A GPU accelerated multiple revolution lambert solver for fast mission design

Abstract

Lambert's algorithm acts as an enabler for a large variety of mission design problems. Typically an overwhelmingly large number of Lambert solutions are needed to identify sets of mission feasible trajectories. We propose a Graphics Processing Unit (GPU) accelerated multiple revolution Lambert solver to combat this computationally expensive combinatorial problem. The implementation introduces a simple initial guess generator that exploits the inherent structure of the well-known Lambert function formulated in universal variables. Further, the approach builds from the concepts of parallel heterogeneous programming utilizing both the central processing unit (CPU) and GPU in tandem to achieve multiple orders of magnitude speedup. The solution strategy is transparent and scalable to specific user resources. Speedups of two orders of magnitudes are found using a state of the art GPU on a single personal workstation, while single order of magnitude speedups are observed using the GPU on a common laptop. Example gravity assisted flyby trajectories are used to demonstrate performance and potential applications.