Cracking Mobile Manipulation: From Failure Diagnosis to Agentic Robot System
From the abstract
Recent vision-language-action (VLA) models have shown strong performance on local robot manipulation, yet their success does not readily translate to long-horizon mobile manipulation. In our experiments, a pretrained subtask VLA achieves 80% success rate on atomic manipulation tasks but only 19% when these skills are composed into mobile manipulation.
From the abstract. Our summary is in progress.