Autonomy in the Real-World: Autonomous Trajectory Planning for Asteroid Reconnaissance via Stochastic Optimization

TL;DR

This paper develops a stochastic optimization-based autonomous trajectory planning algorithm for asteroid reconnaissance missions, enabling safer and more efficient deep-space exploration with minimal ground intervention.

Contribution

It introduces a novel stochastic trajectory optimization method tailored for autonomous deep-space asteroid reconnaissance, transforming the problem for practical nonlinear solver implementation.

Findings

The proposed method outperforms existing benchmarks in numerical experiments.

Validation confirms the model assumptions for the trajectory planner.

The approach enables autonomous mission planning with reduced ground control dependency.

Abstract

This paper presents the development and evaluation of an optimization-based autonomous trajectory planning algorithm for the asteroid reconnaissance phase of a deep-space exploration mission. The reconnaissance phase is a low-altitude flyby to collect detailed information around a potential landing site. Although such autonomous deep-space exploration missions have garnered considerable interest recently, state-of-the-practice in trajectory design involves a time-intensive ground-based open-loop process that forward propagates multiple trajectories with a range of initial conditions and parameters to account for uncertainties in spacecraft knowledge and actuation. In this work, we introduce a stochastic trajectory optimization-based approach to generate trajectories that satisfy both the mission and spacecraft safety constraints during the reconnaissance phase of the Deep-space Autonomous Robotic Explorer (DARE) mission concept, which seeks to travel to and explore a near-Earth object autonomously, with minimal ground intervention. We first use the Multi-Spacecraft Concept and Autonomy Tool (MuSCAT) simulation framework to rigorously validate the underlying modeling assumptions for our trajectory planner and then propose a method to transform this stochastic optimal control problem into a deterministic one tailored for use with an off-the-shelf nonlinear solver. Finally, we demonstrate the efficacy of our proposed algorithmic approach through extensive numerical experiments and show that it outperforms the state-of-the-practice benchmark used for representative missions.