Rapid Trajectory Design in Multi-body Systems Using Sampling-Based Kinodynamic Planning

Abstract

Abstract Existing approaches to trajectory design in a multi-body system often rely on a human trajectory designer with sufficient knowledge of the solution space. However, automated approaches can reduce the designer’s workload and support autonomous decision-making. In the field of robotics, motion planning techniques are used to automate path construction by discretely summarizing an environment via a graph and rapidly generating solution paths via graph search algorithms. In this paper, sampling-based kinodynamic planning is leveraged to automatically design initial guesses for trajectories in a multi-body system. First, a roadmap that summarizes the desired region of the solution space is constructed. The roadmap is a directed, weighted graph composed of nodes and edges that are constructed by randomly sampling valid spacecraft states and constructing trajectory segments or impulsive maneuvers in a multi-body system. Using Dijkstra’s and Yen’s graph search algorithms, the roadmap is repeatedly searched to generate a variety of solution paths that represent geometrically distinct initial guesses for transfers between selected boundary conditions. Continuous and maneuver-enabled trajectories are recovered from these guesses using collocation and local, constrained optimization. This process is demonstrated by constructing a variety of geometrically distinct, planar transfers between Lyapunov orbits near the Moon in the Earth-Moon circular restricted three-body problem.