VILAM: Infrastructure-assisted 3D Visual Localization and Mapping for Autonomous Driving

Explore a conference talk on VILAM, an innovative framework for enhancing autonomous driving systems through infrastructure-assisted 3D visual localization and mapping. Learn how this approach leverages intelligent roadside infrastructures to achieve high-precision and globally consistent localization and mapping for autonomous vehicles. Discover the novel elastic point cloud registration method that enables independent optimization of different parts of the local map, overcoming unique deformation challenges. Understand the lightweight factor graph construction and optimization technique used to correct vehicle trajectory and efficiently reconstruct consistent global maps. Examine the implementation of VILAM on a real-world smart lamppost testbed across multiple road scenarios, and review the impressive results showing decimeter-level localization and mapping accuracy using consumer-level onboard cameras. Gain insights into the robustness of this system under diverse road conditions and access a video demonstration of VILAM in action.

NSDI '24 - VILAM: Infrastructure-assisted 3D Visual Localization and Mapping for Autonomous Driving