**4. Discussion**

The proposed new exoskeleton design has been developed, starting from a thorough analysis of existing exoskeletons. The device is aimed at solving the mechanical problem in a way that adapts to the geometrical and behavioral parameters of the biological hand. The analysis of a variety of designs based on a rigid mechanical structure present in this field of research [1,4,5,20,27,28,45], showed some remarkably distinct types of construction. The solution developed by Chiri et al. [6], considering the direct matching of the finger joint centers, is a good one, due to its compliance with the asymmetric behavior of the human hand's biomechanics. A limitation for this design appears at the MCP joint, where direct matching of the joint cannot be implemented due to the hand anatomy. In most designs of hand exoskeletons, the DIP and PIP joints are easy to actuate via an exoskeleton mechanism, while the MCP joint presents a bigger challenge to actuate properly. For the MCP joint, there are several alternative solutions, based on more complex mechanisms for generating actuated or underactuated movement. Another type of construction, developed by Shields et al. [35], implements a more complex structure based on a mechanism with linkages for remote centers of rotation. Although it seems like a bulky design, considering the large and complex mechanism, it has a huge advantage considering that it saves a lot of space between the fingers, an important design factor in developing a compliant orthotic shell for the exoskeleton fingers. Comparing the design proposed in this work to the actual state of the art in the field, a series of advantages and disadvantages can be identified. In the first place, it has to be mentioned that the device presented in the paper benefits from the advantages offered by an underactuated mechanism, as a result of a low number of actuators. Another advantage of the proposed device is due to the fact that the actuators are not placed on the fingers' orthotic shell. Therefore, this device has the ability to present easily changeable components based on the person's anthropological measurements. Another advantage resides in a more compact design for DIP and PIP joints, compared to other underactuated models (as well as some fully-actuated models) such as those

of Shields et al. [35], Wege et al. [36], and Zhang et al. [55]. One can mention, among advantages, that the proposed design offers a specified center of rotation for the MPC joint, in contrast to the solutions proposed by Chiri et al. [6]. It can be said that the current design can offer a natural movement of the wearer's hand, compared to other fully-actuated exoskeleton models, such as those presented in Zhang et al. [55], where each joint is independently actuated, thus involving a highly complicated control algorithm, due to the fact that the exoskeleton does not adapt very well to different anthropological dimensions and behaviors. On the other hand, one can mention as disadvantages the lack of certainty for each of the joints' positions, which obviously is a negative characteristic. Even though numerous advantages of the designed device are evident, a notable disadvantage is still present in this solution: the mechanism with the remote center of rotation implemented on the MPC joint is still a relatively large one. Further research has to be undertaken to optimize the device's dimensions. It was noted that the new exoskeleton adapts itself to the wearer's fingers, thus producing a compliant actuation, because of the underactuated mechanism implemented in the exoskeleton design. A notable difference in testing the exoskeleton with and without the wearer's hand is that while the device is being used with the wearer's hand, it produces trajectories that tend to be symmetrical between flexion and extension cycles. In contrast, in testing without the hand, the exoskeleton tends to produce asymmetrical flexion/extension cycles.
