Problem
Research questions and friction points this paper is trying to address.
On-orbit obstacle avoidance
Satellite agents
Dynamic environments
Collision risk prediction
Orbit-Planner: Towards Latent World Models for On-Orbit Obstacle Avoidance of Satellite Agents
🤖 AI Summary
This study addresses the limited adaptability of traditional planners caused by constrained onboard navigation observations and dynamic environmental changes in satellite operations. We propose a two-stage latent world model that learns action-conditioned dynamics to infer future states within a latent space, incorporating a physical probe decoding mechanism to accurately reconstruct long-horizon imagined trajectories into physical states. Validated through Isaac Sim simulations, the proposed framework achieves a 91.7% success rate in closed-loop obstacle avoidance tasks. These results demonstrate that our approach effectively resolves efficient planning challenges in dynamic scenarios, significantly enhancing system robustness and generalization capabilities for autonomous satellite navigation.
📝 Abstract
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.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Latent World Model
On-Orbit Obstacle Avoidance
Physics Probe
Action-Conditioned Dynamics
Latent Space Rollouts
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
