Loading…

A framework for joint vehicle localization and road mapping using onboard sensors

This paper presents a modeling framework for joint estimation of a host vehicle state and a map of the road based on global navigation satellite system (GNSS) and camera measurements. We model the road using a spline representation based on lower-dimensional Bézier curves parametrized in generalized...

Full description

Saved in:
Bibliographic Details
Published in:Control engineering practice 2024-12, Vol.153, p.106112, Article 106112
Main Authors: Berntorp, Karl, Greiff, Marcus
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper presents a modeling framework for joint estimation of a host vehicle state and a map of the road based on global navigation satellite system (GNSS) and camera measurements. We model the road using a spline representation based on lower-dimensional Bézier curves parametrized in generalized endpoints (GEPs) with implicit guarantees of continuous lane boundaries. We model the GEPs by a parameter vector having a Gaussian prior representing the uncertainty of the prior map, and provide a systematic way of defining this prior from generic map representations. Both GNSS and camera measurements, such as lane-mark measurements, have noise characteristics that vary in time. To adapt to the changing noise levels and hence improve positioning performance, we formulate the problem as a joint vehicle state, map parameter, and noise covariance estimation problem and present two noise-adaptive linear-regression Kalman filters (LRKFs); (i) an interacting multiple-model (IMM) LRKF and (ii) a variational-Bayes (VB) LRKF. We conduct a Monte-Carlo study and compare the two approaches in terms of estimation precision and computation times. Embedded implementations in an automotive-grade dSpace Micro Autobox-II indicate the real-time validity of both approaches, with turn-around times of between 2–80ms, depending on the problem size and if the map is updated. The results indicate that while the IMM-LRKF shows marginally better estimation accuracy, the VB-LRKF is at least a factor of 2 faster.
ISSN:0967-0661
DOI:10.1016/j.conengprac.2024.106112