The development of the design and construction of the Touch Hand 4.5, with respect to the mechanical structure, electronics and control and software is detailed in the following section. The Touch Hand 4.5 has been developed with a new design based on the design concepts which were implemented in the Touch Hand 4.

The Touch Hand 4 was constructed and developed as a prototype to be used in the Cybathlon 2020 event, yet optimisation was needed. The Touch Hand 4 prototype, which is the basis for optimisation described in this paper, is shown in Fig. 1.

The Touch Hand 4 prototype

The Touch Hand 4.5 features a redesign of the existing mechanical structure of the Touch Hand 4 [13]. With numerous tests of the Touch Hand 4 conducted by different amputees, the feedback allowed for the optimisation and improvements, specifically focussing on the functionality of the device. The development of the hand and socket structure of the Touch Hand 4.5 are detailed in the following section, which also took into account the ability for performing the tasks required by the Cybathlon 2020 event, to perform day-to-day activities. Therefore, the pinch grip, full grip, half grip, and a thumb grip were important to achieve. Even though a weight of 2kgs was being pursued to be picked up, it has been experienced that actuators rated for these forces outputted less than the desired value, so designs were considered for a weight of 5 kgs to be picked up. The size of the hand had to be minimalized, yet contain the actuators, and the weight had to be reduced as much as possible during the design. Aesthetics was important, yet users have indicated in the feedback that functionality is of higher priority.

The challenge detected with the hand structure was the issue that existed between the digits and the knuckle structure. The CAD model representing the index finger before and after the redesign are represented in Fig. 2.

3D CAD model representing the index finger before and after the redesign

The 'index finger' was made thinner, sleeker and pointier, which allowed the finger to be less bulky and ensured a finer grip when completely closed, as well as more compact relative positioning when the ‘fist’ is closed.

he secondary 'ring finger' complex was made to look aesthetically better by becoming two joined fingers and also made this component less bulky, whilst still being actuated with a single linear actuator. The angles between the fingers also had to be readjusted to accommodate proper fitment on closing. The redesigned ring finger complex can be seen in Fig. 3.

3D CAD model representing the ring finger complex before and after the redesign

Following the completion of the design of all the finger complexes, the palm structure was developed. The knuckles on the palm were rotated 45º downward to allow the fingers to close completely into a bicycle grip, and the orientation of the linear actuators were changed to sort out some of the play issues which were initially experienced. The comparative palm structure can be seen in Fig. 4.

3D CAD model representing the palm structure before and after the redesign

he main initial problem with the functionality of the initial hand was the significant play of the 'fingers' on basic operation. The orientation of the linear actuators was changed to facilitate a smoother arc of movement and eliminate any excessive side thrust on the actuator shaft, by looking at various concepts of accommodating this motion.

Further sculpting of the palm surface profile was carried out in order to produce a more effective gripping profile for the tasks that the hand would have to perform. Figure 5, shows the intermediate assembly of all the components, excluding the thumb component. The thumb orientation angle was changed and the thumb component itself had to be redesigned.

The intermediate assembly of the components, excluding the thumb

The palm structure included the reorientation of the linear actuators to eliminate the play issues experienced.

The updated actuator angle was decided using a force analysis of the entire actuating mechanism. Actuonix PQ12-6-100 actuators were chosen as they delivered a 50N output force, which was sufficient for moving the finger mechanisms and provided enough force to hold day-to-day objects. The resultant force was obtained by means of Eq. (1).

$$\mathit\left(28^\circ \right)=\frac$$

(1)

$$Fy=50\times \mathit\left(28\right)=26.5 N$$

$$Fresult=\sqrt^+F^}=56.588 N$$

where: Fy = Side thrust on actuator arm [N] Fresult = Resultant force in the actuating link [N].

The original Touch Hand 4 design had the actuators horizontal to the palm of the hand, with the maximum side thrust being experienced when the actuator was fully extended and the fingers closed. This play caused excessive wear on the actuators and caused them to fail after some use. The mechanism was changed so that the fingers would be closed when the actuator was fully retracted and thus the side force would be experienced in this position rather than in the fully extended position, reducing the wear on the actuator arm, as seen in Fig. 6.

Two images on left shows the original actuation position, two images on right shows the improved actuation position

An Acrylonitrile butadiene styrene (ABS) 3D printed assembly, which can be seen in Fig. 7, was carried out to test the functionality of the updated design. The printing of the Touch Hand 4.5 was done with a Selective Laser Sintering (SLS) printer.

Positioning of linear actuator on the palm structure

Further sculpting of the palm was carried out to give the palm greater functionality in gripping objects, as well as giving the hand an aesthetically more pleasing and organic look. A further 3D print was carried out on the completed hand assembly using new actuator positions, knuckle links and pins. The final weight of the Touch Hand 4.5 was 4.7 kg.

The outcome was that the play issue on the actuator was solved, however a new problem was then identified. The allowing of the actuator to pivot on one end introduced too many movable points and no ground which caused the finger to move out of position, i.e., a new type of play has been inadvertently introduced. This play was solved by removing the knuckle link, so that the actuator connected directly to the finger using a single link, correlating to a simplified bell crank system. Due to the removal of the knuckle link, the finger was not able to move to a fully crunched position due to issues with the mechanical advantage and pivot points of the link attaching actuator to finger. The positioning of the linear actuator can be seen connected to the updated palm structure in Fig. 8.

3D printed model showing the positioning of the linear actuator

he hard plastic fingers of the 3D printed hand did not provide enough grip to allow the fingers to pick up smooth objects, such as a plastic cup, as a result, silicone slip-on covers were used on the fingertips of the hand. The silicone used was a 10-shore hardness silicone, which is very soft and has a skin-like texture. The silicone covers for the thumb, index and ring finger can be seen in the Fig. 9.

Silicone covers for the thumb, index and ring finger respectively

The silicone covers were manufactured using a casting process with split 3D printed moulds. The moulds were manufactured by creating a negative imprint of each finger in a block. Channels for ventilation and for pouring in the silicone were added to the mould. The block was then split in half to allow for the removal of the sleeve once it had cured.

The wrist-snap-fit-clip was used to connect the wrist to the socket of the Touch Hand 4.5. The two tabs located on each side of the component are pressed and fitted onto the female components on the socket. There is an additional locating tab, which is used to ensure that the socket orientation is correct. Figure 10 shows the connection of the snap fit clip to the socket collar.

The snap fit clip clipped into the socket collar

The SLS 3D-Printed socket with ABS material was inspired by similar latticed brace designs. An integrated latticed socket, which was created to fit into the designed collar, was able to provide an improved level of breathability for the lower forearm, further light weighting and an aesthetically pleasing design. The breathability and light weight of the socket allows for comfort, and decreases irritation. A decrease in irritation and breathability allows for a decrease in EMG noise.

Due to limitations experienced with the parametrically-based CAD platforms, a cutting-edge new implicit software platform, called nTopology (New York, USA), was used to carry out the socket design.

In order to produce the 3D printed socket, which can be seen in Fig. 13, the pilot's arm was 3D scanned and an STL file generated. The STL file was imported into nTopology where it was possible to optimize the structure. The import allowed for a CAD model to be generated and shelled to create the socket body. The lattice was then defined on an implicit model and lastly various Boolean functions were utilised in order to achieve the final design. The model was therefore converted to an STL file to allow it to be manufactured.

Once the socket structure had been manufactured, the latticed socket was tried on by the pilot, as seen in Fig. 11.

Construction of the 3D printed socket structure

The iteration included updated collars with integrated fastening devices and the lattice structure, which was thinned and cleaned up considerably. The final weight of the optimised lattice socket was 930 g.

Once all the elements of the hand, wrist and socket structure had been manufactured and assembled, the testing of the Touch Hand 4.5 was conducted.

Figure 12, represents the block diagram of the electronics for the Touch Hand 4.5 prosthetic device:

Block Diagram representing the electronics system of the Touch Hand

As seen in the block diagram in Fig. 12, muscle contractions are picked up by the BITalino EMG electrodes (from PLUX – Wireless Biosignals, Lisbon, Portugal) on the arm. The specific muscles that were used to identify the signals, were the forearm muscles, namely the flexor digitorum superficialis and flexor digitorum profundus muscles. These electrodes generate a signal, which is sent to the Arduino Nano microprocessor (from Somerville, Maine, USA). If the strength of the received signal is above a set threshold value, the microprocessor sends a signal to the Actuonix PQ12-30–12-P miniature electric linear motors (from Actuonix Motion Devices, Victoria, Canada), controlling the fingers, to open or close the hand.

This threshold was implemented to prevent accidental opening or closing of the hand due to involuntary or minor muscle contractions and elaborated in the software design section. The signal sent by the microprocessor to the actuators can be intercepted by switches connected to the ring and index fingers. These switches can be turned on and off irrespective of each other to allow different grips. Once the linear actuator receives the instructions, the thumb and finger complexes actuate accordingly, as per the EMG signals received from the flexor digitorum superficialis and flexor digitorum profundus muscles. The readings of the FSR400 force sensors (from Interlink Electronics, Los Angeles, California, USA), which are below the finger silicone covers, are processed and sent back to the motor control system for interpretation. If the set threshold for the force sensor is exceeded at any time during the operation of the hand, the signal to the motors will be interrupted and the fingers will stop actuating.

There are three switches located on the control box, which is worn on the upper arm of the pilot. The switch, which is located in the middle of the switch pad is an emergency stop or power switch. This switch is required to be switched on before the hand is able to operate.

The remaining two switches are connected to the power wires that actuate the fingers. The one is connected to the index finger and the other to the ring finger. By switching these switches on or off, the user can control which fingers open and close independently. When one or both of the switches are switched off, the signal does not reach the actuator and therefore the chosen finger will not respond to any signal from the Arduino. The finger will only respond to a signal if the respective switch is turned on. The thumb does not have a controlling switch and actuates when any signal from the Arduino is received, as long as the power switch is turned on.

The code flow diagram represents an analysis of the logic behind the data flow of the control system program. There are many different subsystems integrated into the control system of the Touch Hand 4.5. The control system features the microprocessor, the force sensors, and linear actuators. Signals are picked up by the EMG electrodes and processed in each of these subsystems before being interpreted and transferred to the relevant subsystem. The logic flow chart for the entire Touch Hand 4.5 control system is presented in Fig. 13. The sensors are continuously monitored for different threshold values, for the actuators to respond accordingly, depending on the grip that has to be performed. The force sensors and EMG electrode thresholds were identified by a training process with the amputee, to identify the optimal values to be considered depending on the grip they need to perform. The training process is crucial, as different muscle structures from different amputees will differ, therefore different levels of signals will be detected.

Code flow diagram for the control system of the Touch Hand 4.5