IMU + GNSS + Visual-Odometry for Accurate Localisation

[AdityaNG]

GPS alone for localization won't give great accuracy. Using the raw GPS signals (not just lat, lon; also the raw preprocessed signals) along with IMU (inertial measurement unit) / INS (inertial navigation system) along with vehicle trajectory that comes from monocular SLAM (Visual Odometry), we can merge all the 3 together to get better localization.

Task Details

1. For this, we'll need to log raw GPS signal as well from the Android app. We'll also need to see what frequency works best for this application (is 10Hz fine, or should we go higher).
2. CommaAI has written a Laika, an open-source GNSS processing library for Accurate Localisation. They also have a paper on the same, do refer to it. Figure out how to use this and pass our data into it to see how it behaves
3. Write a wrapper class for this library and use it inside the MergedDatasetIterator. Cache the intermediate calculations, final vehicle trajectory, etc. wherever necessary

Python Library: https://github.com/commaai/laika Blog Post: https://blog.comma.ai/dataset-for-hd-maps-comma2k19 Paper: https://arxiv.org/pdf/1812.05752.pdf

Viewing frusta calculated with sensor fusion algorithm (INS + GNSS + Visual-Odometry). Without map-based corrections (left) and with map-based corrections (right).

[AdityaNG]

Android Raw GNSS Measurements: https://developer.android.com/guide/topics/sensors/gnss

[AdityaNG]

Optimization-Based Visual-Inertial SLAM Tightly Coupled with Raw GNSS Measurements

arXiv:2010.11675v5 [cs.RO] 24 Oct 2021

Abstract— Unlike loose coupling approaches and the EKFbased approaches in the literature, we propose an optimization based visual-inertial SLAM tightly coupled with raw Global Navigation Satellite System (GNSS) measurements, a first attempt of this kind in the literature to our knowledge. More specifically, reprojection error, IMU pre-integration error and raw GNSS measurement error are jointly minimized within a sliding window, in which the asynchronism between images and raw GNSS measurements is accounted for. In addition, issues such as marginalization, noisy measurements removal, as well as tackling vulnerable situations are also addressed. Experimental results on public dataset in complex urban scenes show that our proposed approach outperforms state-of-the-art visual-inertial SLAM, GNSS single point positioning, as well as a loose coupling approach, including scenes mainly containing low-rise buildings and those containing urban canyons.