Enhancement of Venus Balloon Science Return through Multi-Agent Autonomy

Abstract:

Venus, often referred to as Earth’s planetary twin, has gained interest as an in-situ exploration target in recent years. It is being studied as the target of two ongoing Discovery-class mission proposals and a Flagship-class mission for the next planetary science decadal survey. Balloons have emerged as an attractive vehicle for the scientific exploration of Venus, ranging from atmospheric science to geophysical exploration, where surface temperature and pressure severely limit the lifetime of landed instruments. Balloons were first floated on Venus in 1985 as part of the Vega mission. Since then, balloon technology has seen significant advancement, and balloons have been included as key elements in several SIMPLEx to Flagship-class mission concepts. A recent study of aerial platforms found that variable altitude balloons offer rich scientific dividends in return for modest technology development.

A comprehensive scientific investigation of Venus can be carried out using multiple in-situ assets. In our presentation, we make the case that multi-agent autonomy can expand scientific return from in-situ explorations of Venus. We use the monitoring of putative volcanic events as a test case and present an optimization framework to monitor volcanic events using a network of three balloons with the ability to autonomously change altitude and an orbiter. The balloons detect events with infrasound microphones; exchange information about candidate events on available communication links; and assign revisits of candidate events to balloons and replan their trajectories through a multi-agent Markov Decision Process framework. The framework is tested in a simulation where the balloon trajectories are modeled as semi-Lagrangian tracers following the mean motion of the Venus atmosphere, except when they autonomously change altitude. The background atmosphere is generated using the IPSL Venus Global Circulation model. We model the temporal distribution of volcanic activity levels on Venus using levels recorded on Earth, and the spatial distribution based on the location of large Venus volcanoes as seen in Magellan data. We show that the proposed framework can enable agile science at close range at ongoing volcanic events, as opposed to similar architectures without autonomous operations or navigation capabilities.