BiPOP preprint pairs language models with a planner to coordinate robot arms

Researchers report that BiPOP, a framework for coordinating two robot arms, outperformed language-model planning baselines on simulated long tasks and ran end to end on the real NICOL robot. The findings appear in an arXiv preprint. A language model converts instructions into a formal task specification. A separate planner assigns actions to arms and schedules them around reach, timing and synchronization constraints, allowing safe actions to run concurrently. The reported comparisons cover success rate, planning time and plan efficiency in simulation; the real-robot demonstration establishes that the full pipeline can run on hardware.