Transformer-Based Warm-Starting for Feasible and Optimal Terminal Approach to Tumbling Objects with Space Manipulators

This paper presents a learning-based warm-starting method for sequential convex programming (SCP) to enhance the terminal approach of space manipulators targeting tumbling objects. By utilizing a causal transformer in the torque-allocation stage, the approach reduces SCP iteration counts by up to 28% and runtime by 23%, while maintaining control-cost distribution across 300 scenarios. This advancement significantly improves computational efficiency and trajectory robustness in robotic servicing applications, making it valuable for practitioners working with trajectory optimization in space robotics.