ABOUT

BMW has partnered with NVIDIA to create digital twins of BMW factories (using NVIDIA technology) to digitise their factory and logistics planning efforts, now and into the future. The purpose of this is to train Artificial Intelligence (AI) systems in a 1:1 scale digital environment without disrupting production on the factory floor. The training data learnt in the digital environment is implemented to a fleet of Autonomous Mobile Robots (AMRs) in the real world. The benefits that digital twins provide include factory layout planning, training workers in VR, optimising workflows and mitigating any errors or risk, running endless simulations of scenarios and technology, integrating metadata, and significantly lowering production costs and environmental impact.

Central to training the AI systems is object recognition. In short, this means that engineers can ‘tell’ the robot if an object is a vehicle, person, container, dolly etc., and how it should interact with each object. The robot continues to develop its object recognition capability as it is trained within the Digital Twin, which uses digital replicas (3D assets) of real-life objects within BMW factories. This enables autonomous robot to identify different objects and make its own decisions in real-time when navigating spaces and performing tasks.

QUT’s involvement in this project has been the creation of a large 3D asset library for software engineers at BMW Group and idealworks. By creating realistic BMW factory 3D models, such as storage containers, dollies, shelving and industrial vehicles, BMW Group can utilise these 3D assets to train AMRs for object recognition within the AI software that it uses. Object recognition is critical for AMRs as they need to perform actions based on real-time data received through various sensors.