1 Introduction

Drastic development of medical devices has contributed to the advancement of modern medical sciences. In addition, modern electrical and mechanical engineering has been widely utilized in various medical devices, increasing their performance. The robotic surgical system is among the most advanced medical equipment using integration of the various engineering technologies (Lanfranco et al. 2004; Corcione et al. 2005; Muller-Stich et al. 2007; Schreuder and Verheijen 2009).

The most widely used robotic surgical system is the daVinci surgical system, which was first introduced to market in 1999 by Intuitive Surgical, Inc. The system performs effective laparoscopy with bending motion, rotating motion, and three-dimensional (3D) endoscopic viewing inside the human body. The surgeon performs the surgery by controlling surgical devices such as forceps, needles, and retractors that are attached to the robot arms. During the surgery, the surgeon stays in a separate space from the patient, watching the real-time image through the endoscope.

The advantages of robotic surgery include precise control beyond the abilities of human hands, minimized incision, and quick recovery due to minimally invasive techniques. Despite these merits, the physical isolation between the surgeon and the patient is an issue to be resolved for stable surgical operation. Physical isolation makes a surgeon insensitive to the force that surgery robots apply to a patient because only visual information is available. This implies that a surgery may have different results from the surgeon’s use of experience and skill. If the surgeon applies excessive force during surgery, unnecessary bleeding or improper damage can occur (Sung and Gill 2001; Ballantyne 2002; Hashizume et al. 2002; Kornprat et al. 2006; Morino et al. 2006).

A device displaying tactile information reportedly solves this problem because haptic stimuli have been found to be more effective than visual data (Dargahi and Najarian 2003; Dargahi et al. 2007; Reiley et al. 2008) in providing tactile information (Bethea et al. 2004; Lamata et al. 2006). Several haptic devices that stimulate the tactile sense have been reported (Kim et al. 2004; Ottermo et al. 2008; Moradi Dalvand et al. 2014b), showing that realistic haptic feedback helps an operator perform with better skills (Tholey et al. 2005; Wottawa et al. 2013; Moradi Dalvand et al. 2014a).

King et al. reported tactile displays for robot-assisted surgery (Culjat et al. 2008; King et al. 2008). The device was composed of a pneumatic tactile display that included six inflating balloons controlled using feedback signals from force sensors that measured the grasping force applied to an organ. They demonstrated that performance could be improved by using the proposed system in an in vivo test with surgeons dissecting animal subjects. One drawback of the haptic display developed by King et al. is that the device can only deliver normal force. It is known that both normal and tangential forces to the human finger are also important for humans to feel real haptic stimulation (Bark et al. 2013).

In our previous work, we reported a tactile display that could express both normal and tangential forces by using pneumatic pressure (Doh et al. 2013; Lim et al. 2014). The tangential display was obtained by actuation of a solid structure tangential to a finger surface. Because a pneumatic pressure regulator passively controlled the tangential actuation, the device suffered from hysteresis resulting in poor position accuracy. In order to solve this problem, we demonstrated a tactile display that was controlled by a position feedback signal. Though this device showed that it was possible to solve the hysteresis problem, it suffered from unstable position sensing because the position sensor was affected by normal pressure from the fingertip (Yoo et al. 2014). Therefore, in this paper we propose a three-axis tactile display with integrated capacitive position sensors that are formed on sidewalls of moving components to obtain accurate tangential movement; they have active feedback control, and circumvent the influence of normal pressure.

2 Design

Figure 1 shows a schematic view and an exploded diagram of the proposed three-axis tactile display. The device is composed of a core body, inner frame, and outer frame. The core body in the middle has nine circular balloons on its top surface, and a pair of elliptical balloons on its side surfaces. Each balloon is made of a 600 μm thick latex membrane covering the holes or pressure chambers, which are connected to silicone tubes for the application of pneumatic pressure. When air pressure is applied to the silicone tubes, the circular balloons on the top surface inflate, and a fingertip placed on top of the device feels normal haptic stimulation. As shown in Fig. 1b, the core body is guided by the inner frame, and moves only in the y-axis direction. Groove structures are formed on the inner surface of the inner frame to guide the rail structure of the core body. A pair of flexible printed circuit board (FPCB) sheets with patterned electrodes is attached to each wall of the core body and inner frame to form a capacitive sensor that measures the position of the core body in the y-direction. The lateral movement of the core body in the y-direction is actuated by the inflation of the balloons on the core body’s sidewall as shown in Fig. 2. Similarly, the inflation of balloons on the inner frame’s sidewall actuates the inner frame in the x-direction, which is monitored by the capacitive position sensor formed on each wall of the inner and outer frames. The outer frame is designed to enclose the inner frame, and restricts its movement in the y-direction. Tangential movement of the device is performed by two pairs of balloons at the sidewalls of the core body and the inner frame as shown in Fig. 2. When external pneumatic pressure is applied to the device, air inflates the balloons, which pushes the structure laterally. The lateral x–y position of the core body is determined by the difference in pressure on the balloons in opposition to each other.

Fig. 1
figure 1

Overall structure of the proposed 3-axis haptic device. a Assembled and b exploded view

Full size image
Fig. 2
figure 2

a Top view of device, b tangential actuation by pneumatic pressure difference applied on each side balloon

Full size image

Figure 3 shows the schematic of the capacitive displacement sensor integrated into the tactile device. The sensor is composed of two FPCB layers: one has a rectangular electrode; and the other has two ramp-patterned electrodes. Two parallel-plate capacitors, C 1 and C 2 , are formed between a rectangular electrode and each of ramp-patterned electrodes. The capacitance of each capacitor varies when the ramp-patterned electrodes move in a lateral direction as the core body or inner frame do by lateral actuation. The ramp-patterned electrodes have been used to obtain a linear output value according to the electrode’s position (Baxter 1997). To improve the sensitivity of the sensor, the slope of the ramp pattern should be increased. Further, the initial capacitance must be large enough to minimize the influence of the parasitic capacitance. Considering these design parameters and the restricted area for the sensor unit, we determined the dimensions of electrodes as shown in Table 1.

Fig. 3
figure 3

Schematic of the capacitive displacement sensor: a layout of rectangular shape and ramp-patterned electrodes, b its cross-sectional view at the A–A’ line

Full size image
Table 1 Dimensions for capacitive position sensor
Full size table

3 Implementation

3.1 Device

Figure 4a shows a photograph of a fabricated three-axis pneumatic tactile display. Each part of the device was made of polycarbonate by using machining whose tolerance was less than 100 μm. The sidewall balloons were made of 600 μm thick latex film that was attached to each side of processed polycarbonate structures. The top surface of the core body was attached with 100 μm thick latex film to form nine balloons generating normal tactile actuation. Silicone tubes were connected to the bottom of the structure to apply pneumatic pressure. All of the structures, including the core body, inner frame, outer frame, and cover were manually assembled to complete the device fabrication. Figure 4b is a photo of the fabricated capacitive position sensor. The sensors were manufactured with FPCB using copper electrodes. The signal lines were positioned over a shielding metal layer to block high-frequency environmental noise. As described in Sect. 2 and shown in Figs. 2 and 4, the sensors for measuring the x- and y-positions were formed in the gaps between the inner and outer frame, and inner frame and core body, respectively. After the complete assembly of the tactile display, the minimum and maximum capacitances were measured as 0.12 and 1.56 pF, respectively.

Fig. 4
figure 4

a Photograph of the fabricated three-axis pneumatic tactile display, b photo of fabricated capacitive position sensor

Full size image

3.2 Feedback control unit

Figure 5a shows a block diagram of the overall feedback-controlled tactile display system. A total of four capacitance values from two capacitive sensors were read and converted into corresponding voltage values successively by a readout circuit. Then data was transferred to a computer through a data acquisition device (DAQ, NI USB-6259, National Instruments) to calculate the current position of the actuation body using LabVIEW (National Instruments) software. A feedback-control algorithm controlled the actuator body by generating control signals for the pneumatic pressure regulators after comparing the current and target positions.

Fig. 5
figure 5

a Overall block diagram of the implemented tactile feedback system, b schematic of the capacitive readout circuit, c a readout circuit board

Full size image

The readout circuit for measuring capacitance values was based on the switched-capacitor method, as shown in Fig. 5b. In this circuit, an amplifier circuit through a switch block sequentially read each of four capacitors. The output voltage signal was stabilized with a regulator and an R–C filter circuit. A complex programmable logic device (CPLD, XC2C64A_7VQG44C, Xilinx) controlled the timing of the switching. Figure 5c is a photo of the readout circuit board, including power unit, clock generator, CPLD, and capacitance readout unit. It took 12 μs to read a single capacitance value, which resulted in an operational frequency of 20.8 kHz.

4 Results and discussion

The inflation of a balloon for normal tactile actuation on the core body was monitored using a camera while applying pneumatic pressure. As shown in Fig. 6, the center deflection was linearly proportional to the applied pressure (King et al. 2008) and was measured at about 1.5 mm under a pressure of 110 kPa.

Fig. 6
figure 6

Center deflection of balloons on top surface of the core body for various pneumatic pressures

Full size image

The capacitive sensor was verified by measuring the voltage output signals from readout circuit. The output signals, V out1 and V out2, from each capacitor formed by a common rectangular electrode and two ramp-patterned electrodes, can be expressed as

$$V_{\text{out1}} = V_{\text{in}} \frac{{C_{s1} }}{{C_{\text{F}} }}$$
(1)
$$V_{\text{out2}} = V_{\text{in}} \frac{{C_{s2} }}{{C_{\text{F}} }}$$
(2)

where V in is input voltage, and C F is feedback capacitance in Fig. 5, which in our design were 5 V and 2 pF, respectively. The position value (P) of the core body can be expressed using V out1 and V out2 as follows:

$$P = \frac{{V_{\text{out1}} }}{{V_{\text{out2}} + V_{\text{out2}} }} = \frac{{C_{s1} }}{{C_{s1} + C_{s2} }} .$$
(3)

The device was designed to have a position value ranging from 0.07 to 0.93, while the actuator body moves ±1.5 mm in the lateral direction. The position value becomes 0.5 when the actuator body is at the center, locating the rectangular electrode at the center of ramp-patterned electrodes. In this work, the output value was measured at seven points with 0.5 mm of lateral movement. As shown in Fig. 7, the position values from the capacitive position sensor varied linearly with the lateral position of core body, showing very good agreement with the theoretical value. The largest discrepancy for the target position was measured as 0.04 mm for a moving distance of 1 mm, which might be caused by parasitic capacitance formed by wiring metal lines in the FPCB capacitive sensor units. We have compensated for this embedded error in the feedback control algorithm to achieve accurate position control.

Fig. 7
figure 7

Position value (P) output from capacitive sensor through readout circuit for various lateral position of the core body

Full size image

The positioning accuracy is shown in Fig. 8. We moved the actuation body to seven different positions in both of x- and y-directions. Blue dots in this graph show the positions of the inner frame moving along the x-axis, while the position on the y-axis is fixed at the center. The data was averaged after measuring the position three times. Our previous tactile display system, without feedback control, showed severe hysteresis, resulting in poor repeatability and inaccurate positioning (Doh et al. 2013). Using capacitive feedback control in this work, the proposed tactile display showed very linear operation and good repeatability in position, with very low hysteresis. The y-directional position control is shown with red dots where the position of the core body moving along the y-axis was measured while maintaining the x-position at the center. The average discrepancy was 0.04 and 0.06 mm for x- and y-directional movement, respectively. The larger discrepancy in y-directional movement might be related to the size of the actuation balloons. The balloons actuating the body in the y-direction were relatively small because of the restricted sidewall area of the core body. As the dimension of balloon decreased, more pneumatic pressure was required, which resulted in large operation error.

Fig. 8
figure 8

Center position of core body in x- and y-direction for various input target positions

Full size image

Two-dimensional control accuracy was monitored by repeatedly measuring the difference between the target position and the actual destination. It was measured three times, for a total of nine points at intervals of 0.5 mm in each axis, as shown in Fig. 9a. In that figure, the measured moving positions are marked as dots with different colors and shapes for three repeated tests. The control accuracy was measured as 0.11 mm in both x- and y-directions. Because a human finger can discriminate a spatial tangential resolution of more than 2 mm, the accuracy of our device will be enough for application in tactile display.

Fig. 9
figure 9

a 2-D plot of output positions from three repeated tests with input target positions (crossing points of dashed lines), b photographs of the actuator at nine different output positions

Full size image

Figure 9b shows photographs of tangential actuation at nine different positions: (−1.5, 1.5), (0, 1.5), (1.5, 1.5), (−1.5, 0), (0, 0), (1.5, 0), (−1.5, −1.5), (0, −1.5), and (1.5, −1.5). Here, the top surface of the core body was covered with a thin paper that had marks measuring the displacement.

Finally, we considered the lateral positioning accuracy when normal pressure is applied to the actuator, because the grasping motion of a finger will push the core body in the normal direction. To evaluate this, we measured the tangential motion of the core body in the y-direction while applying normal force to the core body. The core body was moved to five different target positions while varying the force from 0 to 4 N. As shown in Fig. 10, the positioning was quite accurate regardless of the normal force, which is an important advantage of our feedback control-based actuator.

Fig. 10
figure 10

Accuracy of output positions under various normal forces on the core body

Full size image

5 Conclusions

In this paper, we developed and demonstrated a three-axis pneumatic tactile display integrated with capacitive displacement sensors for feedback position control. By integrating capacitive sensors for feedback control onto the sidewalls of moving structures, we could obtain accurate control of tangential displacement regardless of the normal force applied to the device. The structure was made of polycarbonate with the use of precise machining. We achieved tangential 2-D actuation by applying pneumatic pressure on sidewall balloons, and consequently moving the core body in lateral directions. In addition, nine balloons were formed on the top surface of the core body to generate normal tactile stimulation on a fingertip. The pressure on each balloon was controlled accurately using electro-pneumatic regulators and LabVIEW. The size of the tactile display was 26 × 26 × 18 mm3 and the accuracy of lateral position control was measured at 0.11 mm. In this study, we demonstrated that the accuracy of a three-axis tactile display can be improved by integrating a feedback control unit, and will be a good candidate as a tactile feedback application for robot-assisted surgical systems.