Decentralized Reactive Satellite Constellation Planning for Event Observation

Satellite observation planning is an active field of research that aims to improve the quantity and utility of observations by agile or taskable satellites. Most of the existing work focuses on observing static ground points with constant rewards, which is not sufficient to capture dynamic events on Earth’s surface. We introduce in this work a reactive planning method suitable for observation of time-varying events by a constellation of homogeneous or heterogeneous satellites, and compare it to a non-reactive planning algorithm. We perform this comparison on a generalized scenario with varying parameters such as event duration and number of satellites. We also examine several realistic event scenarios, including lightning storms, fires, and high-flow events in rivers. We show that reactive planning outperforms non-reactive planning in observing and re-observing events, and we show that the reactive planning has tunable hyperparameters that can be adjusted based on the satellite operator’s event observation preferences. A sensitivity analysis shows that reactive planning is most beneficial compared to non-reactive planning in terms of generating more re-observations of events when there are more planes of satellites and when event durations are longer. Ultimately, the proposed reactive planning strategies can greatly reduce our response time to – and increase our spatio-temporal sampling of – short-lived events and scientific processes.