Language models improve planning spacecraft rendezvous operations
OrbitTAMP: Grounding Language Models for Task and Motion Planning in Spacecraft Rendezvous
RoboticsArtificial Intelligence
Summary
Planning how spacecraft meet and operate nearby now requires expert engineers to manually create safe and physically possible flight paths. The authors created a system that uses large language models combined with specialized planners to better understand natural language commands and turn them into valid mission plans. Their approach breaks down commands into reusable steps and checks them against spaceflight rules before producing actual trajectories. Tests show this method recovers correct mission details much more often than using language models alone. This work helps make spacecraft rendezvous planning more scalable and easier to verify.
What this means in practice
- •For spacecraft operators: Generate physically valid spacecraft rendezvous plans directly from natural language instructions to speed mission planning.
- •For robotics system developers: Incorporate language-grounded hierarchical planners to ensure feasible motion plans in complex dynamic environments beyond space.
Authors
Yuji Takubo, Daniele Gammelli, Marco Pavone, Simone D'Amico
Abstract
Spacecraft rendezvous and proximity operations (RPO) are currently planned through an expertise-intensive process in which engineers translate high-level operational intent into safe, dynamically feasible trajectories, creating a bottleneck to scalable operations. Large language model (LLM)-based agents could offer an intuitive interface for this process, although their outputs are not inherently grounded in orbital dynamics, operational constraints, or the structure of admissible spacecraft maneuvers. To exploit their semantic reasoning while ensuring the generated plan's physical validity, this paper presents a hierarchical framework for spacecraft task-and-motion planning (TAMP) that grounds LLM reasoning in a graph of reusable behaviors and domain-specific planning modules. Within this framework, a pretrained LLM maps a natural-language command to a partial mission specification. The associated planners then resolve unspecified decisions within the admissible operational space. Finally, trajectory optimization converts the completed mission specification into a dynamically feasible trajectory. Numerical experiments demonstrate that this architecture substantially improves intent recovery over direct LLM generation, achieving 98% exact recovery of partial mission specifications across all evaluated splits when backed by frontier LLMs. Additional test-time-compute experiments show that, for a compact 9B model, verifier-guided revision increases exact recovery from 75% to 88%, while broader behavior-plan search independently improves selection among admissible trajectory realizations. Overall, these results establish a scalable and auditable foundation for language-driven agentic planning of spacecraft RPO.