Distributed Intelligence: Service-Oriented and Learned Trajectory Planning

Summary

The function offloading is a central use case of the distributed, service-oriented system architecture of autotech.agil. Instead of calculating the trajectories internally, the vehicles use an external planning service provided by the infrastructure. This strategy saves computing resources in the vehicle, reduces energy consumption and enables more efficient vehicle control – especially in localized areas with corresponding service availability.

State of the Art

A holistic framework for the service-oriented function offloading – to the infrastructure, for example – does not yet exist. Existing research focuses primarily on the theoretical optimization of decision-making processes, whether and to what extent the outsourcing of functions can lead to a more efficient use of resources. However, a complete implementation and integration of such concepts in real, connected systems has not yet been demonstrated.

Technological Innovations

For the first time, we are demonstrating a holistic concept for function offloading that was designed, researched and prototypically implemented as part of autotech.agil. The underlying service-oriented architecture enables a high degree of interoperability: the same external trajectory planner can be used by two research vehicles with different automation solutions. This underlines the potential of distributed systems for flexible, scalable and efficient vehicle automation.