GIFT: Generalizable Interaction-aware Functional Tool Affordances without Labels
This addresses the challenge of automating tool affordance learning for robotics, reducing reliance on costly human data, though it is incremental as it builds on existing interaction-based methods.
The paper tackles the problem of learning visual affordances for tool use without human labels or expert demonstrations by grounding affordances in physical interactions, and shows that GIFT outperforms baselines on tasks like hooking, reaching, and hammering, matching a human oracle on two of three tasks with novel tools.
Tool use requires reasoning about the fit between an object's affordances and the demands of a task. Visual affordance learning can benefit from goal-directed interaction experience, but current techniques rely on human labels or expert demonstrations to generate this data. In this paper, we describe a method that grounds affordances in physical interactions instead, thus removing the need for human labels or expert policies. We use an efficient sampling-based method to generate successful trajectories that provide contact data, which are then used to reveal affordance representations. Our framework, GIFT, operates in two phases: first, we discover visual affordances from goal-directed interaction with a set of procedurally generated tools; second, we train a model to predict new instances of the discovered affordances on novel tools in a self-supervised fashion. In our experiments, we show that GIFT can leverage a sparse keypoint representation to predict grasp and interaction points to accommodate multiple tasks, such as hooking, reaching, and hammering. GIFT outperforms baselines on all tasks and matches a human oracle on two of three tasks using novel tools.