Learning Visual Dynamics Models of Rigid Objects using Relational Inductive Biases
This work addresses the problem of improving physical reasoning for robots in cluttered environments, though it is incremental by comparing GNN variants on a simulated task.
The paper tackled learning visual dynamics models for rigid objects in manipulation tasks by using Graph Neural Networks (GNNs) with relational inductive biases, finding that an Auto-Predictor approach without explicit edge attributes outperformed baselines and GN-based models in generalization to novel and increased object scenarios.
Endowing robots with human-like physical reasoning abilities remains challenging. We argue that existing methods often disregard spatio-temporal relations and by using Graph Neural Networks (GNNs) that incorporate a relational inductive bias, we can shift the learning process towards exploiting relations. In this work, we learn action-conditional forward dynamics models of a simulated manipulation task from visual observations involving cluttered and irregularly shaped objects. We investigate two GNN approaches and empirically assess their capability to generalize to scenarios with novel and an increasing number of objects. The first, Graph Networks (GN) based approach, considers explicitly defined edge attributes and not only does it consistently underperform an auto-encoder baseline that we modified to predict future states, our results indicate how different edge attributes can significantly influence the predictions. Consequently, we develop the Auto-Predictor that does not rely on explicitly defined edge attributes. It outperforms the baseline and the GN-based models. Overall, our results show the sensitivity of GNN-based approaches to the task representation, the efficacy of relational inductive biases and advocate choosing lightweight approaches that implicitly reason about relations over ones that leave these decisions to human designers.