On the 24th of January 2024, AI4CSM partner Gerhard Dorn from Virtual Vehicle Research GmbH gave a seminar talk at Silicon Austria Labs about how to realize perception for autonomous driving within AI4CSM using the programming language Julia.
In the AI4CSM project, which aims to enable smart-connected mobility using ECAS vehicles, it is crucial to realize autonomous driving edge perception - the capability to make trustworthy, real-time perception decisions. In the project, Virtual Vehicle Research GmbH develops a prototype Robotaxi with self-developed edge perception in a closed-loop virtual simulation (using CARLA) and real environment (Ford Mondeo).
Gerhard Dorn's talk illustrated:
- the challenge in edge perception starting from the current sensors available to a data analysis of typical Lidar point clouds
- the chosen approach to process data using ML learning on the GPU with the relative new programming language Julia
• introduction to a probabilistic occupancy grid used to map Lidar point clouds to a 2D map.
More information about CARLA simulator: https://carla.org/
More information about Julia Programming language: https://julialang.org/