Evaluation of Planar Inkjet-Printed Antennas on a Low-Cost Origami Flapping Robot

An investigation on antenna solutions for expendable origami paper flapping robots is presented. An origami flapping robotic bird that can be produced using standard A4 paper is employed. Antennas resonating at the two commonly used frequency bands, 2.4 GHz and 5.2 GHz, are designed for the limited space available on the folded origami structure. Two developments of the same dual-band monopole antenna are discussed. The first antenna is located on the robot’s spine and the second on its tail. A space diversity configuration is also studied. The antennas are printed directly onto a photo paper substrate and then folded into an origami robotic crane’s body structure. An ordinary desktop inkjet printer fitted with silver nanoparticle conductive ink cartridges has been employed. CST Microwave Studio™ has been used to design the antennas. A good agreement between the measured and simulated $S_{11}$ results is achieved with a reasonable −10dB impedance bandwidth realized in the three cases studied. The radiation patterns are omnidirectional in the XZ plane which is desirable for the specific application. The diversity configuration has a mutual coupling of <−23dB and a gain of 1.4 dB and 2.8 dB at 2.4 GHz and 5.2 GHz respectively. The aim is to provide a new vision for antennas embedded into expandable flying robots based on traditional origami structures.

In most cases, this antenna should provide omnidirectional radiation pattern and vertical polarisation for efficient communication with the controller [13].
3D inkjet printing with nanoparticle silver ink is a rapidly advancing manufacturing method. It is a layer by layer fabrication process where a design can be printed directly from a digital model. It offers rapid prototyping and can reduce manufacturing process cycles in areas such as electronics, microwave and radio frequency (RF) [14].
This article investigates the integration of inkjet-printed antenna on a flying robot that is produced using traditional origami techniques. This type of flying robots are gaining popularity in the research and development community [28], [29]. A compact dual-band antenna consisting of a triangular shape with a horizontal slot on top of a semi-elliptic monopole is employed. Low-cost inkjet printing on photo paper substrate is used. It is the first time that a geometry has been developed to specifically fit within the folds of a traditional origami bird folded from an A4 paper sheet. Three positions are found to be suitable for the antenna: triangular neck, spine and tail. Consequently, two versions of the same antenna designs are produced. One fits on the robot's spine while the other fits on the tail and/or the neck. The latter is further investigated in a diversity configuration.
The rest of the paper is organized as follows: Section ? describes the design and analyses the antenna located on the spine, section ? studies the antenna located on the tail, section ? describes diversity antenna system solution while section ? is the conclusion and discussion.

II. DESIGN AND ANALYSIS OF THE SPINE ANTENNA A. ANTENNA DESIGN ANALYSIS AND GEOMETRY
A traditional flapping origami robot with potential electromechanical and antenna components locations is illustrated in Fig.1. Fig. 2(a) shows the fold lines along which the origami bird is folded from a 210 mm square (about A4 size) sheet. It also shows the location for a spine antenna. The bands targeted for the flying robot application are 2.4 GHz and 5.2 GHz WLAN/WiFi communication bands [30], [31]. The proposed planar monopole antenna ( Fig. 2(b)) is based on [32], [33] and its geometry is modelled around [34]. It is a leave shaped monopole with a semi-elliptical bottom and a  triangular top with a horizontal slot. Its dimensions are given in Table 1. The lower semi-elliptical section fit in the lower wider section of the triangular spine while the triangular upper part fits in its apex as shown in Fig. 2 (c). The antenna orientation is as that of the spine i.e. upright. A 50 CPW line feeds the antenna. Two rectangular ground plane are on either sides of the CPW feed line.
The lower frequency (f L ) of a planar elliptical disc monopole corresponding to VSWR = 2 is approximated by equating its area to the area of an equivalent cylindrical monopole of the same height L and equivalent radius r: where R 1 and R 2 are the longer and shorter half lengths of ellipse's axes respectively. Equation (1) thus gives: A quarter wavelength monopole antenna has an input impedance equivalent to half of that of a half wavelength dipole antenna. The input impedance of an infinitesimally thin monopole antenna is thus 36.5 + j21. 25 , an inductive impedance. The real impedance is obtained by using a slightly smaller length, L, of the monopole given by [35].
where λ is the free space wavelength and F is the length-to-radius parameter for the stub monopole given by From equations (3) and (4), the wavelength λ is obtained as: Thus the lower frequency f L is: After accounting for the effect of probe length, p, which increase the antenna length and thus reducing the frequency, (6) can accordingly be modified to where f L is in GHz and L, r and p are in cm. For the half lengths axes R 1 and R 2 , the ellipticity ratio, R 1 /R 2 , chosen was 1.3 to restrict the bandwidth to the range of interest. The antenna was fed along the minor axis of the semi ellipse. For elliptical monopole fed along the minor axis, L = 2R 2 and r = R 1 /4. For a f L equal to 2.4 GHz, the radii R 1 and R 2 of the semi-ellipse after optimization were obtained. A planar triangular monopole of height H (height from feed point to apex) and base length W was superimposed on the semi elliptical monopole. H is the height of the equivalent cylindrical monopole of radius r.H and r are calculated from (6), (8) and (10): The actual height H is less than the calculated height due to dielectric substrate and fringing currents effect. It should be noted that the superimposed rounded triangular monopole base is the semi-elliptical disc. The model was simulated on a 0.177 mm thick paper substrate of relative permittivity, ε r , of 3 using CST Microwave Studio TM . Fig. 3 shows the response of the various component of the radiator. The semi elliptical radiator provides a resonant at upper band. The semi-elliptical radiator with superimposed triangular monopole exhibits wideband characteristics. The lower frequency is higher than the targeted 2.4 GHz. This could be due to the rounded (semi-ellipse) base of the triangular monopole which reduces  the size of radiator thus raising f L . By inserting and optimizing the dimension of a horizontal slot across the triangle, the lower resonance point was shifted to 2.4 GHz while maintaining the upper band as well as rejecting frequencies between the two bands of interest. The final antenna has two distinct resonant modes corresponding to the triangular radiating section and the semi ellipse. The gap 'p' between the ground plane and radiating acts as a matching network and improves impedance bandwidth. The optimum impedance bandwidth is achieved when the capacitance due to the gap 'p' between the radiating element and ground plane edges balances with the antenna inductance.
The size of the ground plane is also critical in the design of compact antennas [36], [37]. The surface current distribution on the antenna is shown in Fig. 4. As expected, at 2.4 GHz, Fig. 4 (a), most of the currents is concentrated in the triangular section. At 5.2 GHz, Fig. 4 (b), most of the currents are in the lower section of the semi-ellipse. As with all compact antennas on small ground planes, some currents are also present on the antenna's ground elements. These currents are strongest at the upper edges of the rectangular ground planes at 5.2 GHz than at 2.4 GHz. Taking this into consideration, it can be inferred that the surface current distribution could change with the size and the shape of the ground plane and/or with introduction of cable and/or RF connector to the antenna. This is likely to introduce differences in the measurements in relation to simulations. VOLUME 8, 2020  3 GHz to about 2.6 GHz at the lower band and from 4 GHz to 6 GHz at the higher band. There is no significant change in S 11 from 1 • to 180 • . Only angles of less than 1 • produced an S 11 higher than −10 dB at the target 2.4 GHz to 2.5 GHz band. In the worst scenario, at 0 • , the S 11 was less than −8 dB across the 2.4 GHz band.

C. SPINE ANTENNA FABRICATION AND MEASUREMENTS
The CST model was exported to Gerber (single layer) and then converted to a PDF file using ViewMate. The PDF was then printed using silver nanoparticles ink that is based on chemical sintering. Chemical sintering technology was used to avoid the prolonged heat sintering and the damage it can cause when the substrate used is a paper based material [38]. In chemical based sintering, a polymer latex and halide emulsion solvent dissolves silver nanoparticles of less than 0.1µm. Conductivity appears moments after the solution has dried. This makes the ink based on chemical sintering ideal for inkjet printing of the antennas. A 1200 × 6000 dpi resolution and piezo with 210 x3 colour print head home desktop Brother MFC-J5910DW printer was employed. AgIC-AN01 Silver Nano Ink was loaded in the 3 colour cartridges to produce a thick print layer that can achieve sufficient conductivity. The silver ink provides a resistivity of 0.2 /sq when printed on an AgIC-CP01A4 photopaper [39]. Further analytical tests on a variety of printed tracks were carried out using the equation for the resistivity, ρ: where R is the measured resistance, A is the cross-sectional area and l is the length of the track. The values of resistivity obtained were between 2 × 10 −7 and 3 × 10 −7 which is about the same value to the one provided by the manufacturers for a layer thickness of about 1µm. The printed antenna is shown in Fig. 7(a). The printout was folded into a robotic bird. The antenna was concealed inside the structure to protect the printed metallic layers from environmental damage as shown in Fig. 7(b) Thermal and mechanical stability of the printed layers were provided by AgIc. Heat resistance can be allowed for up to 30 minutes at 100 • C. In cross cut tests (ISO 2409), the mechanical stability is between 0 to 1 (0 is highest and 5 is lowest).
Surface profiles of the antenna's silver ink were obtained using white light interferometry (profilometer). The images at 50x magnification are shown in Fig. 8(a). An average surface roughness of approximately Sa 200nm was obtained. The image shows that the 3D print is not uniform as the substrate can be seen through the silver ink. The droplets are visible and the measured roughness is for both paper substrate and silver droplets. Fig. 8(b) is an optical image of top view of the printed antenna of 1µm silver ink height taken with x10 magnification which shows uneven surface. Black regions are seen within the sample which could be deeper unevenness of ink surface. Fig. 8(c) shows the narrow gap between CPW and the ground plane taken with x20 magnification for the 1µm silver ink height. Some isolated silver ink particles can be seen inside the gap but their isolation prevents short circuit across the two parts. The presence of ink particles in the gap does not have any significant effect on the performance of the antennas as the foregoing tests shows. This could be because the particle sizes are much smaller compared to the wavelength of the frequencies involved.
To test the antenna, a 3.5 mm SMA connector was attached to it using Araldite Rapid Epoxy Adhesive. The antenna was concealed inside the structure to protect the printed metallic layers from environmental damage. A Rohde & Schwarz  ZVL vector network analyser was used for S 11 measurements. Fig. 9 shows the simulated and measured S 11 when the wings are horizontal. The measured lower band has −10 dB impedance bandwidth from 2.2 GHz to 2.7 GHz while the upper band has a bandwidth from 3.9 GHz to 5.5 GHz thus covering both target bands. The slight discrepancy between the simulated and the measured S 11 could be due to fabrication, folding and measurements errors. These includes non-uniformly deposited silver conductive ink and resistive losses of the printed tracks, the bending of the substrate, as well as the SMA connector. Fig. 10 shows the simulated and measured radiation patterns at (a) 2.4 GHz and (b) 5.2 GHz for the robot's horizontal wing position i.e. α is equal to 90 • . The simulated results show that as expected for this type of antennas, the planes xy and yz has nulls in the y axis while plane xz displays an omnidirectional pattern. Tests showed no significant effect on the radiation pattern when the position of the wings is varied from 0 • to 180 • relative to the robot's core. The measured antenna radiation patterns are consistent the simulations. As with planar monopoles [40], it exhibits high cross polarization. The high cross polarization component as a result of excitation order modes, radiation due to Jx current at the top edges of the ground plane near the radiating element. Discontinuity at substrate and metallic radiator results in surface waves adding to cross polarization [41]. Fig. 11 shows the gain vs frequency response for the selected frequency bands. The measured antenna gains were 1.4 dBi and 2.7 dBi at 2.4 GHz and 5.2 GHz respectively, about 0.2 dB and 0.5dB lower than the simulated one. The differences between the simulated and measured gain could be due to fabrication and folding errors, cables and connectors.

III. A SINGLE CPW-FED TAIL ANTENNA A. TAIL ANTENNA DESIGN
An alternative antenna location is the tail of the origami crane. Fig. 12(a) illustrates the location of a tail antenna VOLUME 8, 2020  as well as the potential symmetrical location for a neck antenna, on 210 by 210 mm unfolded photo paper substrate. Fig. 12 (b) show the antenna's structure and dimensions while Fig. 12(c) shows its location on the robot's tail and rotated to fit on the tail's orientation. Like the previous antenna, the antenna is concealed inside structure to protect the printed metallic layers from environmental damage. The antenna is similar in shape and construction to that described in section IIA with slight dimension adjustments to fit in the narrower tail area and achieve the bandwidths of interest. Table 2 show its final dimensions. The same dielectric material and substrate thickness was used.
The simulated S 11 results, Fig. 13, indicates a narrower −10 dB bandwidth at the 2.4 GHz band but a wider −10 dB bandwidth at the 5.2 GHz bands for the angle α of wing    the flapping. The antenna covers the desired frequency bands for the range of angles of flap. They also show a near negligible effect on S 11 by the flapping of the wings. This could be due to the antenna on the tail being removed from the wings bearing section of the origami bird structure. This makes the antenna less obstructed by the flapping of the wings compared to the antenna on the spine (Fig. 7b). Fig. 14(a) and Fig. 14   indicates a high level of surface current at the triangular and rounded sections of the antenna at 2.4 GHz and 5.2 GHz resonance points respectively. A high surface current level is also observed on the ground plane at the two bands due to its small size which makes it a sensitive part of the radiating structure [37].

B. FABRICATION AND MEASUREMENTS
The fabrication procedure used in section IIC was followed for the tail antenna. Fig. 15(a) shows the printed tail antenna on a photo paper and Fig. 15(b), the origami crane after the paper is folded up. S 11 of less than -10 dB at both 2.4 GHz and 5.2 GHz bands compares well to that of the simulation as shown in Fig. 16 for flap angle α of 90 • . Fig. 17 shows simulated and measured antenna radiation pattern at (a) 2.4 GHz and (b) 5.2 GHz bands. Both planes xy and yz have nulls in the y-axis while plane xz indicates omnidirectional radiation. There are differences between the measured and the simulated radiation pattern. This could be due to the small ground plane and the introduction of RF connector/cable to the antenna which affects the current distribution in the ground plane. This changes the impedance and radiation performance. This results in the measured results being slightly different from the simulated results and potentially can make the design validation difficult.    A diversity antenna was created on the neck of the crane, as illustrated in Fig. 19 and also rotated to fit the neck orientation. Each antenna requires an independent RF circuitry and electronics connection and ability to communicate with the control station. The S-parameters i.e. S 11 , S 12 , S 21 and S 22 of the simulated model were obtained and are shown in Fig. 20. S 11 curve is slightly different from S 22 curve. This is due to the robot's wings extending over the front antenna and acting as an extra layer of substrate. S 11 of −10 dB impedance bandwidths from 2.36 GHz to 2.55 GHz at the lower band and 5.07 GHz to 6 GHz at the higher band respectively were realized. S 22 of −10 dB impedance bandwidths from 2.27 GHz to 2.50 GHz and 5.11 GHz to 6 GHz at lower and upper bands respectively were realized. S 12 and S 21 of less than −10 dB at the resonance frequencies indicates a good antenna isolation between the two antennas. Simulated diversity antennas co-polarization radiation patterns are shown in Fig. 21 (a) at 2.4 GHz and (b) at 5.2 GHz bands for planes xy, xz and yz for the tail and neck antenna. The radiation pattern indicates nulls about the y-axis for xy and yz planes and an omnidirectional radiation pattern for the xz plane for both the tail and neck antennas at both bands. There is improvement in coverage in all planes compared to just one antenna on the tail. The radiation patterns of the two antennas are generally in the same direction that could be due to the two antennas being similarly orientated.

B. FABRICATION AND MEASUREMENT
The diversity antennas were fabricated, Fig. 22, as per the procedure described in section 11C. A VNA was used to measure the S-parameters whose plots are shown in Fig. 23. An S 11 −10dB impedance bandwidth from 2.3 GHz to 2.55 GHz at the lower band and 4.5 GHz to 6 GHz at the upper band respectively was achieved for a flap angle α of 90 • . An S 22 −10 dB bandwidth from 2.2 GHz to 2.65 GHz at the lower band and 4.3 GHz to 6 GHz the upper band respectively was also realized. S 12 and S 21 of less than -23 dB at both 2.4 GHz and 5.2 GHz bands indicates good isolation between the two antennas.
The orientation of the axis for the measurement of the radiation pattern of the neck is symmetrically to that for the tail. Fig. 24 depicts the measured radiation patterns of planes xy, xz and yz respectively at (a) 2.4 GHz and (b) 5.2 GHz for the two antennas. It indicates nulls in the xy and yz planes and omnidirection in the xz plane for two frequency bands which is commensurate with simulation results. As expected, the radiation patterns for the two antennas are similar. Improvement in coverage is expected when connected to a diversity antenna system. Table 4 compares previous dual bands WLAN inkjet printed antennas with the one proposed in this work. It can be seen that with comparable gain and bandwidth, the proposed antenna is relatively small in size and is the only one specifically designed to fit the space available in the origami flapping crane.

V. DISCUSSION AND CONCLUSION
The integration of inkjet printed planar monopole antennas on an origami flapping robot has been demonstrated. A wideband monopole antenna consisting of a semi-elliptical shape with a triangular part and a horizontal slot has been developed and tested. The dimensions of the triangular section determine resonance at 2.4 GHz while the dimensions of the semi-elliptical section determine resonance at the 5.2 GHz band, the unlicensed frequency bands for drones' control.
The antenna fits on the available space on the traditional origami structure and is able to operate at the required bands. Two optimal solutions were achieved: one on the spine and the other on tail and/or neck.
A diversity system involving the tail and neck antennas was also realized. This increases coverage in a communication system. All the antennas exhibited nearly omnidirectional radiation pattern in the XZ planes at both frequency bands making them suitable for the purpose. The monopole antennas have been fabricated on standard photo paper substrate. A chemical sintering-based silver nanoparticle conductive ink cartridge was used to print the antennas using an inexpensive and ordinary home inkjet printer and photo paper. This made fabrication of the antenna cheap and fast. The successful outcome promises the potential of integration of antennas with flexible electronic systems using inkjet printing technology.
This holds out prospects of appropriation of instant printing technology for fast integration of antenna and flexible electronic systems on paper substrate for future wireless controlled aerial robots in line with [45]. This promises potential realization of other electronics components on a paper substrate as well as potential realization of fully integrated expendable robot.