Instrument power cycling can be performed in several different ways. Rapid power cycling can reduce the average power draw and provide continuous measurements, at the cost of higher measurement uncertainty. Power cycling can also be performed at longer intervals, from seconds to minutes, allowing more time for the receiver to stabilize, but at the cost of continuous measurements. The use of a convolutional neural network (CNN) can provide the means to produce calibrated measurements in the presence of power cycling over both long and short intervals. CNN and other Artificial Neural Network (ANN) -based calibration approaches have been studied in recent years suggesting they produce calibration errors equal to or lower than those errors produced by using conventional calibration methods [1], [2]. With sufficient training data, the transient response characteristics of a receiver can be incorporated into a CNN model capable of producing calibrated measurements while an instrument is not in equilibrium. In this paper, a simulated radiometer model is used to produce calibration windows with transient response characteristics. A synthetic training dataset is produced to train a CNN which produces calibrated measurements. The CNN calibrator is evaluated against a linear least-squares (LSR) calibration method and shown to produce smaller calibration errors on a second independent synthetic dataset. By enabling the collection of quality calibrated measurements in the presence of rapid receiver power cycling, this technique can improve the power efficiency of space-borne radiometers by enabling measurement calibration prior to temperature stabilization, particularly for radiometer-equipped CubeSats and other SmallSats.

Future smart sensors will be able to utilize the maximum information content from data products, while minimizing the resources required to acquire, downlink, and process data. In-orbit calibration is required for space-borne radiometers in order to correct for gain fluctuations. Many sensors like radiometers are generally only able to produce calibrated scene measurements after reaching steady state. Waiting to reach thermal equilibrium to obtain useful data results in wasted power, excess useless data, and delays in obtaining useful data. Instrument power cycling provides a way to lower power use, but at the cost of pauses in data collection when the instrument is cycled off. Rapid power cycling can be used to reduce the average power draw of a radiometer, at the cost of increased measurement uncertainty. These power cycling techniques have been used on real systems, including the IceCube radiometer [1]. Using a convolutional neural network trained on synthetic data, a simulated radiometer can produce calibrated measurements with lower uncertainties and errors than conventional least-squares-regression (LSR) - based estimators. This approach presents an opportunity to reduce the average power draw of a radiometer by minimizing uncertainties of calibrated data products collected during rapid power cycling.

Recent advances in CubeSat technologies have enabled use of radiometers deployed in constellations of these small satellites for Earth and space science missions. Advantages of CubeSats such as their low cost, low mass and volume, and lower power requirements, however, are confronted by the challenges in calibration of their payloads as well as intercalibration of CubeSat constellations due to higher sensitivity to ambient conditions. This paper describes a novel system-level calibration framework, called “ACCURACy” to calibrate CubeSat based radiometer constellations as a single system in their entirety with minimal errors and uncertainties. Artificial constellation simulations have demonstrated that ACCURACy, while maintaining the accuracy levels of ideal calibration scenarios, leads to lower uncertainties in calibrated radiometer products compared to state-of-the-art calibration and intercalibration techniques based on overlapping measurements of the constellation members.

In the future, smart sensors will be designed to extract maximum value information, while minimizing the resources required to acquire, downlink, and process data. Many sensors like radiometers are only able to produce calibrated measurements after reaching steady state. However, waiting until reaching thermal equilibrium to obtain useful data leads to wasted power, excess useless data, and delays in obtaining useful data. Power cycling a radiometer is one way to circumvent this requirement, but leads to other challenges, as turning power off to instrument not only stops data acquisition until it is powered on, but also past power-on until it reaches thermal equilibrium again. This paper introduces a framework which will use machine learning algorithms to enable the calibration of a radiometer during its transient state after power-on and in the presence of power cycling, aiming to further reduce resource utilization.

Recent advances in space technologies enable science missions using CubeSats equipped with radiometers. Constellations of CubeSats can be used to significant effect, overcoming obstacles in cost, weight, and power. However, these benefits come at a cost, including challenges in calibration due in large part to increased sensitivity of the instrument to ambient conditions. These limitations also mean conventional calibration methods are not always possible. To address this problem, a novel, constellation-level calibration framework called “Adaptive Calibration of CubeSat Radiometer Constellations (ACCURACy)” is being developed. ACCURACy uses instrument-level telemetry data to cluster constellation members into time-adaptive groups of radiometers in similar states and facilitates calibration data sharing within each group for optimum calibration performance. This paper presents a prototype MA TLAB framework using synthetic radiometer data and discusses its calibration performance.

Recent technological developments have enabled usage of constellations of radiometer carrying CubeSats in scientific remote sensing missions. CubeSats, forming such constellations, on the other hand, bring unique challenges in terms of calibration of their instruments as they are easily impacted by ambient conditions. To address this problem, a constellation level calibration framework called “Adaptive Calibration of CUbesat RAdiometer Constellations (ACCURACy)” is introduced in this paper. The framework utilizes machine-learning algorithms such as principal component analysis and density based clustering to separate constellation members into time-adaptive groups of similar-state radiometers based on their telemetry data. Within each group, all radiometers will contribute to a calibration data pool with their absolute calibration measurements. Such shared data pools, which include measurements of different calibration targets at different times, will facilitate frequent N>2-point absolute calibration; thus, reduce and quantify calibration errors and uncertainties.

Although considered as stationary and Gaussian in general, radiometer gain is usually a fluctuating signal with non-stationary properties. Analyses of such non-stationary features is challenging as the radiometer signal cannot be observed independently. On the other hand, time series of post-gain voltages constitute an ensemble set for the radiometer gain which can be used to characterize the radiometer gain. This paper presents a novel technique called "Ensemble Detection" which can analytically retrieve the standard deviation of stationary Gaussian radiometer gain or find an equivalent stationary Gaussian process which represents the non-stationary radiometer gain under different calibration schemes. It has been found that the equivalent Gaussian process for non-stationary radiometer gain heavily depends on the calibration structure and the observation times of the measurand and the calibration references.