Ergodic Trajectory Optimization for Information Gathering

Abstract

Planetary robots currently rely on significant guidance from expert human operators. Science autonomy adds algorithms and methods for autonomous scientific exploration to improve efficiency of discovery and overcome limited communication bandwidth and delay bottlenecks. This research focuses on planning trajectories for information gathering and choosing sampling locations that have the most informative samples. We frame our exploration problem as a mapping problem for spectroscopic data and explore the concept of using low spatial density and low spectral resolution remote data as an information prior. We utilize a Gaussian Process regression model to fuse remote and in situ observations. This allows us to improve our high-resolution predictions across an entire scene without visiting all locations and compute an entropy map to guide exploration. We propose performing informative path planning using ergodic trajectory optimization. We explore the efficacy of the ergodic Spectral Multi-scale Coverage and ergodic Projection-based Trajectory Optimization algorithms. We demonstrate our approach in simulated exploration with real spectroscopic data of Cuprite, Nevada to highlight the advantages compared to traditional planning strategies. We successfully display that the ergodic Projection-based Trajectory Optimization planner outperforms all planners including state-of-the-art a non- myopic Markov Decision Process based planner. We also explore how the planning horizon affects planner performance, varying how often the entropy map is updated and the remaining sample points re-planned. We demonstrate that re-planning after every 1 sample does improve performance however a planning horizon of 4 samples offers a favorable balance between improved information gathering and increased computation time.