Self-Powered Multimodal Sensing Using Energy-Generating Solar Skin for Robotics and Smart Wearables

Wearable electronic devices‐laden systems such as electronic‐skin (e‐Skin) have been explored in recent years to enable advances in applications such as Internet of Things, healthcare, and robotics. The power requirement of multitudes of devices in the e‐Skin is a major hurdle for its wider uptake. Herein, a solar cells‐based energy generating e‐Skin is presented and how the energy outputs of solar cells can be innovatively processed for multimodal sensing is demonstrated. By reading the variations and energy output patterns of the e‐Skin, present on a robotic arm, multiple parameters can be sensed including object motion, color detection, and ambient temperature. With the accurate tracking of shadow sensing, for an object moving in horizontal and vertical directions with respect to the solar skin, information can be obtained such as the velocity and acceleration of moving object. In this regard, the presented e‐Skin can also be seen to have vision capability. The presented multifunctional energy‐generating e‐Skin shows an energy surplus of >1 mW (effective module area of 20 cm2) under white light illumination of 4,450 lux, which is sufficient for continuous powering of portable low‐powered devices. Finally, we demonstrate the e‐Skin application for energy‐autonomous hand gestures recognition in robotics.

act as sensors.With such devices, the e-Skin could allow monitoring of an object's position, speed and acceleration of movement, ambient temperature, etc.With such an architecture, the excess energy generated during the sensors' sleep cycle would be used to power associated electronics leading to highly energy-efficient, self-powered e-Skin.
Solar cells have traditionally been used to convert sunlight into electricity, but recently it has been shown that they could be employed as sensors to gather ambient environmental information, thus enabling the use of photovoltaic cells for applications such as detection of hand gestures and light/shadow. [51,52]Using solar cells as sensors presents key advantages as they do not require external power.On the contrary, they can provide the energy needed to power complex systems.On these lines, we recently reported the first energy-generating e-Skin with intrinsic tactile sensing, i.e., without any touch sensor. [29]In this article, we extend this approach and present a simple and inexpensive solar skin, used with inference algorithms, to detect key indoor sensing parameters.The solar skin is developed by connecting in series four off-the-shelf solar cells over a bendable and biodegradable paper substrate.The realized solar skin is integrated into a robotic arm and used to detect a wide range of dynamic motion (both in vertical and horizontal directions) of an object moving in the proximity of the robot providing parameters such as velocity, direction, and position of the object, light's color detection, and ambient temperature (Figure 1).For motion detection, the output voltage on individual solar cells was compared with each other.The solar skin has been shown to detect dynamic horizontal motion for the object moving at different speeds (1, 5, and 10 cm s À1 ) between two different points (10 and 20 cm) from solar skin.For vertical object motion detection, the solar skin can determine the object motion with different speeds (1, 5, and 10 cm s À1 ) between two points that are at 3 and 40 cm heights.Finally, the solar skin is shown to detect different colors of light by monitoring the voltages generated with different colors (having different wavelengths) and ambient temperature.For temperature sensing, first, a single solar cell is characterized to extract the sensitivity (≈À1.8 mV °CÀ1 ), followed by sensing of ambient temperature using the developed solar skin.The performance of the solar skin as an energy generator was tested under different lighting conditions.With an energy surplus of >1 mW from a module with an effective area of 20 cm 2 (under white light illumination of 4450 lux), the solar skin has shown potential to power-up portable low-power portable electronic devices.The integration of presented e-Skin on an adult human-size humanoid (considering adult skin has an area of 1.5 m 2 ) could generate 0.75 W of energy.

Results and Discussion
The solar cells can read changes in ambient light, just as the photodiodes do. [53]But, in comparison to photodiodes, solar cells are active as they do not require an external power to operate.Moreover, photodiodes suffer from light saturation that limits their use for indoor sensing applications. [54]In the present work, we have developed the solar skin on a paper substrate and placed it over a robotic arm to evaluate its multifunctionality, namely: 1) energy generation and 2) self-powered multimodal sensing (Figure 1).To fabricate the solar skin, first, the as-received commercial monocrystalline silicon solar cells (original size 5 cmÂ5 cm) were laser cut into five pieces to obtain 5 cmÂ1 cm sized cells.Following this, four solar cells with the same dimensions were attached on a paper substrate and connected in series using an electrically conductive and flexible textile tape (Figure 1b,c), thus creating a flexible solar skin able to wrap around a robotic arm.Output terminals were subsequently added to the prototype to measure the total voltage by the solar cell unit (node) in the circuit.Since the voltage read at each terminal node represents the sum of the voltages produced by the preceding solar cells, their individual voltages can be calculated by subtracting the two adjacent measurement nodes (Figure 1b).The output of the solar skin is sampled by a STM32 H7A3ZI-Q microcontroller board via the onboard 16-bit analog-to-digital converter (ADC) at 32 MHz.The solar skin is tested under varying indoor illumination conditions with an intensity between ≈850 and ≈4,500 lux.Table 1 shows the relationship between the light intensity reaching the solar skin and the distance of the source from it.For a human-robot collaboration working in a proximity to each other, safety and comfort are the two important aspects to consider. [55]The flexible aspect of the solar skin should take care of the user's comfort and for safety the robot must be able to avoid dynamic obstacles with high reliability.Toward safety, we have carried out motion sensing by moving an object in a proximity of a robot and obtaining essential information such as the direction of the movement, position of the object, speed, and acceleration.This set of information is critical to ensure safer HMI.

Horizontal Motion (Shadow) Sensing
In this study, object motion was inferred from the designed solar skin.With collected data, we were able to obtain the direction of movement, as well as the speed at which the object's shadow was cast on the module.By combining the information from each cell, we obtained the velocity and acceleration at which the object was moving relative to stationary cells placed on a robotic arm.By understanding the output of the solar skin module during a moving object, we also developed a strategy for a safe HMI.It is an essential first step toward extending the sensing capabilities to more complex scenarios, i.e., when both the module and the obstacle are in motion relative to each other.Additionally, restricting the sensing conditions to a single source of light allowed us to comprehensively characterize, and understand the operational requirements of solar cells for indoor applications.During the horizontal motion sensing, we also extracted other parameters, such as the direction of movement, from the sequential response of the cells.
Figure 2 shows the solar cells output voltages when an object traveled in lateral direction (parallel) between the skin module and a white light source.Here, the light source is a 4 W LED lamp placed at 50 cm distance from the solar skin module.Under this arrangement, the light intensity reaching the solar skin was ≈1680 lux, as measured using a RS Pro RS-3809 commercial light meter.The testing arrangement is schematically shown in Figure 2a.The temporal patterns of solar cells output, when the object moved with varying speed (s) at a fixed distance (d) from the solar skin, are shown in Figure 2b-g.For this set of experiments, we performed two types of experiments (d was fixed at 10 and 20 cm).These temporal patterns in Figure 2b-g allowed to accurately detect the object's lateral motion.The magnified data from these results are shown in Figure 2h (d = 10 cm) and Figure 2i (d = 20 cm).The data evidently show the direction of movement of the object, and this is also reflected in Figure 2h,i.The first node (V N1 ) responds first when the object is moving from node 1 to node 4. When the object moves in the opposite direction (node 4 to node 1), the first node responds last in the sequence.Thus, the direction can be inferred, along with other parameters such as velocity (inferred from the transient response time of any node), acceleration of the object (first-time derivative Table 1.Magnitude of light intensity reaching the solar skin and the distance of light source from the solar skin. of velocity), and position (from the time-series data produced by the motion of the object).
The sensing of solar cells is based on the observation that an approaching object interferes with the light rays reaching the solar skin in a way that depends on the shape and speed of the object.This will leave its distinguishable signature in the monitored output voltage or photocurrent's time series data.In the experiment, shadow events were successfully detected more than 5 times to show the repeatability in performance metric.Once the motion of an object is detected, the time difference .It is to be noted that the given distance is from solar skin.e-g) Temporal output voltage patterns for object moving with varying speed when paced at a fixed distance of 20 cm between the light source and solar skin: e) 1 cm s À1 , f ) 5 cm s À1 , and g) 10 cm s À1 .h-i) Temporal output voltage curves summarizing the effect of speed when the object is placed at: h) 10 cm and i) 20 cm.
between the response of each solar cell can provide insight into the speed with which the object is moving, as well as other parameters such as its acceleration.Further, the information could also reveal the angle of the light source relative to the sensor module.Since the solar cells are connected in series, once the object starts casting shadows onto the first cell in the module, the voltage seen at the last cell will start to decrease as the object moves toward it.Then, the time taken by the voltage of any cell in the module to decay and stabilize is measured.Finally, the lengths between the edge of the first solar cell to the next cells in the sequence were measured to be 2.6 cm (to the edge of the second node), 4.2 cm (to the edge of the third node), and 6 cm to the end of the 4th node.These distances were divided by the measured decay time to find the speed with which the object was moving.For the 3 speeds of the obstacle measured, the voltage reduction time of the last solar cell in the series was measured across each node in the solar skin module.Table 2 shows the calculated time and corresponding speed of the object for each of the test cases.The difference in computed speeds seen by different solar cells can be attributed to issues such as resolution and losses in precision and the robotic arm (used to control the speed of the object) applies an acceleration profile to the motion.In a real-world situation, the discrepancy in speed could provide insight into the motion of an object with varying speeds.

Vertical Motion Detection
When an object comes between the light source and the skin module, its shadow leads to spatial variation of the generated output voltage and could provide the shape and motion of the object.Further, on approaching the solar skin, object's shadow dimensions will change, and this too would change the magnitude of the output voltage.Using such variations in the output voltage, it is possible to obtain the object's distance from the solar skin.Also, if the approaching speed of the object is varying, the pattern of variations in the output voltage could provide information such as speed and acceleration (thanks to the rapid response of the solar cells).To examine the solar skin performance for vertical motion (shadow) sensing, the light source was fixed at the 50 cm distance from the solar skin while an object was moved between 3 and 40 cm at different speeds (1, 5, and 10 cm s À1 ).The schematic representing the testing arrangements and the electrical data generated is shown in Figure 3.In this test case, compared to the horizontal shadow sensing, all the solar cells respond at the same time.The maximum output voltage of ≈0.3 V (from node 4) is monitored when the object is at the maximum distance from the skin module (40 cm).When it reached  the 3 cm distance, the output voltage from node 4 was reduced to ≈0.15 V for experiments performed at each speed.Similarly, output voltage from all the other nodes showed a decreasing trend when the object approached the solar skin.It is interesting to note that the output voltage from all the nodes changed almost linearly for the entire measured distance (between 3 and 40 cm) and for all the speeds.When the object reaches at 3 cm distance, the output voltage reaches the lowest value and then is saturated.
Because of the high linearity, the vertical shadow sensing can be accurately determined between 3 and 40 cm distance from the solar cells.This shows the proximity-sensing capability of skin for static as well as moving objects.Three cycles were monitored for each speed showing repeatability in sensing response.Further, if the voltage change with respect to time is analyzed, it is possible to obtain the speed and acceleration with which the object is moving relative to the skin module.Interestingly, distinct output voltage patterns were noted for different speeds.
For instance, sinusoidal-like pattern was noted for an object moving with the slowest speed (1 cm s À1 ) (Figure 3b).Likewise, the obtained output voltage signals looked like square pulses at the highest speed (10 cm s À1 ).Therefore, the output voltage patterns could also provide a qualitative information regarding the object's speed and acceleration while the monitored time between the maxima and minima of the output voltage could provide quantitative information.

Light Color Detection
The solar cells respond selectively to light wavelengths (from ultraviolet (UV) to near-infrared (NIR), including monochromatic visible light) and hence to the type of light they are exposed to. [56]In fact, they could be designed to provide optimum output for light available in specific environments, and in this regard, the cells for indoor and outdoor light energy harvesting could be different.Considering this, the presented solar skin could also be used to detect a wide range of light wavelengths.Shorter wavelengths (in extreme UV range) are expected to pass through a solar cell.In contrast, wavelengths that are too long (in IR region) will not have enough energy to "excite" the electrons in the solar cell to produce voltage output.Considering this background, we performed experiments to see if the developed solar skin would create a unique output voltage to a specific monochromatic wavelength.Different colors of light, having different wavelengths (different energy levels), were used and the magnitude of output voltages produced was observed.In the present case, we have used green, red, and blue light colors and compared the generated output voltage with white color.The results are shown in Figure 4.It is to be noted that the distance of the light source from the solar skin was fixed to 10 cm for all color illuminations.The results show that the voltage produced increases as the wavelength of a light decreases (when compared with monochromatic light illumination only).This is because the energy of light (photons) is inversely proportional to its wavelength and directly proportional to its frequency.For instance, red color having the highest wavelength in the visible light spectrum produces the lowest magnitude of voltage (≈0.5 V for node 4).The output voltage from node 4 increases to ≈0.9 V when illuminated with green light color.The same node yielded ≈1.1 V output voltage when light color was changed to blue.Finally, when illuminated with the white light, which is a combination of all colors in the visible spectrum, the cell/node generated the highest output voltage of ≈2 V.All the other nodes followed the similar trend of output voltage when illuminated with different colors.Accordingly, the solar skin could be successfully employed to detect indoor light colors.

Temperature Detection
The developed solar skin can also be used to measure the ambient temperature autonomously.Before evaluating the solar skin capability for practical applications, we obtained their sensitivity for temperature measurements in the range of 35-95 °C.The as-received off-the-shelf solar cell (5 cm Â 5 cm) was integrated on a paper substrate temperature.It may be noted that the temperature-sensing performance of cells could be reduced on a paper substrate, particularly when flexible interconnects are used to extract the light-induced charges.For these measurements, a solar cell on a paper substrate was placed over a hot plate and a white light energy source was placed at a 10 cm distance.The change in V OC with time is monitored for every 10 °C step change in temperature between 35 and 95 °C.Once the temperature reached to 95 °C, the hot plate was brought back to 35 °C.The experimental data are shown in Figure 5a.The data evidently show a decrease in V OC with increasing temperature.The response time between two readings is ≈70s (this includes the time required to reach the set temperature).The V OC value, at each measured temperature was noted and plotted with temperature, giving a temperature resolution of ≈À1.8 mV °CÀ1 for a wide detection range (Figure 5b).The decrease in V OC is attributed to the reduction in silicon bandgap with increasing temperature.The temperature dependence of band gap in semiconductors could be described using the Varshni relation as described in Equation ( 2): [57] E g ðTÞ ¼ E g ð0Þ À αT 2 ðT þ βÞ (1) where E g (T ) is the bandgap of the semiconductor at temperature T, E g (0) is the banggap value at T = 0 K, and α and β are constants.According to the Varshni relation, bandgap decreases with an increase in temperature.The decrease in bandgap of the semiconductor led to an increase in short-circuit current while V OC decreased.The obtained temperature sensitivity of À1.8 mV °CÀ1 compares well with the data reported in the literature for the single p-n junction-based silicon solar cells. [58]xperimental results display that the solar skin can effectively perceive the temperature change over a wide range.To further evaluate the temperature sensing capability of the solar skin for practical application, we evaluated its response to hot air.Figure 5c displays the photograph captured using a commercial IR camera of solar skin integrated over the robotic arm.The IR image confirms the skin temperature of ≈26 °C placed at room temperature.By blowing the hot air (the blow gun was set at 100 °C) on the solar skin, a real-time output voltage change curves were recorded from each output node to describe the temperature sensing transient response to continuous temperature change (Figure 5e). Figure 5d shows the IR photograph of the solar skin when the hot air was blown.The temporal output voltage data arising from different solar sensing nodes shows a dominant decrease in the output signal with respect to the node directly exposed to the hot air.The skin temperature, as measured by the IR camera, was ≈45 °C.The changed voltage magnitude from node 4 is proportional to the obtained temperature sensitivity of the solar cell.Further, as can be seen from this set of data the solar cell farthest from the hot gun showed significantly less change in the output voltage.This means that such a skin architecture could be used for self-powered temperature mapping in the future.

Energy Generation and Reliability
The solar cells used in the presented solar skin are commercially available monocrystalline PV cells.The purchased cells had an initial area of 25 cm 2 .To construct the solar skin, they were cut using a laser engraving machine to obtain 5 cm 2 size cells and characterized for energy generation.Figure 6 shows the electrical characterization of the solar skin in terms of output current, voltage, and power when illuminated by the white LED placed at varying distances (30-70 cm) from the skin module.As expected, under dark conditions, the solar skin has the electrical characteristics of a p-n junction diode.It is interesting to note that under bending (when placed on the arm of a robotic platform (Figure 6a)), the solar skin showed similar electrical characteristics to the one obtained under planar conditions (Figure 6b).Under white light illumination with varying light intensity, the I-V curve shifts as the solar skin begins to generate electrical power.The greater the light intensity, the greater the amount of shift which is controlled by varying the distance between the robot and the light source.For instance, the maximum power is generated when a white light lamp is placed at 30 cm distance from the solar skin module.At this distance, the measured intensity of light is 4450 lux, measured using a commercial light meter (Table 1).We can identify the short circuit current (Figure 6c), and the maximum power point (PMPP) (Figure 6d) with the varying light intensity.The output power reached its peak (PMPP > 1 mW) when the light source was placed at 30 cm.The larger the distance between the light source and the solar cell, the smaller the amount of energy produced.This is because light spreads out as soon as it leaves the source, meaning smaller and weaker light reaches the solar skin.This is evidenced by  , c) micrograph from a commercial infrared camera of solar skin wrapped on a robotic platform showing the maximum solar skin temperature ≈26 °C, d) measured temperature of the solar skin using commercial infrared thermometer when exposed to hot air gun set at 100 degree, and e) real-time solar skin response when exposed to hot air gun.The data show higher V OC variation of solar cells when directly exposed to hot air (maximum temperature ≈45 °C).
Figure 6d where the output power reaches its peak ≈150 μW when the light source was placed at 70 cm (light intensity of 850 lux).
The monocrystalline Si solar cells used here have excellent long-term reliability as they have an expected lifespan of ≈25 years and a typical efficiency of ≈20%.They offer higher stability than their polycrystalline counterparts and a hightemperature sensitivity. [59,60]Modules made of such solar cells are by far the leading solar cell technology in the market because of their higher efficiency and robust performance in outdoor conditions.However, the modules installed outdoors are exposed to variety of environmental stresses and hence, as a result, could suffer performance degradation.Environmental factors such as thermal stress and crack formation pose potential failure points for the proposed system.However, it has been shown that dust accumulation and material degradation are the leading aging factors of performance degradation of solar cells. [61,62]he impact of such issues greatly diminishes in the case of indoor operations.Further, the efficiency of the cells can be maintained through regular or automated maintenance, [63] making the proposed system suitable for deployment in real-world scenarios.Although the power output is generally lower in indoor light conditions, the performance of silicon solar cells can be more stable and robust.The operational principle of the presented e-Skin is based on the difference between the recorded voltage produced by each individual solar cell.Hence, any readable voltage produced by the device (i.e., > 50 μV per cell for the microcontroller used), should activate all presented detection features.The wear-out factors such as delamination and corrosion of metallic interconnection could also lead to degraded performance or failure of e-Skin.In the current study, we have used strong textile conductive tapes as interconnects as they offer enhanced stability under bending conditions.However, longterm (aging) studies are needed to analyze the effect of humidity and other environmental parameters on the textile interconnects and solar skin performance stability.

Application in Gesture Recognition
Finally, we have demonstrated that the solar skin integrated over the robotic arm could be used to perform hand gestures recognition in an energy-autonomous manner.We track the output voltage signal from the solar skin using a STM32 H7A3ZI-Q microcontroller which is sufficient for detecting hand gestures.Supporting information Movie#1 demonstrates that each gesture causes unique patterns in the ambient light levels.Using the solar skin, we digitized the shadow events for different hand gestures.The collected voltage patterns signal for each hand gestures were transferred to a computer using serial communication between the microcontroller and a Python script for data validation.For embedded system runtime, we have implemented a bespoke dense neural network algorithm training on lightweight datasets via backpropagation.Since the algorithm is running on a self-powered microcontroller, efficient and low-latency execution is paramount.One of the challenges we noted is related to the detection of gestures with high accuracy under low ambient light conditions (e.g., below 100 lux).Advanced machine learning methods such as autoencoding [64] or deep reinforcement learning [65] could be adopted for classifying complex gestures.Our future work will develop a simple binary classifier to identify these gestures.The key challenge is to keep the complexity of such algorithms within the constraints of low-power devices such as wearables with limited processing abilities.Another limitation can arise during scenarios where the light conditions change drastically, or when the light source is not stationary. [29]In such cases, the implemented sensing algorithms could result in false or incorrect detections.Some mitigation of these effects can be achieved by frequently running a calibration algorithm.For example, during a moving light source, the variation of light on various cells will be the same.This means the relative variation among solar cells will be negligible, and, hence, the solar skin will work normally.However, this may change in the presence of multiple moving light sources.Further processing of the output data of the solar skin could also help in this type of complex scenario.However, a more effective solution could be to integrate a microcontroller on each solar skin patch and run a small part of a dense neural network onto it. [66]Upon connecting multiple modules to each other, a more complex network architecture could be achieved, and the performance of the overall classification tasks could be greatly increased.To keep the computational cost low, and for this solution to be feasible, the voltage bias of the solar cells could be used as neural network weights, thus transforming the outside world inputs of the network into latent space hardware translations. [67]This procedure could also show potential for leveraging the sensing capabilities of different types of sensors integrated in the solar skin module, while maintaining the self-powered nature of this technology.Moreover, a single solar skin module could infer data from its surroundings, such as the distance from the light source or the angle of incidence of the light source, and use this information as another input in the network.For example, if the module can infer the distance from the light source, it would then be able to use this information to compute the relative position in space of the source of light, and then perform a self-powered mapping of its own environment. [68]Naturally, this procedure presents several limitations before it can be implemented.Data inference cannot be achieved with 100% accuracy and hence computing several other parameters based on inaccurate data will compound and propagate errors in the system.If the data inferred is not probabilistic, then it cannot represent new information. [69]owever, this effect could be minimized by creating a distributed network of solar skin devices and algorithmically validating the data inferred between them to arrive at a more precise result, thus reducing the overall compounding error in the system. [70]his new design paradigm will be explored in a future version of this solar skin module.

Conclusions
In this work, we have demonstrated the dual functionality of solar cells to develop a novel paradigm for flexible, conformable, energy-autonomous solar skin that is integrated into a robotic platform to endow it with capabilities for human-machine interaction applications.The developed solar skin was used for shadow sensing while the object was moving in horizontal and vertical directions with respect to the solar skin.In such scenarios, the proposed skin would be useful for improved safety and security for humanÀrobot coexistence and/or robot-robot co-existence for continuous operation of robots (thanks to its energy autonomy characteristics).Further, the solar skin was used for ambient temperature and light color detection.Finally, human finger gestures were monitored using the solar skin on robots.We note, however, that in the present design, we have used a larger and limited number of cells for sensing, which limits the spatial resolution.In the future, we will integrate multiple smaller and flexible solar cells into the solar skin to improve the architectural design while using the neural network approach to achieve higher accuracy gesture recognition.

Experimental Section
Solar Skin Design and Fabrication: First, the solar cells, Sol Expert 50 mm Â 50 mm monocrystalline, were laser cut to obtain smaller (50 mm Â 10 mm) sensing elements to construct the solar skin.Following the cutting process, the photovoltaic cells were attached to a paper substrate using electrically conductive and flexible textile tape (Holland Shielding Systems BV), thus creating the bendable e-Skin sensing system.
The solar cells were connected in series to enable us to measure the accumulated voltage of the sequence and provide a discrete series of outputs to be used for data inference.In this configuration, the module size can be easily scaled as all the output nodes follow this rule: This procedure theoretically allows us to achieve any desired final output voltage, while also obtaining a discrete uniform sequence of internal voltages from the module to perform data inference on.
Sensor Characterization: The system was then characterized using a remotely controlled white LED lamp (4 W) and the voltages generated by the solar cells were logged using a microcontroller (Stm32 Nucleo H7A3ZI-Q) with a 16bit ADC, giving a voltage readout resolution of 50.4 μV.Multiple tests were subsequently executed to characterize the sensor performance for the required application.For motion detection tests (horizontal and vertical), we have integrated an object on an industrial robot arm UR5 (Universal Robots, Odense, Denmark).Universal Robot Polyscope programming was used to control the arm movements including the speed of the movements while recording the voltage output in response to an object approaching the solar skin module.
Horizontal Motion Detection: In the horizontal motion case, the light source and the sensing module were kept in the same position.The object was moved at speeds of 1, 5, and 10 cm s À1 over the solar cells.
Vertical Motion Detection: For detecting vertical motion, the light source was kept fixed, 50 cm above the solar sensing module, and a light-obstructing object was moved vertically using a Universal Robot UR5, with a constant speed of 1, 5, and 10 cm s À1 .
Color Detection: In this experiment, we isolated the colors of the light shown on the solar module and measured the voltage produced by a single cell.This experiment was performed while the light source was kept at a fixed-point relative to the module (10 cm).
Temperature Detection: Temperature sensing characterizations of fabricated solar skin on the flexible paper substrate were performed in the ambient environment.For temperature sensitivity measurements, data was collected by interfacing the solar cell placed on a hot plate with a semiconductor parameter analyzer (B1500A, Agilent).

Figure 1 .
Figure 1.Multimodal solar skin performing dual functionality of energy generation and self-powered sensing on a robotic platform: a) schematic/optical image illustrating the dual functionality, b) circuit diagram of the solar skin module, and c) optical micrograph of the module.d-g) Schematics displaying energy autonomous sensing: d) ambient temperature detection, e) vertical motion detection, f ) horizontal motion detection, and g) visible light color detection.

Figure 2 .
Figure 2. The electrical characterization of solar skin during object's horizontal motion (shadow): a) schematic representation of the testing arrangements during horizontal motion sensing.b-d) Temporal output voltage patterns for object moving with varying speed when paced at a fixed distance of 10 cm between the light source and solar skin: b) 1 cm s À1 , c) 5 cm s À1 , and d) 10 cm s À1.It is to be noted that the given distance is from solar skin.e-g) Temporal output voltage patterns for object moving with varying speed when paced at a fixed distance of 20 cm between the light source and solar skin: e) 1 cm s À1 , f ) 5 cm s À1 , and g) 10 cm s À1 .h-i) Temporal output voltage curves summarizing the effect of speed when the object is placed at: h) 10 cm and i) 20 cm.

Figure 3 .
Figure 3.The electrical characterization of solar skin during vertical motion of object (shadow): a) schematic representation of the testing arrangements during vertical motion sensing.b-d) Temporal output voltage patterns for objects moving with varying speeds between the light source and solar skin in 3-40 cm distance: b) 1 cm s À1 , c) 5 cm s À1 , and d) 10 cm s À1 .It is to be noted that the given distance is from solar skin.

Figure 4 .
Figure 4. Temporal output voltage from various nodes of the solar skin when illuminated with different light colors.

Figure 5 .
Figure 5. Temperature sensing using solar skin: a) V OC at different times when the temperature varied between 35 and 95 °C, b) peak V OC -temperature graph showing linear change and sensitivity of À1.8 mV °CÀ1, c) micrograph from a commercial infrared camera of solar skin wrapped on a robotic platform showing the maximum solar skin temperature ≈26 °C, d) measured temperature of the solar skin using commercial infrared thermometer when exposed to hot air gun set at 100 degree, and e) real-time solar skin response when exposed to hot air gun.The data show higher V OC variation of solar cells when directly exposed to hot air (maximum temperature ≈45 °C).

Figure 6 .
Figure 6.Energy generation characteristics of fabricated solar skin under planar and bending conditions: a) optical image of the solar skin wrapped on a robotic arm, b) I-V measurements of the solar skin under dark conditions, c) I-V measurements under dark and illumination conditions.Under white light illumination, the distance between white light and solar skin was changed to vary the intensity of light.d) Power versus voltage curve under different intensities of white light.

Table 2 .
Calculated speeds based on the voltage drop of the 4 th solar cell in the skin module.