The Inventory Verification through Detectors on Robotic Inspection Platforms project is developing methods and technologies to enable robotic platforms to autonomously perform Safeguards-relevant inspection tasks. The two inspection tasks we are working to address are automated design verification and radiation-based nu- clear material accountancy inspections. Our approach to development focuses on integrating our Scene Data Fusion (SDF)-based radiation detection systems, with robotic platforms and developing autonomy algorithms within the SDF software ecosystem. The SDF technology, developed at Lawrence Berkeley National Laboratory (LBNL), is approximately ten years old and couples radiation imaging techniques with computer vision tech- nologies to create context-informed 3D maps of radioactivity. It has been most widely applied to contamination mapping and radiation search applications, but has also been applied to object inspection in materials accoun- tancy problems in the past. This project represents LBNL’s first effort at integrating SDF with autonomous systems in a safeguards context. For the design verification task, we focus on indoor applications, where we are integrating our detector systems with a Boston Dynamics Spot robot which is given navigation instructions by the SDF computer. The Spot/SDF system will autonomously create a 3D map of a facility, concurrently note areas that present heightened radiation hazards, and automatically locate clutter objects, as well as pipes and ducts. For accountancy inspection tasks, we plan to leverage previously-developed computer vision technolo- gies to identify and count nuclear material containers, either using the Spot system indoors or a similar small unmanned aerial vehicle (UAV)-based system for outdoor inspections. Once the nuclear material containers are identified, we are developing autonomy algorithms that automatically navigate to the containers and con- duct SDF-based radiation measurements to assess whether the containers’ contents match declarations within pre-specified confidence levels (e.g., empty vs. full). When the number of containers is too high for each to be individually surveyed, we will implement randomized sampling strategies in order to gain various levels of statistical precision on a broad set of items, yet randomly select a subset for additional scrutiny. This project, funded by Defense Nuclear Nonproliferation Research and Development (DNN-R&D), is in its second year of a three-year project.