Authors on arXiv eess.SY report improved low-speed electric-vehicle velocity estimates with a hybrid UKF

Authors on arXiv eess.SY report improved low-speed electric-vehicle velocity estimates with a hybrid UKF

The authors describe two ways to estimate an electric vehicle’s longitudinal velocity below 1 km/h, where production wheel-speed sensors lose resolution. Their unscented Kalman filter (UKF) combines sensor readings with a vehicle model. One approach models the drivetrain as a torsional spring-damper, rather than treating the motor and wheels as rigidly connected; the other uses a neural network-based virtual wheel-speed sensor as a pseudo-measurement. The authors say both approaches reduce velocity root-mean-square error against two baseline configurations in real-world test drives with two series-production electric vehicles. They also report that the estimates carried over to the second vehicle without re-tuning on its data. This matters for automated parking and traffic-jam assistance, which operate at low speeds: the paper addresses a measurement limitation in that operating range. The reported comparison covers two vehicles and an inertial reference; it does not establish how the approaches perform across a wider range of vehicles or conditions.

Publication: 866d03ec-9a9d-49ea-8682-a25e8f0632b7

attribution — pass: The title, summary and claims attribute the methods and results to the paper's authors on arXiv eess.SY. The performance and transfer results remain the authors' reported findings, not established facts.

Authors:Hendrik Schäfke, Daniel O. M. Weber, Thomas Seel, Askar Vagapov

summary — uncertain: The summary omits the central mechanism: the electric motor's resolver gives finely resolved motor speed down to standstill, but that speed reaches the wheels through a compliant drivetrain. Without this, a general reader cannot tell why drivetrain modeling helps or what the neural virtual sensor uses. It also drops the reported magnitudes and leaves root-mean-square error and pseudo-measurement unexplained. Smallest repair: add one sentence on the resolver and compliant drivetrain, give the reported percentages, and add tooltips for root-mean-square error, pseudo-measurement and longitudinal velocity.

The motor speed from the electric machine's resolver stays finely resolved down to standstill but reaches the wheels only through a compliant drivetrain.

date — pass: The originalPublished date, 2026-11-05, matches the arXiv v1 submission date. The feed date of 2026-11-06 is later and correctly not used.

[Submitted on 5 Nov 2026]

support — pass: The abstract supports the claims: sensor resolution loss below 1 km/h, the two estimation routes, error reduction against both baselines on both vehicles, and zero-shot transfer without re-tuning. The significance sentence is limited to the stated measurement limitation for parking and traffic-jam assistance. It does not claim a demonstrated benefit to those functions. The stated limitation to two vehicles is fair.

Carried over zero-shot to Vehicle B without re-tuning on target-vehicle data, they lower it by at least 41% and 51%.

media — pass: The selected media is the verified arXiv abstract page for this paper. It is the direct original source.