Methods for Assessing and Optimizing Solar Orientation by Non-Planar Sensor Arrays

Jiang Wang, Xingang Fan, Yongchao Zhang, Jianyu Yang, Yuming Du, Jianxin He2019 · DOI

TL;DR

This paper introduces a novel mathematical framework for assessing and optimizing the accuracy of solar orientation determination using non-planar sensor arrays. By developing a precise formulation for orientation error and its supremum, the authors provide a robust method for evaluating and designing sensor arrays that outperform existing approaches, particularly for space applications.

What Problem Does This Paper Address?

Accurately determining the Sun's orientation is crucial for a wide range of applications, from spacecraft attitude estimation and planetary rover navigation to optimizing solar power plants. Traditional methods often rely on planar sensor arrays, which can require multiple sensors and increase payload mass, a significant concern for small satellites. Non-planar sensor arrays, utilizing photodiodes on various surfaces, offer a more compact solution but have historically suffered from susceptibility to interference and a lack of rigorous methods for assessing and optimizing their orientation accuracy. Existing approaches to optimize sensor array configurations often focus on minimizing estimated variance or bias, but a unified, assessment-based approach to evaluate the fundamental orientation error and its supremum has been missing. This paper aims to fill that gap by providing a theoretical foundation for understanding and improving the performance of non-planar solar sensor arrays.

Methods

The core of this paper lies in its mathematical formulation of orientation error. The authors first establish a relationship between the measured irradiance and the Sun vector, accounting for sensor characteristics. They then introduce an interference vector ($\epsilon$) that corrupts the measurement. A key innovation is the derivation of the orientation error ($\Delta r$) as a function of the orientation matrix ($H$) and the interference vector. Crucially, they decompose the interference vector into components within and orthogonal to the sensor array's measurement space, showing that only the component within the space affects the error.

This leads to the definition of the orientation error angle ($\theta$) and its maximum possible value, the supremum ($\theta_{sup}$). The paper then defines the "full impact supremum" ($\theta_{sup\_fi}$), which represents the worst-case orientation error under full interference projection. This $\theta_{sup\_fi}$ is shown to be inversely proportional to the minimum non-zero singular value ($\sigma_{min}$) of the orientation matrix.

To assess performance, the authors introduce an "interference coefficient" ($\kappa = 1/\sigma_{min}$) for finite interference energy and an "average interference coefficient" ($\kappa_a$) for finite average interference energy. A smaller coefficient indicates better performance. For optimization, the goal becomes selecting an orientation matrix (a combination of sensor planes) that minimizes these coefficients. The paper proposes that optimal sensor arrays should have all sensor planes illuminated and their orientation matrix should have equal non-zero singular values.

Datasets

The paper utilizes simulated data for evaluating its methods. It presents five different sensor array configurations (Arrays 1-5) with varying sensor plane arrangements. These configurations are then used in simulations to compare the proposed optimization method against those based on minimum estimated variance and bias. A field experiment was also conducted using a custom-built sensor array with 16 solar panels mounted on a regular 16-pyramid.

Key Insights and Ideas

  • Unified Error Formulation: The paper provides a comprehensive mathematical framework to quantify orientation error, directly linking it to the sensor array's geometry (via the orientation matrix) and the nature of interference.
  • Supremum as a Performance Metric: Defining the "supremum" of orientation error offers a worst-case performance guarantee, which is more robust for critical applications like spacecraft attitude determination than average error metrics.
  • Interference Coefficients: The introduction of $\kappa$ and $\kappa_a$ provides intuitive metrics for assessing sensor array performance based on their geometric configuration, independent of the specific interference magnitude.
  • Optimal Array Design Principles: The findings suggest that maximizing the number of illuminated sensors and ensuring equal non-zero singular values in the orientation matrix are key to achieving optimal performance.
  • Assessment vs. Optimization: The paper clearly distinguishes between assessing the performance of an existing array and optimizing its design, offering methods for both.

Key Results

  • The proposed method for assessing orientation error and its supremum was validated through simulations and a field experiment.
  • Simulations showed that sensor arrays designed using the proposed method (minimizing $\theta_{sup\_fi}$) outperformed those designed based on minimum estimated variance or bias. For instance, Arrays 2 and 3, optimized using the proposed method, consistently showed smaller orientation errors than Arrays 4 and 5.
  • The field experiment demonstrated that smaller interference coefficients ($\kappa$) directly correlated with smaller orientation errors. Specifically, an orientation matrix with $\kappa_{min} = 0.8$ yielded the smallest orientation errors compared to matrices with larger $\kappa$ values.
  • The study confirmed that in scenarios with finite average interference energy, the performance is primarily determined by the average interference coefficient ($\kappa_a$), and among those with the same $\kappa_a$, the one with the smaller $\kappa$ performs better.

Limitations and Future Work

  • Interference Model: While the paper models interference, the exact nature and distribution of interference in real-world scenarios can be complex and may require further refinement. The assumption of interference being projected onto the column space of the orientation matrix is a simplification.
  • Sensor Calibration: The paper assumes similar sensors with a constant measurement coefficient ($\eta$). In practice, sensor calibration and individual sensor errors could introduce additional complexities.
  • Computational Complexity: While the paper focuses on theoretical formulation and optimization principles, the practical implementation of finding the optimal sub-matrix for a large number of sensors might involve significant computational effort.
  • Future Work: The authors suggest further research into more complex interference models and the integration of sensor calibration into the optimization process. Exploring the application of these methods to different types of non-planar sensor arrays and real-world mission scenarios would also be valuable.

Summary updated

BibTeX
@misc{wang2019methods,
  author = {Jiang Wang and Xingang Fan and Yongchao Zhang and Jianyu Yang and Yuming Du and Jianxin He},
  title = {{Methods for Assessing and Optimizing Solar Orientation by Non-Planar Sensor Arrays}},
  year = {2019},
  doi = {10.3390/s19112561},
}

Loading related content...

Figure