Reproduce “Path Planning and Reinforcement Learning-Driven Control of On-Orbit Free-Flying Multi-Arm Robots” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Álvaro Belmonte-Baeza, José Luis Ramón, Leonard Felicetti, Miguel Cazorla, Jorge Pomares
A hybrid framework for free-flying multi-arm robots integrates trajectory optimization with a reinforcement learning-driven controller, enabling robust and adaptive motion in uncertain on-orbit environments. This approach achieves precise tracking of complex paths, with the RL policy demonstrating mean tracking errors of 1.5-2.5 cm for the robot base, outperforming classical control methods (7.6-9.2 cm) for high-dimensional manipulation tasks.
Key insights
- Integrating thruster control into the trajectory optimization significantly enhances motion smoothness, stabilizes the robot base, and drastically reduces interaction forces at arm docking points.
- A single, generic reinforcement learning policy can effectively track diverse and complex motion profiles (e.g., surface locomotion, spacecraft approximation) generated by a simplified planner, demonstrating robustness to model discrepancies and coupled dynamics.
- Specific RL training strategies like curriculum learning, domain randomization, and observation noise injection are critical for developing robust policies that generalize well across varied conditions and facilitate sim-to-real transfer for space robotics applications.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.