Plan Recovery through Object Substitution


video_stillRobots executing plans in changing or new environments currently lack the flexibility to use novel objects in a context-aware manner. We present object substitution as a solution to repairing plans in open-world robotic applications. The key insight of our work is that considering the task context is important when performing a substitution. In our evaluation we show that our approach models valid substitutions accurately and that the learned models are resilient to task variations.