51st Lunar and Planetary Science Conference (2020) 2503.pdf ADVANCING THE SCIENTIFIC FRONTIER WITH INCREASINGLY AUTONOMOUS SYSTEMS. R. B. Amini1, J. Castillo-Rogez and J. Day, 1Jet Propulsion Laboratory, California Institute of Technology, [email protected]. Introduction: A close partnership between people autonomous systems to adjust to conditions over the and semi-autonomous machines has enabled decades of duration of the science mission. space exploration, but to continue to expand our Benefits and Advantages: One principal advantage horizons, our systems must become more capable. of autonomous systems is their ability to (re)plan their Increasing the nature and degree of autonomy - allowing own operations and respond to their environments. On- our systems to make and act on internal decisions - board planning enables missions and observations that enables new science capabilities. In the current need response to hazards, extreme and unknown Planetary Science Decadal (Vision and Voyages), terrains, and transient phenomena on timescales faster increased autonomy was identified as one of eight core than possible than by involving ground systems. Rapid multi-mission technologies required for future missions. response to hazards is of particular importance for Both scientific and technological progress over the last increasing the productivity of long-range mobile decade has reinforced the necessity and demonstrated platforms. These include subsurface missions and other the feasibility of autonomous systems for planetary missions exploring hazardous environments, like cliffs, exploration. Specifically, autonomy is poised to enable caves, and crevices on planets, satellites, and small new planetary missions and access to new bodies. environments, increase science return, increase Other missions that are enabled by on-board resilience of mission systems, and reduce mission cost. planning are those where the typical ground planning Six of the pre-decadal studies recently selected by cycle incurs a prohibitive cost or science penalty. For NASA explicitly call for surface exploration at a wide instance, mission concepts for Venus or icy moon range of planetary bodies as the natural next step in the landers are tightly time- or energy-constrained. These exploration of targets whose scientific value was missions can dramatically increase productivity by demonstrated by reconnaissance missions over the past autonomously determining how and what to sample decades. These missions require operating in with on-board planning and data processing, instead of challenging environments (e.g., Mercury, Enceladus, waiting for uplink of sampling instructions from a Triton) and/or require long-range mobility (e.g., Ceres, science team on the ground. Although not resource Moon). Other selected concepts, like the Pluto Orbiter restricted, prospective fast flyby missions of interstellar and Kuiper Belt exploration missions must address the visitors like ‘Oumuamua and Comet Borisov may need challenges of operation at an extreme distance. Over the to rely on autonomous targeting for close approach longer term, NASA’s vision includes accessing the navigation and science planning depending on prior subsurface oceans of known ocean worlds. Successful knowledge of the target and the available observation accomplishment of these missions will require window. deployment of increasingly autonomous systems. Transient phenomena, such as dust devils and Continued technology development, both internal to outgassing plumes, can be detected and observed in NASA and strategically selected external investments, real-time, instead of reliance on serendipity. This provides a rich field of solutions and capabilities to extends to constellation missions that may require advance the use of autonomy in NASA missions. coordinated observations of transient phenomena Advances in goal-directed operation, model-based observed from multiple vantage points, spatially reasoning, and situational awareness allows operators distributed sensors in the case of fields and particles in and scientists to focus on objectives and oversight, giant planet systems, or distributed transmitters and while the deployed system determines how to safely receivers on multiple platforms, such as the CONSERT perform its assigned objectives. Progress in bistatic radar experiment on Rosetta and Philae. development of systems engineering processes, system Demonstrated Progress: Experience with the and environmental models, and formal behavior Curiosity rover offers a compelling and extensible case specifications enable more rigorous analysis and study in leveraging autonomy to increase mission “correct-by-construction” design specifications, leading science return. Curiosity has adopted limited to guarantees of system behavior. Further, advances in autonomous features that have significantly benefited artificial intelligences and machine learning enable on- the mission. For instance, the AEGIS experiment on board learning and model adaptation that allow Curiosity, which autonomously selects targets for its ChemCam instrument and has resulted in additional 51st Lunar and Planetary Science Conference (2020) 2503.pdf science return. However, Curiosity’s ultimate reliance 2020, Mars Helicopter will perform the first powered on traditional planning results in 48% of sols being flight beyond Earth – soon to be followed by Dragonfly underutilized. By adding system-level planning that at Titan. These missions are pathfinders for new integrates path planning, data processing, and health technologies and modes of exploration. The portfolio of monitoring features, Curiosity would be able to utilize studies for the Planetary Science Decadal Survey further most of this underutilized time [1]. These productivity conveys the ambitions of the community. It is expected enhancements are not limited to surface missions and that the resulting concepts will set priorities for NASA’s apply to orbiter and flyby missions as well. Autonomy technology investments for the next decade, which can permit more efficient mapping campaigns by should include drivers for expanding existing tailoring science observations to current spacecraft state autonomous software, the supporting hardware (e.g., instead of relying on overly conservative a priori computer), and testbeds that provide sufficient estimates. validation and testing of autonomous capabilities. In the last decade, new autonomous capabilities In order to advance the frontier of robotic space have been demonstrated that will improve spacecraft explorations, NASA must pioneer resilient, self-aware, resiliency, reducing the risk of long-duration missions and autonomous systems able to weigh risk and make and missions that cannot reliably communicate with decisions locally to ensure that tomorrow’s missions are Earth. In the current operating paradigm, entrance into a success. Application of advances in autonomous a spacecraft safe mode suspends science observations systems technology will dramatically increase science and restricts spacecraft functionality for periods of a few return by extending the reach, the productivity and the days or even weeks, depending on the nature of the robustness of NASA missions. These technologies are anomaly. Based on historical data, about 50% of safings critical to enabling missions such as rovers on Mars that can be mitigated as the system decides to permit can traverse a thousand kilometers, rugged submersibles continued science operations or restores its own under the ocean at Europa, long-duration balloons in the functionality [2,3]. Recent work in model-based atmospheres of Titan and Venus, and autonomous diagnosis is an example of advancements that allow an explorers of the Kuiper Belt. These systems will have autonomous system to achieve directed goals while far greater reliability, reduced mission risk and working around anomalies [4]. significant reductions in development and operations New Opportunities and Next Steps: As the barrier cost. to space is further reduced with lower cost launch opportunities and commercial-off-the-shelf space This work is being carried out at the Jet Propulsion systems, new mission architectures utilizing large Laboratory, California Institute of Technology, under numbers of assets are now feasible. For instance, these contract to NASA. Government sponsorship can include constellations at Venus, Moon, and Mars acknowledged. and missions that survey large numbers of small bodies [5]. The operational complexity in managing a large References: number of satellites can be cost-effectively addressed [1] D. Gaines, et al. "Productivity Challenges for using autonomy. Rather than developing and testing Mars Rover Operations." (2016). JPL Technical Report, detailed command sequences for each spacecraft, https://ai.jpl.nasa.gov/public/documents/papers/gaines_ autonomous capabilities such as goal-based report_roverProductivity.pdf commanding can be deployed that allow spacecraft to [2] T. Imken, et al. "Modeling Spacecraft Safe Mode develop context-sensitive plans based on the directed Events." In 2018 IEEE Aerospace Conference, pp. 1-13. goals. This approach has the promise of reducing IEEE, 2018. required infrastructure investments and operator effort, [3] R. Amini, et al. "Enhanced and Enabled Astro- especially for networked constellations. However, the physical Observations with System-Level Autonomy." benefits of autonomy can only be realized by addressing American Astronomical Society Meeting Abstracts.