Density-based Feasibility Learning with Normalizing Flows for Introspective Robotic Assembly

Abstract

Machine Learning (ML) models in Robotic Assembly Sequence Planning (RASP) need to be introspective on the predicted solutions, i.e. whether they are feasible or not, to circumvent potential efficiency degradation. Previous works need both feasible and infeasible examples during training. However, the infeasible ones are hard to collect sufficiently when re-training is required for swift adaptation to new product variants. In this work, we propose a density-based feasibility learning method that requires only feasible examples. Concretely, we formulate the feasibility learning problem as Out-of-Distribution (OOD) detection with Normalizing Flows (NF), which are powerful generative models for estimating complex probability distributions. Empirically, the proposed method is demonstrated on robotic assembly use cases and outperforms other single-class baselines in detecting infeasible assemblies. We further investigate the internal working mechanism of our method and show that a large memory saving can be obtained based on an advanced variant of NF.

Method

At the top, t-distributed Stochastic Neighbor Embedding (t-SNE) shows that samples mapped by NF are ”normalized” and pulled together to a compact cluster. At the bottom, Cosine Similarities between feasible and infeasible assemblies are more distinct after the transformation, verifying the ”normalization”.

Bibtex:

@inproceedings{feng2023density,

   title = {Density-based Feasibility Learning with Normalizing Flows for Introspective Robotic Assembly},

   author = {Feng, Jianxiang and Atad, Matan and Rodriguez Brena, Ismael Valentin and Durner, Maximilian and Triebel, Rudolph},

   booktitle = {18th Robotics: Science and System 2023 Workshops},

   url = {https://arxiv.org/abs/2307.01317},

   year = {2023},

   keywords = {Feasibility learning; Normalizing Flows;}

}