An End-to-End System for Crowdsourced 3d Maps for Autonomous Vehicles: The Map** Component
Authors:
Onkar Dabeer,
Radhika Gowaikar,
Slawomir K. Grzechnik,
Mythreya J. Lakshman,
Gerhard Reitmayr,
Kiran Somasundaram,
Ravi Teja Sukhavasi,
Xinzhou Wu
Abstract:
Autonomous vehicles rely on precise high definition (HD) 3d maps for navigation. This paper presents the map** component of an end-to-end system for crowdsourcing precise 3d maps with semantically meaningful landmarks such as traffic signs (6 dof pose, shape and size) and traffic lanes (3d splines). The system uses consumer grade parts, and in particular, relies on a single front facing camera a…
▽ More
Autonomous vehicles rely on precise high definition (HD) 3d maps for navigation. This paper presents the map** component of an end-to-end system for crowdsourcing precise 3d maps with semantically meaningful landmarks such as traffic signs (6 dof pose, shape and size) and traffic lanes (3d splines). The system uses consumer grade parts, and in particular, relies on a single front facing camera and a consumer grade GPS. Using real-time sign and lane triangulation on-device in the vehicle, with offline sign/lane clustering across multiple journeys and offline Bundle Adjustment across multiple journeys in the backend, we construct maps with mean absolute accuracy at sign corners of less than 20 cm from 25 journeys. To the best of our knowledge, this is the first end-to-end HD map** pipeline in global coordinates in the automotive context using cost effective sensors.
△ Less
Submitted 31 March, 2017; v1 submitted 29 March, 2017;
originally announced March 2017.
