Neural operator learning for collision-aware trajectory planning of spacecraft swarms

2026-07-31

AI Summary

This work addresses the poor scalability and limited generalization of trajectory planning for large-scale spacecraft swarms operating in congested orbits. The authors propose a permutation-equivariant neural operator that, via a single forward pass, directly maps the spatial distributions of spacecraft, targets, and debris into globally collision-free, fuel-efficient trajectories for the entire swarm. By integrating self-supervised physics-based losses with batch Gauss–Newton optimization, the method achieves zero-shot generalization without requiring labeled optimal trajectories. It scales seamlessly from 10 training agents to 1,000 and robustly handles varying debris densities and adversarial threats. In scenarios involving over 11,000 cataloged objects, the generated trajectories match the accuracy of single-agent optimal control solutions, significantly reduce minimum inter-agent distances, and successfully mitigate worst-case threats that conventional blind baselines cannot resolve.

Abstract

Autonomous spacecraft swarms must plan fuel-efficient, collision-free maneuvers in increasingly congested orbits, yet classical trajectory optimization scales poorly as pairwise safety constraints multiply with swarm size, and learning-based planners rarely transfer across swarm sizes or debris densities. Here we introduce a permutation-equivariant neural operator that maps distributions of spacecraft, targets and debris to collision-aware trajectories for an entire swarm in a single forward pass, paired with a batched Gauss-Newton finish that enforces exact orbital dynamics. The operator is trained without optimal-trajectory labels, combining self-supervised physics objectives with adversarial threats generated against its own rollouts. Trained on ten spacecraft, it generalizes zero-shot to swarms of 1,000 amid more than 11,000 catalogued objects, matching a per-agent optimal-control solver's accuracy, evading worst-case threats that a debris-blind baseline cannot, and reducing proximity within the swarm several-fold. Physics-grounded operator learning thus offers a fast, scalable alternative to optimal control for crowded orbits.

Problem

• spacecraft swarms
• collision-aware trajectory planning
• orbital congestion
• scalable trajectory optimization
• autonomous navigation

Innovation

• neural operator
• permutation-equivariant
• collision-aware trajectory planning
• self-supervised learning
• zero-shot generalization