Deep neural network based inertial odometry using low-cost inertial measurement units
- Submitting institution
-
University of Oxford
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11840
- Type
- D - Journal article
- DOI
-
10.1109/tmc.2019.2960780
- Title of journal
- IEEE Transactions on Mobile Computing
- Article number
- -
- First page
- 1
- Volume
- N/A
- Issue
- -
- ISSN
- 1536-1233
- Open access status
- Compliant
- Month of publication
- December
- Year of publication
- 2019
- URL
-
-
- Supplementary information
-
https://www.youtube.com/watch?v=L5LtE-PQuHk&feature=youtu.be
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
4
- Research group(s)
-
-
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper, extending our AAAI’18 conference version, presents the first deep learning approach to inferring the motion of a device based on inertial (accelerometer and gyroscope) data. It is the first method to break the cycle of continuous error propagation used in physics-based odometry, by learning motion characteristics and systematic error drift simultaneously. Benefits include applicability to multiple types of devices/attachments (wearables, robots, trolleys); and robustness to non-periodic motion and highly dynamic conditions, including those encountered in extreme emergency environments. Navenio Ltd recently licensed the patent application to build an asset-tracking product for hospitals.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
