Compositional Servoing by Recombining Demonstrations
This work addresses task transfer and multitask capability in robotic manipulation for high-precision scenarios, representing an incremental improvement over existing visual servoing methods.
The paper tackles the problem of weak task transfer in learning-based manipulation policies by introducing a framework that formulates visual servoing as graph traversal, using recombined demonstrations to achieve higher task-specific success rates in simulation and real-world experiments.
Learning-based manipulation policies from image inputs often show weak task transfer capabilities. In contrast, visual servoing methods allow efficient task transfer in high-precision scenarios while requiring only a few demonstrations. In this work, we present a framework that formulates the visual servoing task as graph traversal. Our method not only extends the robustness of visual servoing, but also enables multitask capability based on a few task-specific demonstrations. We construct demonstration graphs by splitting existing demonstrations and recombining them. In order to traverse the demonstration graph in the inference case, we utilize a similarity function that helps select the best demonstration for a specific task. This enables us to compute the shortest path through the graph. Ultimately, we show that recombining demonstrations leads to higher task-respective success. We present extensive simulation and real-world experimental results that demonstrate the efficacy of our approach.