The exploration of space requires the best of the nation's technical community to provide the technologies that will enable human and robotic exploration beyond Low Earth Orbit (LEO): to establish a lunar presence, to visit asteroids, to extend human reach to Mars, and for increasingly ambitious robotic missions such as a Europa Lander. Autonomous Systems technologies provide the means of migrating mission control from Earth to spacecraft, habitats, and robotic explorers. This is enhancing for missions in the Earth-Lunar neighborhood and enabling for deep space missions. Long light-time delays, for example up to 42 minutes round-trip between Earth and Mars, require time-critical control decisions to be closed on-board autonomously, rather than through round-trip communication to Earth mission control. For robotic explorers this will be done through automation, while for human missions this will be done through astronaut-automation teaming.Long-term crewed spacecraft and habitats, such as the International Space Station, are so complex that a significant portion of the crew's time is spent keeping it operational even under nominal conditions in low-Earth orbit, while still requiring significant real-time support from Earth. The considerable challenge is to migrate the knowledge and capability embedded in current Earth mission control, with tens to hundreds of human specialists ready to provide instant knowledge, to on-board automation that teams with astronauts to autonomously manage spacecraft and habitats. For outer planet robotic explorers, the opportunity is to autonomously and rapidly respond to dynamic environments in a timely fashion.Machine learning has made spectacular advances for terrestrial applications, exceeding human capabilities in tasks such as image classification. Machine learning could become an increasingly important aspect of space exploration, from finding novel patterns in the science data transmitted from robotic spacecraft, to the operation of sustainable habitats. Machine learning and inferencing calls for new computing paradigms; for space, radiation tolerant processors will be enabling.Subtopics:In order to enable on-board autonomy, both software advances and computing advances need to be addressed.The autonomous agent subtopic addresses this challenge by soliciting proposals that leverage the growing field of cognitive computing to advance technology for deep-space autonomy.Fault management is an integral part of space missions. The fault management subtopic spans the lifecycle of fault management for space missions from design through verification and validation to operations. In the past, the predominant operational approach to detected faults has been to safe the spacecraft, and then rely on Earth mission control to determine how to proceed. New mission concepts require future spacecraft to autonomously decide how to recover from detected anomalies and continue the mission. The fault management subtopic solicits proposals that advance fault management technology across architectures, design tools, verification and validation, and operations.The sustainable habitat subtopic calls for machine learning technology in order to substantially improve diagnostic and prognostic performance for integrated systems health management. This subtopic solicits technology for long-term system health management that goes beyond short-term diagnosis technology to include advances machine learning and other prognostic technologies. Enhancing the capability of astronauts is also critical for future long-duration deep space missions.The Deep Neural Network accelerator and Neuromophic computing subtopic addresses extrapolating new terrestial computing paradigms related to machine learning to the space environment. For machine inferencing and learning computing hardware proposals, metrics related to energy expenditure per operation (e.g., multiply-add) and throughput acceleration in a space environment are especially relevant.The subtopic on swarms of space vehicles addresses technologies for control and coordination of planetary rovers, flyers, and in-space vehicles in dynamic environments. Co-ordinated swarms can provide a more robust and sensor-rich approach to space missions, allowing simultaneous recording of sensor data from dispersed vehicles and co-ordination especially in challenging environments such as cave exploration.