleton. It was shown that better performance can be achieved when the stiffness tensors of the connections can be fully designed. For instance, lower rotational stiffness of the shank connection along the shank and normal to the sagittal plane improves both comfort and tracking performance of the robot [26].

Although optimal design of the stiffness of the connections improves the mechanical performance and comfort of the exoskeletons, it seems difficult for practical reasons. The stiffness tensor of the connection element has six independent stiffness components which need to be realized in a single element. The geometrical constraints for such element are also very limiting since the compactness is a critical requirement for an exoskeleton.

To put this knowledge into practice, instead of manipulating the stiffness tensor of the connection, one may add limited backlashes in certain directions by adding simple kinematic or even flexible joints to the conventional connections. The implementation of a limited backlash in a certain direction is more practical than designing the stiffness. Also, the idea was originated by observing that [26] the performance improvement is mostly achieved by lowering the stiffness in certain directions which in extreme case implies backlash.

The objective of this research is to explore the effects of the backlash parameters (e.g. the axis of the backlash and its range) on the user comfort and the tracking performance, for a lower extremity exoskeleton executing normal gait cycle. For this purpose, a model was developed for the interaction of the powered exoskeleton and the lower extremities. Experimental data were used to identify the model including the geometric stiffness matrices for three human subjects at the thigh and shank connections. The exoskeleton efficiency indices were then calculated for different backlash parameters and types based on the kinematic model simulating the limbs and exoskeleton motions.