OrbitTAMP: Grounding Language Models for Task and Motion Planning in Spacecraft Rendezvous

Yuji Takubo; Daniele Gammelli; Marco Pavone; Simone D'Amico

cs.ROOct 1, 2026

OrbitTAMP: Grounding Language Models for Task and Motion Planning in Spacecraft Rendezvous

Authors: Yuji Takubo, Daniele Gammelli, Marco Pavone, Simone D'Amico

Organizations: Stanford University, Stanford, CA 94305, USA · Department of Aeronautics and Astronautics, Stanford University, Stanford, CA 94305, USA · Italian Institute of Artificial Intelligence (AI4I), Turin, Italy · NVIDIA Research, Santa Clara, CA 95051, USA

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.

Figures & tables

Fig. 1 : Overview of the proposed framework. The intent parsing initializes the partial mission specification $\mathcal{M}_{0}$ ; behavior sequencing produces $\mathcal{M}_{1}$ ; waypoint generation produces $\mathcal{M}_{2}$ ; and trajectory optimization generates the final state and control trajectories. All three specifications are successive states of the same MissionConfig .

Fig. 2 : Example behavior–waypoint graph.

Domain	Description	$a\delta\lambda$ [m]	$a\delta e_{y}$ [m]	$a\delta i_{y}$ [m]
$E_{0}$	E/I-separated, Central	$[-5,5]$	$\pm[30,70]$	$\pm[30,70]$
$E_{+}$	E/I-separated, $+V$ -bar	$[100,250]$	$\pm[30,70]$	$\pm[30,70]$
$E_{-}$	E/I-separated, $-V$ -bar	$[-250,-100]$	$\pm[30,70]$	$\pm[30,70]$
$F_{+}$	Flat, $+V$ -bar	$[100,250]$	$[-5,5]$	$[-5,5]$
$F_{-}$	Flat, $-V$ -bar	$[-250,-100]$	$[-5,5]$	$[-5,5]$

TABLE I : Waypoint domains used in the case study.

ID	Behavior primitive	Directed transitions	Duration (orbits)
1	Station-keeping	Self-loop at all domains	$[1,3]$
2	Drift, slow short, $+V$ direction	$E_{0}\rightarrow E_{+}$ , $E_{-}\rightarrow E_{0}$	$[2,4]$
3	Drift, slow short, $-V$ direction	$E_{0}\rightarrow E_{-}$ , $E_{+}\rightarrow E_{0}$	$[2,4]$
4	Expand R/N separation	$F_{+}\rightarrow E_{+}$ , $F_{-}\rightarrow E_{-}$	$[1,3]$
5	Shrink R/N separation	$E_{+}\rightarrow F_{+}$ , $E_{-}\rightarrow F_{-}$	$[1,3]$
6	Approach from $-V$ -bar	$F_{-}\rightarrow E_{0}$	$[1.5,4]$

TABLE II : Behavior primitives used as directed edges in the behavior-waypoint graph.

	Random		Heuristic		Learned $p_{\phi}$
	Fixed	Variable	Fixed	Variable	Fixed	Variable
SCP success [%]	84.0	95.0	95.7	95.1	77.1	95.6
Fuel [ $\mathrm{cm/s}$ ]	$15.31\pm 8.83$	$11.05\pm 5.37$	$9.37\pm 5.42$	$6.73\pm 4.19$	$7.00\pm 4.08$	$\mathbf{6.65\pm 4.13}$
SCP iterations	$6.86\pm 8.01$	$9.98\pm 7.43$	$\mathbf{5.85\pm 7.25}$	$8.05\pm 6.13$	$6.72\pm 7.27$	$7.21\pm 7.12$

TABLE III : Waypoint-generation performance on 1,000 held-out test cases; fuel and SCP iterations are the mean

\pm

standard deviation over the cases in which the SCP converges.

Fig. 3 : Preference win rate of the structured behavior–duration planning against the best of $K$ uniformly sampled graph-valid candidates over 100 scenarios.

Behavior selector	Waypoint proposal	Waypoint contract in SCP	SCP success	Fuel $[\mathrm{cm/s}]$	SCP iterations	Aux. SCP calls/request	Percentile [%]		Online time [s]
Behavior selector	Waypoint proposal	Waypoint contract in SCP	SCP success	Fuel $[\mathrm{cm/s}]$	SCP iterations	Aux. SCP calls/request	Strict	Baseline	Online time [s]
Uniform random (best of 64)	Representative waypoint	Fixed	100/100	$7.65\pm 7.04$	$2.90\pm 3.80$	0	–	–	–
Leg library	Representative waypoint	Fixed	100/100	$5.12\pm 8.06$	$3.37\pm 7.37$	60	98.9	99.6	74.77
Uniform random (best of 64)	Min.-displacement heuristic	Variable	100/100	$3.82\pm 3.94$	$4.75\pm 4.30$	0	–	–	–
Learned $Q_{\psi}$	Min.-displacement heuristic	Variable	97/100	$2.07\pm 2.79$	$3.57\pm 3.93$	0	96.2	96.5	3.02

TABLE IV : Behavior-duration planning and corresponding trajectory performance after trajectory optimization. Fuel and SCP iterations are the mean

\pm

standard deviation over successful trajectories.

Entry	Representative realization
Command	Drift towards the target, traverse the near-target region for 1 orbit, and then withdraw toward $+V$ -bar. Prioritize minimizing fuel over maximizing observation quality, and complete the entire operation within 5 orbits.
Scenario	$\boldsymbol{w}_{0}=[\,0,\;-156.24,\;0,\;-38.97,\;0,\;-60.99\,]$
	$r_{\mathrm{KOZ}}=30\,\mathrm{m}$ (default), $M_{0}=0.0^{\circ}$
$\mathcal{M}_{0}$	$\mathcal{P}=$ [fuel, observation], $\mathcal{H}=\emptyset$ (default)
	$T\in[\bot,\,5.0]~\text{orbits}$
	$\boldsymbol{b}=[2,1,7]$ , $\boldsymbol{d}=[\bot,1.0,\bot]~\text{orbits}$ , $\boldsymbol{w}=[\bot,E_{0},F_{+}]$

TABLE V : Representative progressive completion of a natural-language command into

\mathcal{M}_{0}

\mathcal{M}_{1}

, and

\mathcal{M}_{2}

, followed by the SCP-optimized waypoint sequence.

Panel of Figure 9 — Fig. 4 : Representative trajectory optimized via SCP based on the completed MissionConfig (cf. Table V ).

Fig. 5 : Architecture comparison against the proposed hierarchical architecture, tested with three dataset splits.

Dataset split	Backend	Exact $\mathcal{M}_{0}$ [%]	Atomic match [%]
Dataset split	Backend	Exact $\mathcal{M}_{0}$ [%]	Precision	Recall
Development	Qwen3.5-9B	75	90.6	85.1
	Ministral3-8B-Instruct	70	88.1	89.0
	GPT-5.6 Terra (high)	98	99.0	99.5
Held-out instance	Qwen3.5-9B	78	91.1	89.8
	Ministral3-8B-Instruct	70	89.1	90.3

TABLE VI : Intent-parsing performance on the prompt-development split and two held-out splits.

Fig. 6 : End-to-end mission satisfaction and pipeline completion across semantic parsers on 100 commands per split. All parsed specifications are screened by the same deterministic boundary/path verifier at $N_{\mathrm{rev}}=0$ .

Panel of Figure 13 — Fig. 7 : Effect of semantic test-time compute on intent parsing with Qwen3.5-9B and the development-split commands.

Fig. 8 : Effect of planning-search budget over the top- $N_{\mathrm{plan}}$ behavior-duration candidates ranked by $Q_{\psi}$ .

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Style	Command realization
Formal	Circumnavigate the target through the near-target operating region. Prioritize maximizing observation quality over minimizing completion time.
Natural	Move the chaser around the target along a circumnavigation path. Give observation quality higher priority than completion time.
Concise	Chaser task: target circumnavigation. Priority order: observation quality, then completion time.
Radio	Circumnavigate target, near-field. Prioritize observation quality over completion time, all phases.

Appendix

TABLE I : Illustrative renderings of the same reference mission intent across the supported language styles.

Split	Mission instances	Language styles
Development	Development sample	Formal, natural, concise
Held-out instance	Independently sampled	Formal, natural, concise (seen)
Held-out structural	Independently sampled	Radio register (unseen)

Appendix

TABLE II : Command-benchmark splits used for language-conditioned evaluation.

Hyperparameter	Value
$\epsilon$	$10^{-3}$
$\{\rho_{0},\rho_{1},\rho_{2}\}$	$\{0.0,0.25,0.7\}$
$\{\alpha_{1},\alpha_{2},\beta,\gamma\}$	$\{2,2,1.5,0.9\}$
$\{r^{(1)},r_{\min},r_{\max}\}$	$\{0.2,10^{-9},10\}$
$\{w^{(1)},w_{\max}\}$	$\{100,10^{20}\}$
$\#$ max. iter	100

Appendix

TABLE III : Hyperparameters of

\texttt{SCvx}\star

Explore similar work

Jun 2, 2026math.OC

Semantic Constraint Synthesis for Adaptive Trajectory Optimization via Large Language Models

Trajectory optimization is a critical component for enabling safe and reliable autonomous operations in space exploration. As space missions increase in frequency, complexity, and scope, there is a growing need to rapidly formulate mathematically sound trajectory optimization problems that accurately reflect mission objectives and operational constraints. However, translating mission intent into tractable analytical formulations for trajectory optimization requires substantial domain expertise. This paper presents a framework that leverages large language models (LLMs) to translate natural language descriptions of mission requirements and constraints into executable trajectory optimization code and corresponding mathematical formulations. Experiments in spacecraft rendezvous scenarios demonstrate a high success rate in reconditioning a convex trajectory optimization problem from semantic mission requirements. Ultimately, this work highlights the potential of LLMs to bridge high-level intent and formal optimization models, enabling more flexible and efficient trajectory design of spacecraft.

Eleanor Brosius, Yuji Takubo, Daniele Gammelli +2

Stanford University 496 Lomita Mall, Stanford, CA, 94305

Intent-Aligned Autonomous Spacecraft Guidance Robust Trajectory Optimization Modeling

Ask PDF

Jan 16, 2026cs.AI

AstroAgentBench: Evaluating Agentic Planning on Space Mission Planning Tasks

Recent LLM-for-Space systems address mission planning, scheduling, operations support, simulator control, and autonomy, but their evaluations use different task contracts, control settings, simulators, and success criteria. We introduce AstroAgentBench, a seven-family benchmark for executable space mission planning in the domains of scheduling, observation planning, constellation design, and relay support. For each case, an agent submits a planning artifact that is checked by an external verifier for schema, timing, geometry, resources, and mission value. Results report validity and normalized scores, with comparisons to task-specific solver references. Across five LLM agent systems and 35 held-out cases, the strongest systems approach or exceed solver-reference scores on several families, while weaker systems often fail to produce high-value valid plans and even strong systems lose quality on geometric, product-level, or design-heavy tasks. Trace analyses separate two failure points: task-contract misformulation and weak solution construction. Successful runs instead calibrate agent-written implementations against verifier feedback and adapt search to case-specific structure. Ablations show that procedure injection and memory accumulation help selectively, when they supply the missing formulation, calibration, or search support.

Weiyi Wang, Xinchi Chen, Jingjing Gong +2

Fudan University · OpenMOSS Team · Shanghai Innovation Institute

Spacecraft Human-Centric Mission Planning Large Language Model Planning Large Language Model Agents

Ask PDF

Sep 24, 2026cs.AI

GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI

Large Language Models (LLMs) typically exhibit a performance profile where reliability degrades as task complexity increases. We address the challenge of generating high-quality natural language executable plans for complex tasks by introducing

\textbf{GRASP}

, a strategy-aware, multi-stage planning framework. GRASP decouples the planning pipeline across specialized, context-isolated modules: it pre-compiles global macro-guidelines (GenPlan), explores alternative localized strategies within isolated context windows (RevPlan), and independently evaluates trajectories using a multi-criteria discriminator (VerPlan). Empirical evaluations show that GRASP consistently establishes a new state-of-the-art frontier across diverse datasets, yielding substantial accuracy gains over direct LLM planners on Natural Plan Calendar Scheduling (

\sim

12.4

\%$$\uparrow

), ZebraLogic (

\sim

30.8

\%$$\uparrow

), and SciBench Math. Crucially, under multi-task scaling-where standard planners suffer immediate performance collapse-GRASP completely flattens the multi-task degradation penalty. In interleaved dual-task environments, GRASP achieves an absolute accuracy gain of up to 16.7

\%

over direct LLM planners. Furthermore, by isolating context and enforcing strict macro-regularization, GRASP outperforms frontier reasoning models (such as GPT-5-mini) by a margin of 14.5

\%

Arunabh Srivastava, Mohammad A., Khojastepour +2

Amir · University of Maryland, College Park, MD · NEC Laboratories America, Inc.

Large Language Model Planning Data Generation

Ask PDF