Synthetic Data for Enhancing Perception and Cooperative Localization in Autonomous Vehicles

Christos Anagnostopoulos, Nikos Piperigkos, Alexandros Gkillas, Aris S. Lalos

Abstract: The design and development of complex cyber-physical systems, as well as systems of systems, demand a broad array of methods and tools, with simulations playing a crucial role in developing and evaluating multiple system aspects. While real-world testing remains essential, simulation offers a valuable supplementary approach, facilitating faster development by providing scalable, reproducible results at minimal cost. In this context, we extend an open-source simulation framework by integrating object detection and Simultaneous Localization and Mapping (SLAM) algorithms, enabling the implementation and evaluation of approaches in cooperative localization and the enhancement of low-quality sensor data. The incorporation of synthetic data in our evaluation pipeline yielded promising results, demonstrating improvements in localization accuracy and sensor performance for autonomous vehicle applications. The dataset is uploaded on the following repository https://gitlab.com/mips_group1/carla-dataset1.git.

External IDs:dblp:conf/aivr/Anagnostopoulos25