Evaluating the next generation of warehouse robotics using generative world models

Warehouses represent complex, dynamic environments requiring efficient navigation and task execution by autonomous robots, with minimal downtime and human intervention. Modern SLAM (Simultaneous Localization and Mapping) techniques can provide a robust mapping of the warehouse itself. However, there are also a huge number of dynamically interacting entities (including other robots, human operators, vehicles, etc). This, coupled with significant variations across regions and customer requirements, can lead to complex emergent situations at deployment that were not considered during development. This project seeks to address these challenges by developing flexible generative world models which are capable of rapidly simulating diverse warehouse environments using varying sensor loadouts and metadata.

By leveraging generative diffusion models, multimodal generative models, and large language models, the project will create adaptive, scalable world models that can simulate and predict warehouse dynamics. This enables robots to train, test, and adapt in virtual environments before deployment, significantly reducing real-world testing requirements and robustness. This digital twinning capability also makes it possible to safely test resilience to rare emergency scenarios, such as hardware failures or human obstacles. 

Academic Institution

This is an opportunity to join the Centre for Vision, Speech and Signal Processing at the University of Surrey. This is the largest such UK institute, and is ranked 1^st for Computer vision research in the UK and 3^rd in Europe (csrankings.org). The supervisory team includes award-winning and world-renowned academics, with the applicant joining a large and tightly knit research team of 10+  peers working on various related topics (http://personalpages.surrey.ac.uk/s.hadfield/).

Industrial Sponsorship and Research Impact

This project is funded at a 50% rate by Locus Robotics. Locus are world-leaders in Warehouse Automation. They have a fleet of more than 50,000 active robots in daily operation around the world, and supply warehouse logistics systems for more than 50 internationally recognisable customers including DHL, UPS and Boots. In addition to funding studentship costs, Locus are also providing access to their hardware, warehouses, and engineer time to support the project. There is also an opportunity to undertake a paid internship in industry, helping build impact for your research activities and getting real-world research experience.

Open to any UK or international candidates. Up to 30% of our UKRI funded studentships can be awarded to candidates paying international rate fees.

You will need to meet the minimum entry requirements for our PhD programme.