A Workflow for Map Creation in Autonomous Vehicle Simulations

TL;DR

This paper introduces a streamlined workflow for creating simulation-ready maps for autonomous vehicle testing, reducing resource requirements and increasing flexibility across different simulators.

Contribution

The paper presents a novel, resource-efficient workflow for map creation in AV simulations, demonstrated with a 3D parking lot map, and discusses future enhancements including SLAM integration.

Findings

Successfully generated a 3D parking lot map for AV simulation

Workflow reduces computational resources compared to existing methods

Future plans include SLAM integration and broader simulator compatibility

Abstract

The fast development of technology and artificial intelligence has significantly advanced Autonomous Vehicle (AV) research, emphasizing the need for extensive simulation testing. Accurate and adaptable maps are critical in AV development, serving as the foundation for localization, path planning, and scenario testing. However, creating simulation-ready maps is often difficult and resource-intensive, especially with simulators like CARLA (CAR Learning to Act). Many existing workflows require significant computational resources or rely on specific simulators, limiting flexibility for developers. This paper presents a custom workflow to streamline map creation for AV development, demonstrated through the generation of a 3D map of a parking lot at Ontario Tech University. Future work will focus on incorporating SLAM technologies, optimizing the workflow for broader simulator compatibility, and exploring more flexible handling of latitude and longitude values to enhance map generation accuracy.