Traditional planning algorithms used in satellite constellations and other sensor webs typically assume a static reward grid where observing each ground point provides a constant and typically uniform value. However, remote sensing of dynamic events on Earth’s surface requires a dynamic reward grid. This in turn requires a reactive planning approach that replans as needed when the reward grid changes. In this paper, we present a simple reactive planning strategy that builds upon existing work in the satellite observation planning field, and we examine its ability to observe dynamic events on Earth’s surface with a heterogeneous satellite constellation. We compare the approach to traditional non-reactive planning strategies. We also test the sensitivity of the reactive planning approach to two parameters - the timeliness of the information about events occurring, and the number of observation types required to complete a heterogeneous observation.

Human modification of the landscape affects total suspended solids (TSS) concentrations in water. The quantitative extent of these changes remains poorly understood, partly because of the challenges associated with observing TSS dynamics in inland waters over large scales. While many current missions and sensors provide usable data to estimate inland water quality (e.g. Landsat series, VIIRS, Sentinel-2), future missions present the opportunity to increase transferability and accuracy of TSS estimation. Here we degrade assumed ideal spectral data to evaluate the optimal data quality for TSS retrieval using an optical sensor configuration. We also perform wavelet analysis and a river size distribution analysis to study temporal and spatial data quantity requirements, respectively. We find that while the highest resolution data always gives the best retrieval accuracy, some factors are more essential in TSS estimation than others and can simplify mission design. Specifically, fine hyperspectral resolution is key in improving retrieval accuracy and a finer spatial resolution allows exponentially more river surface area to be observed. A revisit period of approximately 5 days or less best captures TSS pulse events, such as floods. Understanding the optimal mission specifications for observing inland water quality, especially TSS, will assist in developing and proposing future optical satellite missions.

This work explores a new concept of operations for observation of Earth events by a satellite sensor web that is able to reason about its capabilities and plan observations based on detected or requested events on Earth’s surface. An intelligent agile satellite sensor web is shown to produce more than twice the number of observations of a nadir-looking sensor web.

Monitoring of short-lived and highly-dynamic processes and events such as floods or forest fires has gained an increasing interest in Earth Observation, particularly as global climate change is affecting these processes. The observation of such dynamic events is often limited by the response time of human operation of Earth-Observing satellites or UAVs.To address this bottleneck, this paper presents a Modified Asynchronous Consensus Constraint-Based Bundle Algorithm (MACCBBA) for observation task allocation in Sensor Web mission concepts for Earth Observation. This algorithm allows for the decentralized allocation of observation tasks amongst a network of Satellites and UAVs based on recently measured data processed on board, or on messages received from other sensors or from the ground. This algorithm is also capable of reaching a feasible plan in a dynamic communications network such as the ones present in some Sensor Web mission concepts and allows for complex temporal constraints and dependencies between tasks to model the value of near-simultaneous co-observations by complementary or synergistic sensors.

In this paper we develop a Convolutional Long Short-term memory (ConvLSTM) model, a time series deep learning neural network, to predict soil moisture, with an add-on module of passive microwave (radiometer) soil moisture retrieval using the Tau omega model. We incorporate antecedent observations, landscape properties, and forcing factors such as precipitation, landcover, clay fraction, and brightness temperature in the prediction scheme. A regularization Monte Carlo Dropout layer is added to the network to remove stochasticity and avoid overfitting during the training phase. This dropout layer also provides a Bayesian approximation to quantify uncertainty during forecasting. The model is validated at four Soil Moisture Active Passive (SMAP) Cal/Val locations using performance metrics such as Root Mean Square Error (RMSE) and Bias to evaluate effectiveness of the proposed method. This model is developed as a component of the Science Simulator within the Distributed Spacecraft with Heuristic Intelligence to Enable Logistical Decisions (D-SHIELD) project.

This article explores the tradespace for a constellation of heterogeneous smart satellites intended to measure soil moisture using a combination of L and P band radars, radiometers, and reflectometers. Orbit inclination, repeat cycle, number of satellites, and number of planes were treated as input variables to create a set of architectures for evaluation. Attempting to optimize multiple output variables (cost, average revisit time, maximum revisit time, and percent coverage) results in a complex tradespace with suitable options at various cost caps. Therefore, several cost ranges are examined to find the best constellation for a given cost cap. It was found that a relatively simple constellation of three satellites in one plane offers acceptable performance at a low cost. This preliminary submission shows results for a homogeneous constellation, while the final paper will include satellites with various instrument configurations.

We describe a suite of scalable software methods and frameworks to helps schedule payload operations of a large constellation, with multiple payloads per and across spacecraft, such that the collection of observational data and their downlink, constrained by the constellation constraints (orbital mechanics), resources (e.g., power) and subsystems (e.g., attitude control), results in maximum science value for a selected use case. Constellation topology, spacecraft and ground network characteristics can be imported from design tools or existing constellations and can serve as elements of an operations design tool. Our framework includes a science simulator to inform the scheduler of the predictive value of observations or operational decisions. Autonomous, realtime re-scheduling based on past observations needs improved data assimilation methods within the simulator.

This paper presents a method to effectively adapt the baseline of a synthetic aperture radar based on Deep Reinforcement Learning in distributed Earth observation missions. We describe the approach, which uses the Proximal Policy Optimization algorithm and provides initial results for a toy example built around a hypothetical mission to measure the vertical structure of forests using a formation of 7 satellites carrying L-band synthetic aperture radars. We demonstrate that using a reward function based on expected science return over time and fuel usage; our Deep Reinforcement Learning planner can create plans with positive scientific returns while minimizing fuel usage.

D-SHIELD is a suite of scalable software tools that helps schedule payload operations of a large constellation, with multiple payloads per and across spacecraft, such that the collection of observational data and their downlink, constrained by the constellation constraints (orbital mechanics), resources (e.g., power) and subsystems (e.g., attitude control), results in maximum science value for a selected use case. Constellation topology, spacecraft and ground network characteristics can be imported from design tools or existing constellations and can serve as elements of an operations design tool. D-SHIELD will include a science simulator to inform the scheduler of the predictive value of observations or operational decisions. Autonomous, realtime re-scheduling based on past observations needs improved data assimilation methods within the simulator.

Both ESA/Eumetsat and NASA/NOAA have had a GEO and a polar component of space-based operational weather monitoring for decades. However, constellations of a few dozen small satellites in LEO can now provide sub-hour revisit times similar to those achieved from GEO and better spatial resolution than GEO for the same cost as a large GEO satellite. On the other hand, the large instruments typically used for GEO-based monitoring must be broken down into components in order to be allocated to smaller satellites. This paper describes a methodology to systematically study this large and rich architecture space and discuss under what circumstances and for what purposes a GEO component may still be superior to LEO. The focus is on coverage-cost tradeoffs, whereas other important considerations are left for future work.

Adaptive operator selection (AOS) is a high-level controller for an optimization algorithm that monitors the performance of a set of operators with a credit assignment strategy and adaptively applies the high performing operators with an operator selection strategy. AOS can improve the overall performance of an optimization algorithm across a wide range of problems, and it has shown promise on single-objective problems where defining an appropriate credit assignment that assesses an operator's impact is relatively straightforward. However, there is currently a lack of AOS for multiobjective problems (MOPs) because defining an appropriate credit assignment is nontrivial for MOPs. To identify and examine the main factors in effective credit assignment strategies, this paper proposes a classification that groups credit assignment strategies by the sets of solutions used to assess an operator's impact and by the fitness function used to compare those sets of solutions. Nine credit assignment strategies, which include five newly proposed ones, are compared experimentally on standard benchmarking problems. Results show that eight of the nine credit assignment strategies are effective in elevating the generality of a multiobjective evolutionary algorithm and outperforming a random operator selector.

System Architecture [1] is the high-level design of a system, defining the main elements of function and form, the mapping between function and form, and the interfaces between elements of function and form, and between the system and the surrounding context. Several studies have highlighted the importance of system architecture [2], as it is a point in the design process of unique leverage: most of the lifecycle cost of a system is committed after the system architecture phase[3], and a similar argument can be made for performance. The author introduced the VASSAR framework for knowledge-intensive architecture optimization [4]. The VASSAR framework is essentially an architecture evaluation and optimization framework, which combines a rule-based engine for evaluation with several heuristic optimization strategies. The VASSAR framework has been applied to multiple large aerospace systems [5-7]. The performance of several domain-independent and domain-specific heuristics was also compared in some preliminary experiments [8].

The goal of the Soil Moisture and Ocean Salinity mission over land is to infer surface soil moisture from multiangular L-band radiometric measurements. As the canopy affects the microwave emission of land, it is necessary to characterize different vegetation layers. This paper presents the Reference Pixel L-Band Experiment (REFLEX), carried out in June-July 2003 at the Vale/spl grave/ncia Anchor Station, Spain, to study the effects of grapevines on the soil emission and on the soil moisture retrieval. A wide range of soil moisture (SM), from saturated to completely dry soil, was measured with the Universitat Polite/spl grave/cnica de Catalunya's L-band Automatic Radiometer (LAURA). Concurrently with the radiometric measurements, the gravimetric soil moisture, temperature, and roughness were measured, and the vines were fully characterized. The opacity and albedo of the vineyard have been estimated and found to be independent on the polarization. The /spl tau/--/spl omega/ model has been used to retrieve the SM and the vegetation parameters, obtaining a good accuracy for incidence angles up to 55/spl deg/. Algorithms with a three-parameter optimization (SM, albedo albedo, and opacity) exhibit a better performance than those with one-parameter optimization (SM).