On the move indoors and outdoors with a dynamic outdoor robot

Many driverless transport vehicles today are designed for low dynamics and are designed for either the interior or exterior of company premises. They have either high performance, dynamics or flexibility. So far, all three characteristics together cannot be found in practically any single vehicle.

Researchers at the Fraunhofer Institute for Material Flow and Logistics (IML) have recently demonstrated a highly dynamic robot that can transport large pallet loads omnidirectionally at high speed. It is designed to leave the protected and defined environment of warehouses and to operate throughout a company’s premises.

The O³dyn vehicle’s navigation works seamlessly at the transition between indoor and outdoor areas. It is implemented using environment-based and radio-based localisation algorithms.

The mobile transport robot was developed at Fraunhofer IML using an AI-based simulation — a completely new process that the researchers also presented at the digital NVIDIA GTC developer conference in March.

Highly complex processes are simulated in real time through highly parallel processing of modern graphics cards and form the basis of a new league of algorithms known as simulation-based AI. Simulations use models for abstraction. For the development of such highly dynamic systems as the two transport robots, the behaviour of the simulated transport vehicle is synchronised with that of the real transport vehicles in a special, particularly suitable test environment, with the help of high-performance motion capturing. This way, the simulation model is optimised. If the difference between the model and reality is reduced, the simulation turns into a digital reality for the AI and the robot becomes the digital twin (cyber-physical system) of the simulation.

Using traditional methods, the development of new robots takes several years. With the help of AI-based simulation, however, the robots can be trained with new requirements from intralogistics during the development period. Due to the digital reality approach, the development of the hardware can be decoupled from the programming of the system behaviour during the development period. If the robots then go into production, they are still state of the art or at the level of the latest intralogistics technology — a significant added value compared to classic robot development.

Researchers at Fraunhofer IML had already used this technology for the first time as part of the LoadRunner project in order to develop the AI for swarm control purely virtually, and thus opened up a completely new branch of research.