Artificial Intelligence and Machine Learning in Planetary Science

Machine Learning for Small Bodies in the Solar System is the first book devoted entirely to the application of artificial intelligence and machine learning to the study of asteroids, comets, and trans-Neptunian objects. Covering topics from celestial mechanics and asteroid family identification to object detection, spectroscopy, and autonomous data analysis, it combines methodological foundations with practical implementations, including code examples and publicly available repositories. The volume is intended both as an introduction for researchers entering the field and as a reference for experienced planetary scientists seeking to incorporate modern AI techniques into their research. As the era of data-intensive surveys led by facilities such as the Vera C. Rubin Observatory begins, the methods presented in this book provide an essential foundation for the next generation of solar system science.

This work establishes the first comprehensive benchmark for evaluating multimodal large language models on resonance classification in celestial mechanics. Its standardized datasets, evaluation methodology, and comparison of commercial and open-source models provide a foundation for future research on generative AI in dynamical astronomy. Essential reading for anyone interested in the intersection of large language models and Solar System dynamics.

A landmark paper in autonomous spacecraft operations, this work demonstrated that AI planning and execution systems could reliably control a spacecraft in a real mission environment. By validating the Remote Agent architecture beyond laboratory experiments, it established a foundation for autonomous space exploration and remains an essential reference for researchers interested in AI-driven mission planning, spacecraft autonomy, and intelligent control.

A landmark paper that introduced Physics-Informed Neural Networks (PINNs), establishing one of the most influential frameworks for integrating physical laws with deep learning. By embedding differential equation constraints directly into neural network training, this work opened new avenues for data-driven scientific computing and inverse modeling. Essential reading for researchers interested in applying machine learning to physics, engineering, and celestial mechanics.

A pioneering application of DeepONets to asteroid thermophysical modeling, this work shows how neural operator architectures can dramatically accelerate surface temperature calculations while preserving high accuracy. Its methodology paves the way for efficient coupling of AI-based thermal models with orbital dynamics, making it an important contribution to the growing use of scientific machine learning in planetary science.

A pioneering application of machine learning to asteroid dynamics, this work showed that scikit-learn classifiers could successfully identify three-body mean motion resonances. While the field has since advanced considerably, the paper remains a useful reference for researchers interested in applying classical machine-learning techniques to problems in celestial mechanics.

By comparing several supervised machine-learning classifiers for asteroid family identification, this study showed that data-driven methods can effectively complement traditional clustering approaches. The techniques remain relevant today, as they readily scale to the rapidly expanding asteroid databases produced by modern surveys.

An important methodological contribution, this paper was among the first to apply vision transformers to the identification of secular resonances in asteroid dynamics. It illustrates the potential of modern deep-learning architectures to complement traditional dynamical analyses and represents a significant step toward AI-driven classification of resonant populations.

One of the first studies to apply machine learning to the dynamical classification of Kuiper Belt Objects, this work showed that gradient-boosting methods can provide fast and accurate classifications with minimal loss of reliability. It remains an important reference for researchers interested in scalable approaches to the analysis of the rapidly growing KBO population.

A pioneering application of machine learning to asteroid family identification, this work showed that supervised clustering algorithms can efficiently recover known families and identify new candidate groups. The paper remains a valuable reference for understanding how machine-learning methods can complement traditional HCM approaches, particularly as the size of asteroid catalogs continues to increase.

This paper applies optimized artificial neural networks to the identification of asteroid resonances, demonstrating how machine-learning models can automate a challenging classification task in celestial mechanics. Beyond its technical contribution, it introduced an innovative methodology for resonance identification and was awarded the CELMEC Prize for innovative methods in dynamical astronomy. It remains a useful reference for researchers interested in applying neural networks to dynamical astronomy problems.

This chapter provides an in-depth review of machine-learning methods for identifying cometary activity in wide-field astronomical surveys. By reviewing both classical approaches and modern AI techniques, it offers an accessible introduction to the challenges of detecting active solar system objects and discusses future applications to next-generation surveys such as the Vera C. Rubin Observatory. A valuable starting point for researchers entering this rapidly developing field.

An innovative application of AI to spacecraft guidance and control, this work introduces an adaptive control strategy for stable hovering around asteroids with uncertain gravity. It highlights the growing role of intelligent algorithms in enabling autonomous proximity operations, inspection, and future sampling and landing missions.

One of the early applications of machine learning to the dynamical classification of trans-Neptunian objects, this work illustrates how AI can streamline the analysis of complex orbital behavior. Its methodology remains relevant as modern surveys continue to expand the known population of distant Solar System bodies, making automated classification increasingly important.

This paper demonstrates the transformative potential of the Vera C. Rubin Observatory for the discovery of long-period and hyperbolic comets, showing that next-generation surveys could dramatically increase detection rates. Beyond its scientific results, it provides valuable insight into how modern survey capabilities and data-driven methods will reshape studies of the distant Solar System. An important reference for researchers interested in the future of small-body discovery and survey science.