TEDi Policy: Temporally Entangled Diffusion for Robotic Control

Diffusion models have been shown to excel in robotic imitation learning by mastering the challenge of modeling complex distributions. However, sampling speed has traditionally not been a priority due to their popularity for image generation, limiting their application to dynamical tasks. While recent work has improved the sampling speed of diffusion-based robotic policies, they are restricted to techniques from the image generation domain. We adapt Temporally Entangled Diffusion (TEDi), a framework specific for trajectory generation, to speed up diffusion-based policies for imitation learning. We introduce TEDi Policy, with novel regimes for training and sampling, and show that it drastically improves the sampling speed while remaining performant when applied to state-of-the-art diffusion-based imitation learning policies.

Blazer: Laser Scanning Simulation using Physically Based Rendering

Line laser scanners are a sub-type of structured light 3D scanners that are relatively common devices to find within the industrial setting, typically in the context of assembly, process control, and welding. Despite its extensive use, scanning of some materials remain a difficult or even impossible task without additional pre-processing. For instance, materials which are shiny, or transparent. In this paper, we present a Blazer, a virtual line laser scanner that, combined with physically based rendering, produces synthetic data with a realistic light-matter interaction, and hence realistic appearance. This makes it eligible for the use as a tool in the development of novel algorithms, and in particular as a source of synthetic data for training of machine learning models. Similar systems exist for synthetic RGB-D data generation, but to our knowledge this the first publicly available implementation for synthetic line laser data. We release this implementation under an open-source license to aid further research on line laser scanners.