Limitations of the design and testing protocols
Since the exoskeleton was designed to provide a rigid support, individual digit movement was constrained to concentric and eccentric trajectories in a single plane of movement. Additionally, the digits were limited to 15 degrees of freedom in order to decrease weight and cost by reducing the number of active motors necessary. In the effort to reduce weight, complexity, and cost, adduction and abduction motions of the fingers were not assisted. The entire device, including the batteries and control system, weighed approximately 0.91 kg, a potentially significant weight to be carried on the arm of an individual with any amount of reduced arm strength; however, this weight is comparable to that of other similar devices that assist motion of 5 digits, and the weight falls within the typical 0.7–5.0 kg range of other hand exoskeletons [22, 27, 36, 37].
Both Bowden cable-based and pneumatic systems require motors or compressors to generate any concentric or eccentric forces. As for soft pneumatic actuators, the artificial muscles, such as the McKibben muscle, requires air transfer conduits and a compression system. As compared to the system in our described device, neither of these ideas save weight per-se. The pneumatic system is more complex in design, and the mass of the compressor(s) and valves are usually quite significant [15, 28, 38,39,40,41,42]. This type of system also moves a portion of the weight and volume from the forearm to a remote pack and/or the fingers themselves [15, 28, 29, 38, 41], which also reduces the ‘realistic’ feel or likelihood of providing the ‘structural support’ to the fingers that this device aims to provide. Using a rigid exoskeleton as opposed to a soft one also increases the likelihood of the unpowered device reducing the EMG produced for a certain force as shown by some of our subjects in Figs. 10 and 11. This movement of weight and volume to fingers, or devices that do not have a ‘realistic’ feel of independent finger movement can result in increased EMG for force provided, as described in [43]. By putting the heavier components directly onto the wrist as opposed to something like a backpack or holster, we can reduce movement restrictions and perceived movement restrictions due to hanging wires, as was present in an early version of this device [23], and other devices which use Bowden-cable systems [28, 29, 38, 41].
Along with increased weight which impeded the functional test of lifting a light weight, the joints of the device provided additional friction to digit movement. Although the control system compensated for the added friction on the concentric efforts, actuator activation only applied assistive forces during these concentric efforts. Therefore, the user was required to contribute to all eccentric, or digit extension, efforts without any motor assistance, although the non-adjustable rubber bands assisted in this movement. As stated earlier, if the user were to remove the tips of their fingers from the FSRs, the exoskeleton digits would automatically extend, but would not force the extensor movement should the user resist the rubber bands, even without engaging the motors.
In order to optimize the functioning of the FSRs and linear actuator function for each subject, the lengths of the polymer cables were adjusted to best fit the subject’s combined finger and palmer lengths in order to allow for an optimal contact of the fingertips and the FSRs, and therefore maximizing the user’s interface with the control system of the device. This required the investigator to disconnect, shorten, and re-attach 5 polymer cables on the device for each subject after checking the fit of the device, but prior to recording any data. Finally, the force produced solely by the user was not recorded, the forces reported were the combination of the user’s effort and the assistive force of the device. It was determined that the separate recording of user and device force would have required additional force sensors and wiring placed inside the device that would increase the weight and friction of the joints resulting from the additional wires.
Objective assessment of device performance
In previous studies, a grasping effort showed a statistically significant reduction in forearm muscle activation during grasping efforts for both the three-fingered [11] and five-fingered versions of this device [31]. However, in the previous pilot study with fewer participants using the five-fingered device, the 15 N pinching effort did not provide a statistically significant reduction in forearm muscle activation [31]. Now, in a sample of ten subjects with a slightly modified control scheme, while wearing the device the user needed a significantly reduced amount of forearm muscle activation for a grasping effort by 67% (Fig. 8a), a 15 N palmer pinching effort by 30% (Fig. 8d), and a 15 N pinching effort by 67% (Fig. 8b). These percent differences are comparable to reduction in EMG recorded while using a similar exoskeleton device, for either the hand or arm, whose intent is to augment a user’s force production [21, 44]. There are devices, however, that allow the user to produce minimal muscle effort, but these are device that intend to only minimally voluntary movement, where the user implies movement and the device moves semi-autonomously [15, 16].
Although there was no statistically significant increase in the force/EMG ratio in the 8 N pinching effort (Fig. 8c), the average pinching force/EMG ratio increased in both the trials where subjects wore the unpowered structure and the trials where subjects wore the powered device trials as compared to the barehanded trials (35% and 32% increase respectively). These percent differences in the 8 N pinching effort are comparable to Kadowaki, et al., where it was found that their soft exoskeleton device assisted with 20% of the pinching effort [45]. This is also comparable to the reduction in EMG produced in devices of similar structure and function, but for a different limb, for example by the major hip flexor, minor knee extensor muscle in a study investigating a powered hip exoskeleton device during walking [46]. This implies that wearing the device, powered or unpowered, provided enough support to the fingers during light pinching efforts to reduce muscle activation. It is also possible that there was reduction in activation in the other muscles of the arm and hand that were not investigated. For example, the exoskeleton also assisted in the movement of the thumb, especially in the pinching efforts, but the EMG of these muscles, such as the abductor pollicis, were not recorded. Additionally, in comparison to the HERO Grip Glove [14], this exoskeleton device itself, with no human interaction, produced 17.2 N of grip force, compared to their 12.7 N, but only 5.0 N of pinch force compared to their 11.0 N. This would appear to indicate that the thumb exoskeleton digit may be the limiting factor in the reduced pinching forces produced by this device. This is also supported by the increase in force produced when performing single/multi-finger grasps as opposed to finger-thumb pinches. These single-digit user-independent grasping forces, however, are comparable to other, similar powered assistive devices for the hand, although some with fewer digits than five, with forces ranging from 5 to 12 N per digit [20, 47, 48].
Overall EMG increased when lifting a water bottle while wearing the exoskeleton device, powered or unpowered, when compared to the trials in which the subjects were barehanded (Fig. 8f). However, the majority of subjects showed a decrease in EMG production when the device was powered on compared to when the device was unpowered, while a two saw an increase (Fig. 11).), indicating that powering the device still had a beneficial impact when grasping an object over wearing the unpowered device. This increase in EMG from bare-handed to wearing the device is likely attributed to the weight of the device, as the subjects also had to lift the device along with the water bottle in the trials in which they wore the exoskeleton device, and the weight of the water bottle was significantly smaller than the weight of the device. However, in order to reduce the weight of the device, the ability to move all five fingers independently would be lost, such as with Yoo et al. [17], where only one motor was used to control multiple fingers to decrease weight and cost. Based on some participant feedback, however, it may be possible to have one motor control both the pinky and the ring finger. Some subjects expressed that they did not feel their pinky finger contributed to their gripping capability, so having the pinky driven in parallel with ring finger motion may be a feasible method to remove some weight.
While there was no overall trend of improvement when lifting a tote bag off of the ground (Fig. 8e), most subjects showed reductions in EMG production when the device was powered as compared to their trials with the unpowered device or bare-handed, while two subjects saw an increase when comparing the powered device trials to the unpowered device trials (Fig. 10). This suggests that for some individuals, wearing the device while powered was beneficial in reducing EMG to lift certain objects, while others had a difficult time wearing or controlling the device for these purposes. For example, subject 8 had difficulty in using the powered exoskeleton for both of these tests as shown in Figs. 10 and 11.
Again, it is possible that there was reduction in activation of other muscles of the arm and hand were not investigated. In a study investigating an exoskeleton for the arm, the EMG of 16 upper limb muscles were recorded, and it was found that in different movement patterns, different muscles showed a decrease in EMG with the use of the device [49]. In the future, a more extensive array of EMG electrodes could be used to determine if different muscles of the hand and forearm showed reduction in activation during functional tests.
Assessment of potential functionality for ADLs
Many studies assessing the functionality of novel exoskeletons for assistance in ADLs and rehabilitation assess theoretical sensory feedback [30, 50], joint torques and grip forces in controlled motion of the device [18, 36, 51], and force/EMG measurement for controlled full-hand grasps [11, 31]. This type of assessment, however, does not necessarily correlate to a device’s usefulness in performing ADLs such as lifting various objects. Although the device appeared to increase the muscle activation needed to lift an object lighter than the device itself and showed no significant increase or decrease in the muscle activation needed to lift an object heavier than the device itself; this generalization was not true on a subject-by-subject basis.
The majority of subjects saw a reduction in EMG when lifting an object heavier than the device when the device was in the powered state compared to their trials with the unpowered device or bare-handed (Fig. 11). This is likely attributed to both the subject’s grip on the object and the fit of the device. If the device was fit poorly to the individual, the subject would be more likely to lose contact with some of the FSRs in the fingertips and therefore not be fully assisted in their grip. This would also cause the subject to adjust their hand’s position in the device during the test, increasing their muscle activation. This spike in muscle activation due to repositioning would also occur when the subject adjusted their grip on the tote bag. The subjects that were more adept at controlling the device and felt more comfortable wearing it were less likely to attempt to reposition either the device or the bag, resulting in a decrease in EMG when wearing the powered device. So, while the device is not well suited to assist subjects lifting an object lighter than the device, it was beneficial when assisting subjects who felt comfortable wearing the device in lifting an object heavier than the device itself.
This would imply that the intended users of the device, or subjects who require assistance in ADLs, would require a training regimen involving repetitively using the device to complete tasks that the device would be used for. This kind of training regimen is commonly used in studies evaluating a device’s usefulness in assisting stroke patients perform ADLs [5]. Post-task-oriented training, subjects with impairments due to stroke have shown improvement in their hand functions [5], so this type of training would be beneficial for subjects who might feel uncomfortable using the device initially. In the future, subjects should undergo a task-oriented training regimen after initial grasping efforts, then their ability to perform functional tasks should be re-assessed to account for initial comfortability in using the device. Additionally, in future studies with this device, subjects with upper limb impairments should be recruited to investigate how effectively the device assists in their realistic ADLs.
Value added
This device is a low-cost rigid exoskeleton that allows for individual finger movement and significantly reduces forearm EMG. Many force amplifying exoskeletons fit one or two of these categories, but the fact that our device fits all three is what adds value to the field. As stated previously, while wearing the device, the user required a significantly reduced amount of forearm muscle activation for a grasping effort by 67% (Fig. 8a), a 15 N palmer pinching effort by 30% (Fig. 8d), and a 15 N pinching effort by 67% (Fig. 8b), which is a comparable reduction while using a similar exoskeleton device whose intent is to augment a user’s force production [21, 44]. Although there was no statistically significant increase in the force/EMG ratio in the 8 N pinching effort (Fig. 8c), the average pinching force/EMG ratio increased in both the trials where subjects wore the unpowered structure and where subjects wore the powered device trials as compared to the barehanded trials (35% and 32% increase respectively, similar to a comparable soft exoskeleton device [45]). Even though exoskeletons with 1 or 2 fingers are simpler to implement, most ADLs require at least 3 fingers to be assisted by an exoskeleton [26]. Exoskeletons with 3 or 4 fingers could assist with most ADLs, the realism for the user would decrease as the number of fingers decrease [26]. Additionally, having fewer fingers limits the motions the user can make with their hand, as well as limit the objects the user can lift. For example, a 4 or 3-fingered exoskeleton could assist a user in picking up objects of uniform shapes (e.g., cup, reusable water bottle) but not objects that are oddly shapes or have varying thicknesses (e.g., cell phone, wine glass). Increasing the number of independently controlled exoskeleton digits would allow for the control to lift objects such as this and increase the mobility to the point where it feels natural to the user.
Using only commercially available components is only one part of making the exoskeleton cost-effective by removing the need for custom components. While this would make up some of the cost, what makes this device cost-effective is the manufacturing method. Previous designs from our laboratory were machined from aluminum, which not only resulted in a heavier exoskeleton, but also required greater time, cost, and experience to manufacture. This fabrication method also limited part and assembly design due to the inherent nature of subtractive manufacturing. By switching to thermoplastic, the device was lightened significantly, but the manufacturing method then also had to be reconsidered. Subtractive manufacturing of plastic would have largely the same problems as subtractive manufacturing of metals, and forming/molding would require an excess of 20 different molds (as each part in our device is unique to each finger) and would limit the design and assembly of the parts, as interconnected pieces cannot be formed/molded. Therefore, a main design consideration that makes the exoskeleton cost effective is the move to additive manufacture, specifically fused deposition modeling (FDM). Not only is material cost low, labor cost and time is also much lower. By using additive manufacturing, the need for expert machining, design and assembly restrictions, and the cost of developing molds was removed. This design fits into the niche of where additive manufacturing makes production cost effective, with low production volume and high complexity.