Orbit-Planner: Towards Latent World Models for On-Orbit Obstacle Avoidance of Satellite Agents

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Zhijian Li, Chao Ren, Peijin Wang, Xian Sun
• Venue: arXiv: cs.RO
• Published: 2026-08-17
• Collected: 2026-08-18
• URL: http://arxiv.org/abs/2608.16651v1
• PDF: https://arxiv.org/pdf/2608.16651v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

The paper presents Orbit-Planner, a system designed to help satellite agents navigate in space by predicting collision risks with limited onboard observations. It introduces a two-stage latent world model for on-orbit obstacle avoidance that learns spacecraft dynamics and uses physical probes to decode state changes from imagined trajectories. The system has shown success in long-horizon latent rollouts and closed-loop obstacle-avoidance navigation, with a success rate of 91.7%.

What the Document Says

• Satellite agents need to predict collision risks using limited onboard observations
• Conventional planners rely on predefined maps and fixed environmental assumptions
• Orbit-Planner is a two-stage latent world model for on-orbit obstacle avoidance
• Orbit-Planner learns action-conditioned spacecraft dynamics for future-state rollouts in latent space
• Physics Probe is introduced to decode physical state changes from imagined latent trajectories
• Experiments demonstrate that Orbit-Planner can perform long-horizon latent rollouts and recover physical states from imagined trajectories
• In closed-loop obstacle-avoidance navigation in Isaac Sim, it attains a success rate of 91.7%
• Code is available at https://github.com/ZhijianLi2003/Orbit_Planner

Key Concepts Mentioned

• Satellite agents
• Collision risks prediction
• On-orbit navigation tasks
• Latent world model
• Action-conditioned spacecraft dynamics
• Physical state changes
• Imagined latent trajectories
• Closed-loop obstacle-avoidance navigation
• Isaac Sim
• Code repository

Claims Made by the Document

• Orbit-Planner can perform long-horizon latent rollouts and recover physical states from imagined trajectories
• In closed-loop obstacle-avoidance navigation in Isaac Sim, Orbit-Planner attains a success rate of 91.7%

Evidence Used

• Results from experiments as described in the paper

Methods / Approach

• Learning action-conditioned spacecraft dynamics
• Using Physics Probe to decode physical state changes

Limitations Stated by the Authors

• UNKNOWN - not stated in the source

Limitations Noted by the Agent

Agent interpretation — not from the source.

• The paper does not provide details on how the system handles unexpected or rare events
• It is unclear if the success rate of 91.7% is compared to other existing methods

Open Questions

• How does Orbit-Planner handle unexpected or rare events in dynamic on-orbit scenarios?
• How does the success rate of 91.7% compare to other existing methods for obstacle avoidance in satellite navigation?

Notes for Future Q&A

May include interpretation. Labeled where it does.

• The paper focuses on the presentation and evaluation of Orbit-Planner, but it lacks information about how the system handles unexpected or rare events
• It is important to consider that the success rate might be influenced by the specific simulation environment used (Isaac Sim)

Raw Abstract / Excerpt

Satellite agents for on-orbit navigation tasks need to predict collision risks using limited onboard observations. However, conventional planners often rely on predefined maps and fixed environmental assumptions, limiting their adaptability in dynamic on-orbit scenarios. In this paper, we propose Orbit-Planner, a two-stage latent world model for on-orbit obstacle avoidance. Orbit-Planner learns action-conditioned spacecraft dynamics to perform future-state rollouts in latent space, and introduces a Physics Probe to decode physical state changes from imagined latent trajectories. Experiments demonstrate that Orbit-Planner can perform long-horizon latent rollouts and recover physical states from imagined trajectories. In closed-loop obstacle-avoidance navigation in Isaac Sim, it attains a success rate of 91.7%. Code is available at https://github.com/ZhijianLi2003/Orbit_Planner.