Development and Evaluation of Regolith Mass Estimation Sensor Based on Photoresist Effect

Abstract

This paper presents the design, implementation, and laboratory validation of an optoelectronic-based mass estimation sensor for regolith sampling devices. The sensor integrates multiple photoresistors into the walls of a shovel of a sampling device, where the sensors detect varying levels of light occlusion caused by the deposited regolith. By analyzing the output signals from these photoresistors, the sensor estimates the mass of the sampled regolith. The device is designed to handle a typical sample mass range of 100–300 g. Laboratory tests demonstrated that the sensor can estimate the regolith mass with a relative error of approximately 23%, which is suitable for early-stage applications where rapid, non-invasive mass estimation is essential. The shown level of accuracy underscores the potential for further refining the calibration process, enhancing sensor sensitivity, and integrating multi-sensor approaches to improve performance. This conceptual study highlights the feasibility of using optoelectronic sensors for regolith mass estimation, paving the way for future innovations in ISRU missions and other granular material sampling applications. Future work will focus on the optimization of photoresistor placements, refining the calibration process, and enhancing sensor sensitivity to improve the accuracy of mass estimation.

1. Introduction

In the context of In Situ Resource Utilization (ISRU) projects, efficient regolith sampling is crucial for planetary exploration and resource extraction missions [1]. Regolith sampling devices (RSDs) are tasked with collecting soil or other granular material from planetary surfaces [2,3], but the ability to accurately estimate the collected mass in situ remains a technical challenge. Accurate in situ mass estimation is essential for optimizing material collection efficiency in ISRU missions and ensuring that sufficient samples are collected for scientific analyses. This capability helps prevent overloading or under-collecting, thereby improving mission reliability and resource allocation. Traditionally, strain gauge-based and piezoelectric force sensors are employed to measure the mass of the sampled regolith through force comparison methods.

However, several factors limit the feasibility of using strain gauge sensors in extraterrestrial environments, including their sensitivity to environmental conditions and difficulty meeting stringent uncertainty requirements.

Firstly, the reduced gravity on celestial bodies like the Moon or Mars significantly diminishes the measurable force exerted by the regolith mass on the sensor. This challenge is even more pronounced for missions like OSIRIS-REx or Martian Moon Exploration, which target the smallest celestial bodies [4,5,6]. This makes it difficult to achieve the desired measurement accuracy, set for the RSD to be at least 25 g. Secondly, in the context of force sensors, accuracy is affected by several factors, including nonlinearity, hysteresis, and inter-channel coupling [7]. These sources of error generally contribute to an overall uncertainty level of at least 0.01% of the full measurement scale, with some configurations reaching up to 3%, depending on the application and sensor design. While this level of precision can be sufficient for many terrestrial applications, it poses a challenge for space missions with stringent requirements. During the development of the RSD, we determined that the mass sensor would need an accuracy level as low as 0.0001% of the full scale to meet operational demands, especially under conditions requiring the sensor to endure launch loads without an HDRM. Achieving such an extraordinarily low margin of error with existing sensor technology is currently unfeasible.

Finally, sampling devices are often deployed as end-effectors on manipulator arms or other deployment mechanisms. Due to strict mass and volume constraints inherent in space missions, there is limited space for incorporating Hold-Down Release Mechanisms (HDRMs), which are essential to reduce launch loads for traditional applications equipped with force sensors. Therefore, there are two opposite requirements: first to withstand substantial launch loads—up to 65 GRMS (root mean square acceleration)—necessitating a stiff interface; second: a need for compliance to detect small mass variations.

These opposing requirements—having a stiff and robust design to survive launch loads while also achieving high measurement accuracy—pose a significant engineering challenge, leading to the development of an alternative solution: the exploration of optoelectronic sensors, specifically photoresistors, integrated into the walls of the sampling shovel. The proposed sensor detects the degree of light occlusion caused by the collected regolith. By analyzing the signals from these photoresistors, which vary depending on whether the sensors are fully covered, partially covered, or not covered at all by the regolith, the sensor estimates the mass of the sampled material. A pre-determined calibration function, derived from controlled laboratory experiments, is employed to convert the sensor readings into a mass estimate.

This conceptual study focuses on the feasibility of a low-Technology Readiness Level (TRL) approach for regolith mass estimation. The primary objective was to develop and validate a simplified system that could provide initial proof-of-concept results within the ±25% accuracy margin specified by ESA in a project focused on RCE development [8], acknowledging that further development will be necessary for operational readiness.

It is essential to note that the performance of the system was evaluated under controlled laboratory conditions and did not encompass additional durability or environmental testing. While the study demonstrated the conceptual approach, factors such as dust accumulation, variable lighting conditions, and bulk density changes in regolith could affect the overall accuracy.

The remainder of this paper is organized as follows. Section 2 describes the materials and methods used in this study, detailing the construction and operation of the optoelectronic mass estimator, including the selection of photoresistors, mechanical integration, electrical circuit design, parameters of the regolith analog used and the calibration process using a multidimensional polynomial fitting approach in MATLAB R2023b. Section 3 presents the results from two measurement series: the first utilized to obtain the calibration coefficients and the second conducted to evaluate the sensor’s performance, including the estimation process and accuracy assessment using the Percentage Relative Difference method. Section 4 discusses the implications of these findings, limitations of the current conceptual design, and recommendations for future research, emphasizing that this work represents an early-stage, low-TRL solution that will require further development to reach operational readiness.

By presenting this alternative approach, we aim to contribute to the broader discourse on mass estimation technologies suitable for ISRU applications and planetary exploration, demonstrating that even low-TRL solutions can offer promising pathways for future development.

2. Materials and Methods

This section outlines the design and operational principles of a mass estimation sensor that employs photoresistors to measure the mass of regolith samples as well as the design of the calibration function. The sensor operates by detecting variations in light intensity, which occur due to the partial coverage of photoresistors mounted on the side of the RSD clamp during sampling.

2.1. Photoresistor Selection

For the construction of the mass estimation sensor, GL5516 photoresistors were selected due to their small size and favorable resistance-luminosity characteristics. The GL5516 photoresistors have a diameter of 5.1 mm and a height of 2.1 mm (excluding connectors), making them an ideal fit for integration into the side walls of the RSD clamp. The detailed dimensional specifications and resistance-luminosity characteristics of these photoresistors are presented in Figure 1. Their ability to reliably change resistance with varying light intensities is critical for the sensor’s functionality.

2.2. Mechanical Design

The mechanical design of the sensor was optimized to recently develop an engineering model of the Rotary Clamshell Excavator [8]. Five photoresistors were placed within the internal side of the wall of the RCE’s shovel. These photoresistors are mounted in 5.2 mm diameter holes drilled along a symmetrical radius of the shovel wall, facing the interior where the regolith is collected. As shown in the Figure 2, the holes are spaced 10 mm apart, with the first hole positioned 8 mm from the external edge of the wall. This arrangement ensures that when regolith is sampled and deposited inside the clamp, it partially covers the lower photoresistors, thereby reducing the light reaching them and causing a measurable change in their resistance.

2.3. Electrical Circuit

The electrical design of the sensor employs a voltage divider configuration to accurately measure the changes in light intensity detected by the photoresistors. Each of the five photoresistors is connected in series with a fixed 10 kΩ resistor, forming individual voltage dividers. This arrangement is powered by a stabilized 10 V supply, as shown in Figure 3. As the resistance of a photoresistor changes in response to varying light levels—caused by the regolith partially covering the sensor—the voltage drop across the photoresistor also changes.

The voltage drop over each photoresistor is directly measured using an Analog-to-Digital Converter (ADC). By measuring this voltage, the sensor can determine the resistance of the photoresistor at any given moment, which correlates with the amount of light reaching the sensor. The more regolith that covers the photoresistor, the less light it receives, leading to a higher resistance and a corresponding change in the voltage drop.

This method allows for precise detection of light intensity variations without the need for complex circuitry. The voltage readings from the ADC are then used in conjunction with a pre-determined calibration function to estimate the mass of the sampled regolith. This approach simplifies the electrical design while providing the necessary sensitivity to detect changes in the regolith mass within the sampling device.

2.4. Lunar Regolith Analog

AGK-2010 was created as a lunar soil analog by the Department of Drilling and Geoengineering at the AGH University of Science and Technology in collaboration with the Space Research Centre of the Polish Academy of Sciences [10]. This analog was developed to minimize costs while maintaining fidelity to the properties of authentic lunar regolith, using CHENOBI and JSC as reference standards. These American analogs are recognized for their resemblance to lunar regolith samples collected during the Apollo 17 mission.

AGK-2010 exhibits a grain-size distribution comparable to that of CHENOBI, as determined using standardized sieve methods. This distribution ensures that AGK-2010 represents the granular structure typical of lunar regolith, which includes fine particles with varying sizes. The bulk density in a loose state of AGK-2010 is approximately 1.295 g/cm³, only 1.45% lower than that of CHENOBI (1.314 g/cm³).

The internal friction angle and cohesion values of AGK-2010 align closely with those of CHENOBI. AGK-2010 displays a friction angle of 37.67° and cohesion of 3.85 kPa, compared to CHENOBI’s 37.27° and 3.78 kPa, respectively. These mechanical similarities ensure that AGK-2010 can adequately simulate the shear strength and stability of lunar regolith.

An important aspect of lunar regolith is the sharp-edged nature of its mineral grains, which contributes to its unique behavior under mechanical stress. AGK-2010 was designed to replicate this feature, ensuring that it accurately reflects the regolith’s abrasive and cohesive properties, which is significant for realistic testing scenarios.

2.5. Operational Scenario

The calibration was performed under controlled laboratory conditions to simulate the actual operating environment of the regolith sampling device (RSD). During the calibration, the RSD’s shovel acquired samples of regolith AGK-2010 while the voltage drop across each photoresistor was continuously recorded. To maintain consistency and minimize external light interference, the ambient lighting in the laboratory was dimmed. A stationary LED diode was employed to uniformly illuminate the acquisition area, providing a consistent light source for the photoresistors.

As shown in Figure 4, while the shovel rotates during the sampling operation, the photoresistors mounted on its side wall traverse through regions of varying illumination caused by the shovel’s mechanical constraints. These constraints create distinct lit and shadowed areas within the shovel’s interior. The placement of the LED light source was carefully adjusted to illuminate the area where the photoresistors come to rest after the acquisition cycle. This setup ensures that any changes in the voltage readings of the photoresistors are predominantly due to the presence of regolith covering the sensors rather than fluctuations in ambient lighting. The examples of surface profiles of sampled regolith are shown in Figure 5.

During baseline operations—when the shovel rotates without acquiring regolith—the photoresistors move through these lit and shadowed regions, resulting in characteristic voltage signal patterns. These patterns are evident in the recorded voltage signals and are crucial for understanding the photoresistors’ responses under normal operating conditions. An example of the voltage signals from all photoresistors during baseline operation is presented in Figure A1 in Appendix A.

2.6. The Depiction of Light Inside the Shovel

The shovel’s mechanical design inherently creates regions of light and shadow due to its geometry and movement during operation as shown in Figure 4. The stationary LED light source was strategically positioned to illuminate the area where the photoresistors are located at the end of the shovel’s rotation. This positioning ensures that, after the shovel stops moving, the photoresistors are exposed to consistent lighting conditions, which is essential for accurate voltage measurements.

As the shovel rotates, the photoresistors pass through varying light conditions. Initially, they may be in shadowed regions, but as the shovel moves, they enter lit areas illuminated by the LED. This transition results in observable changes in the voltage signals recorded from each photoresistor. The voltage decreases when the photoresistors are exposed to light and increases when they are in shadow or covered by regolith. Figure A2, Figure A3, Figure A4, Figure A5 and Figure A6 in Appendix A illustrate the voltage signals from all photoresistors during the shovel’s rotation without regolith, highlighting the effect of light and shadow regions on the sensor outputs.

Understanding these light patterns is vital for interpreting the voltage signals accurately. It allows for differentiation between voltage changes due to mechanical movement and those resulting from regolith coverage. This understanding forms the basis for developing a reliable calibration function that correlates voltage readings to the mass of the sampled regolith.

2.7. The Determination of the Calibration Function

The following section outlines the experimental approach used to derive the calibration function, which relates the photoresistor voltage readings to the actual mass of the regolith samples. A series of measurements were conducted where the RSD acquired nine different regolith samples of varying masses. For each acquisition, both the mass of the sample and the voltage signals from each photoresistor were recorded.

The voltage signals from each photoresistor during these acquisitions are depicted in Appendix A. Analysis of these signals revealed that the most informative data for mass estimation are the final voltage values obtained after the shovel has completed its rotation and come to a stop. At this point, the photoresistors are in a stable position and the lighting conditions are consistent, making the voltage readings more reliable.

To minimize the influence of environmental light variations and reflections, the ratio of the initial voltage (before acquisition) to the final voltage (after acquisition) was calculated for each photoresistor. This normalization process reduces the impact of any fluctuations in ambient light and focuses on the relative change caused by rotation of the shovel and the presence of regolith covering the sensors. Therefore:

$$U_{Pn}={\displaystyle \frac{U_{Pn final}}{U_{Pn initial}}},$$

(1)

where U_Pn—voltage ratio from photoresistor n; U_{Pn final}—final voltage value read from photoresistor n; and U_{Pn initial}—initial voltage value read from photoresistor n.

During the analysis, it was observed that the photoresistor designated as P1, located closest to the shovel’s rotational axis, exhibited no significant correlation with the acquired regolith mass. This lack of correlation is attributed to P1 never being fully covered by regolith, even during larger acquisitions. Instead, regolith tends to build up in front of the shovel, casting a shadow inside but not directly affecting P1. Consequently, P1′s voltage readings were inconsistent and did not contribute meaningfully to mass estimation.

To address this issue, an artificial variable was introduced to effectively exclude P1 from the calibration without discarding valuable data from the other sensors. This variable, defined as U_p1 = U_p2 + U_p3 + U_p4 + U_p5, aggregates the normalized voltage ratios of the four remaining photoresistors, all of which showed a meaningful correlation with the regolith mass.

Using MATLAB R2023b, a multidimensional polynomial fitting approach was employed to model the relationship between voltage values and the sample mass. The least squares method was used to fit a second-degree polynomial to the data, resulting in the following calibration function:

$$\begin{gathered} Mass=c_1+c_2·U_{p1}+c_3·U_{p2}+c_4·U_{p3}+c_5·U_{p4}+c_6·U_{p5}+c_7·{U_{p1}}^2+c_8 \\ ·{U_{p2}}^2+c_9·{U_{p3}}^2+c_{10}·{U_{p4}}^2+c_{11}·{U_{p5}}^2 \end{gathered}$$

(2)

This calibration function provides a mathematical model to estimate the mass of the sampled regolith based on the aggregated and normalized voltage readings from the photoresistors. By focusing on the final voltage values and normalizing them, the calibration accounts for any environmental variations and enhances the accuracy of the mass estimation.

The polynomial degree of 2 was chosen to capture any nonlinear relationships between the voltage readings and the regolith mass. The use of a multidimensional polynomial fitting approach allows for a more precise modeling of the data, improving the reliability of the mass estimates provided by the sensor.

3. Results

Two measurement data series were conducted to calibrate the mass estimation sensor and evaluate its performance. The first series aimed to obtain the calibration coefficients necessary for developing the mass estimation model. The second series was designed to acquire data that would allow for the evaluation of the sensor’s performance and validate the model’s predictive capability.

3.1. Calibration Data Acquisition

The first measurement series consists of nine regolith sampling operations. The signal read from photoresistors as well as mass of collected sample were saved and used to calculate the calibration. The values of the calibration coefficients obtained are presented in

Table 1. Calibration coefficients’ values.

Coefficient	c1	c2	c3	c4	c5	c6
Value	−1.7721	−6.4091	9.6412	−1.9357	−0.2334	0.8492
Coefficient	c7	c8	c9	c10	c11
Value	0.3697	−0.5221	−0.0013	0.0076	−0.0463

.

3.2. Evaluation Data Acquisition

Following the calibration process, a second series of measurements was conducted to evaluate the performance of the mass estimation model derived earlier. This series aimed to validate the sensor’s ability to accurately estimate the mass of regolith samples under the same controlled conditions as the calibration but with new data points to test the model’s predictive capability.

A total of 20 additional sampling operations were performed using the RSD. During each operation, the shovel acquired a regolith sample while the voltage signals from the photoresistors were continuously recorded. The same controlled lighting conditions were maintained—the laboratory lights were dimmed, and the acquisition area was illuminated using the stationary LED diode to ensure consistency with the calibration setup.

For each sample, the actual mass of the regolith was measured using a scale immediately after acquisition. This provided a reference for comparing the estimated masses calculated using the calibration function.

3.3. Mass Estimation

The mass of each regolith sample was estimated using the calibration equation derived from the previous calibration process. The normalized final voltage ratios for each photoresistor were calculated, focusing on the values obtained after the shovel had come to a complete stop.

As previously determined, the photoresistor P1 did not contribute meaningfully to the mass estimation due to its placement and lack of coverage by the regolith. To account for this, the artificial variable U_p1 = U_p2 + U_p3 + U_p4 + U_p5 was utilized.

Using the calibration function—a second-degree polynomial fitted using the least squares method in MATLAB—the estimated mass m_e for each sample was calculated.

3.4. Evaluation of Sensor Performance

The estimated masses were then compared to the actual masses measured. Figure 6 illustrates the estimated mass plotted alongside the actual acquired mass for all 20 samples, facilitating an easy comparison between the two datasets. Figure 7 presents the relative error visualization, with a distinction between overestimation and underestimation of the sample mass.

3.5. Accuracy Assessment

To quantitatively assess the accuracy of the mass estimation, the Percentage Relative Difference (PRD) method was employed. The PRD was calculated using the following formula:

$$PDR={\displaystyle \frac{1}{n}}\sum \left|{\displaystyle \frac{m_r-m_e}{m_r}}\right|·100\%=22.6\%$$

(3)

where m_r—actual mass and m_e—estimated mass.

Additionally, standard deviation was also determined with the formula:

$$ST{D}_{DEV}=std\left(\left|{\displaystyle \frac{m_r-m_e}{m_r}}\right|·100\%\right)=19.2\%$$

(4)

The average PRD across all 20 samples was found to be 22.6%, indicating the relative accuracy of the mass estimator. This level of accuracy is compliant with the required ±25% target specified by ESA’s Statement of Work.

4. Discussion

The validation results demonstrate that the optoelectronic mass estimation sensor can estimate the mass of regolith samples with a reasonable degree of accuracy. The 22.6% relative error aligns with the ±25% target specified by ESA’s Statement of Work, which is sufficient for applications where rapid and approximate mass estimation is required, such as detecting overfilling or ensuring adequate sample collection. However, we recognize that this level of accuracy may not be sufficient for all applications, particularly those requiring precise sample mass measurements for detailed scientific analysis.

Several factors could contribute to the observed estimation errors:

• Variability in regolith properties: Differences in particle size, shape, and compaction can affect how the regolith covers the photoresistors, thereby influencing voltage readings. Variations in bulk density could lead to increased errors, effectively causing the sensor to estimate the captured volume rather than the mass. We have emphasized that the calibration process assumed consistent material properties, which may not always be present in operational scenarios.
• External light influence: Despite controlled lighting conditions, minor fluctuations or reflections could impact the photoresistor signals. Real-life environments, such as lunar or Martian surfaces, would introduce additional challenges like irregular lighting, shadows, and dust accumulation. We acknowledge that these factors could impact accuracy and have included a discussion on potential future enhancements, such as using inter-shovel light sources and incorporating dust-resistant coatings or cleaning mechanisms.
• Sensor limitations: The resolution and sensitivity of the photoresistors and the ADC may limit the accuracy of voltage measurements. Durability concerns, especially in abrasive regolith environments, could also affect long-term sensor performance. We did not conduct durability testing in this initial study. Protective measures, such as abrasion-resistant coatings and recessed sensor housings, are essential areas for future development.
• Dust accumulation: Accumulated dust on the sensors can impact accuracy. We identified two scenarios where dust could interfere with measurements: dust accumulation between acquisitions, which was accounted for with a reference measurement, and dust covering sensors during the digging process, which could lead to overestimations. Addressing this issue with design solutions, such as periodic cleaning mechanisms or electrodynamic dust shields, will be a focus in further iterations.
• Calibration: Both the chosen calibration function as well as the minor size of calibration dataset used to determine calibration coefficients have significant impact on mass sensor accuracy.

To improve the sensor’s accuracy, the following approaches could be considered:

• Enhanced calibration models: Incorporating higher-degree polynomials or additional variables that account for environmental factors may refine the calibration function. Simulations using the Discrete Element Method (DEM) could assist in determining the optimal form of the calibration function. The acquisition of a larger dataset used for calibration should also positively impact the accuracy of the fitted function.
• Regolith density variability: During the development of the RSD, we observed that regolith sampled by the excavating device tends to become loose, leading to a decrease in bulk density. Measuring the resulting density and investigating its impact on the final mass estimation error could provide valuable insights for refining the calibration function.
• Incorporating data during digging phase: Focusing solely on initial and final voltage readings may overlook valuable data captured during the shovel’s entry into the regolith. Analyzing voltage changes during this phase could provide additional insights and improve overall accuracy.
• Sensor improvements: using photoresistors with higher sensitivity or integrating secondary sensing modalities, such as radio frequency-based sensors, could enhance the robustness and accuracy of the system.
• Light control: Optimizing the light source and effectively isolating the photoresistive sensors from external light can enhance the robustness and accuracy of the system. Such isolation could be achieved by investigating effective light covers for the sampling device, utilizing unique light parameters such as polarity, or generating light pulses at a known frequency and applying a band-pass filter in the signal acquisition module.
• Signal processing techniques: applying advanced filtering or signal analysis methods may help reduce noise and improve the reliability of voltage measurements.
• Optimization of sensor placement: The determination of the area of the shovel where the regolith falls most repeatedly should improve the reliability of the collected signal. Optimizing the placement of photoresistors within the shovel to ensure consistent regolith coverage is crucial for improving measurement reliability.
• Environmental influence: Advancing the sensor to higher TRL levels will require addressing real-life environmental factors expected at the deployment site but not covered in this study. These include electrostatic adhesion of regolith particles, temperature dependencies affecting the performance of photoresistive sensors, cosmic radiation, and other environmental variables.

Overall, this study demonstrates the feasibility of a low-TRL optoelectronic sensor for regolith mass estimation. While the current design shows promise for preliminary applications, future development should address sensor placement, enhanced calibration, and integration of multi-sensor systems to improve robustness and accuracy, especially for real-world extraterrestrial missions.