- 01Reads project dataStages, timesheets, costs, tests and documents recorded in Paylists.
- 02Searches published workOpenAlex, Europe PMC, PubMed, arXiv, Crossref, OSTI, NASA NTRS, NIST and other sources.
- 03Recommends answersEach answer is supported by references for your team and R&D tax advisor to review.
What is the R&D aim of the project?
Lister recommended answerTo develop a robotic perception system capable of estimating object pose accurately enough for autonomous picking under changing real-world conditions.
What was the scientific or technological uncertainty?
- 01
Describe exactly which uncertainties you need to resolve
Lister recommended answerWhether pose accuracy can remain within the required limit when lighting, clutter, surface reflectivity, orientation and object geometry change.
- 02
Describe what made it a project to resolve a specified uncertainty
Lister recommended answer3D cameras and object-pose algorithms already existed, but published performance could not establish robustness under the project's nuisance conditions.
- 03
Describe the steps you plan to take to resolve them
Lister recommended answerEstablish ground-truth poses, vary lighting and object presentation, test alternative filtering and registration algorithms and validate on unseen objects.
- 04
What difficulties are foreseen for the project
Lister recommended answerReflective surfaces, occlusion, low geometric texture and inconsistent depth data can cause systematic pose errors.
- 05
Information about any major changes to the aims, or new uncertainties that have been encountered, while running the project.
Lister recommended answerTesting may reveal that the dominant limitation changes from sensor noise to segmentation or registration under specific object classes.
- 06
How the uncertainties and other challenges were overcome (or not).
Lister recommended answerChanges would be frozen and independently validated against new objects and changed nuisance conditions.
What advance in science or technology was sought?
- 01
A review of the current state of knowledge
Lister recommended answerNIST had already documented significant robotic-perception performance challenges involving lighting, geometry and object properties.
- 02
How the project sought an advance in a qualifying area of science or technology
Lister recommended answerThe project sought a measurably more robust perception capability across the defined picking envelope.
What was the result, and what evidence is still needed?
Result
The final system met pose accuracy on the validated object and lighting classes but retained an identified boundary for highly reflective objects.
Conclusion
The advance is a bounded perception capability, not the general use of AI or machine vision.
Evidence still needed
Ground-truth methodology, sensor configuration, object set, lighting conditions, error distributions, failed picks and frozen model versions.
Supporting references
NIST, A Standards Roadmap for 3D Imaging in Robotic Assembly Applications.
nist.gov ↗NIST, Filtering Organized 3D Point Clouds for Bin Picking Applications.
nist.gov ↗NIST, Perception Performance of Robotic Systems.
nist.gov ↗
What does Paylists record for this project, and who reviews it?
Paylists records the project stages, timesheets, costs, tests and documents, and Lister recommends an answer to each Additional Information Form question with supporting references. Your team, your competent professional and your R&D tax advisor can log in to review them.