Load torque estimation for an automotive electric rear axle drive by means of virtual sensing using Kalman filtering

Robert Kalcher*, Katrin Ellermann, Gerald Kelz

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


Load torque signal information in hybrid or battery electric vehicles would be beneficial for control applications, extended diagnosis or load spectrum acquisition. Due to the high cost of the sensor equipment and because of the inaccuracies of state-of-the-art estimation methods, however, there is currently a lack of accurate load torque signals available in series production vehicles. In response to this, this work presents a novel model-based load torque estimation method using Kalman filtering for an electric rear axle drive. The method implements virtual sensing by using measured twist motions of the electric rear axle drive housing and appropriate simulation models within a reduced- order unscented Kalman filter. The proposed method is numerically validated with help of sophisticated multibody simulation models, where influences of hysteresis, torque dynamics, road excitations and several driving manoeuvres such as acceleration and braking are analysed.

Original languageEnglish
Pages (from-to)1-30
Number of pages30
JournalInternational Journal of Vehicle Performance
Issue number1
Publication statusPublished - 2022


  • Battery electric vehicles
  • Bev
  • Electric rear axle drive
  • Hev
  • Hybrid electric vehicles
  • Kalman filtering
  • Load torque estimation
  • Mbs
  • Multi-body simulations
  • Reduced-order unscented kalman filter
  • Roukf
  • Ukf
  • Unscented kalman filter
  • Vehicle systems modelling
  • Virtual sensing

ASJC Scopus subject areas

  • Modelling and Simulation
  • Automotive Engineering
  • Fuel Technology
  • Safety, Risk, Reliability and Quality
  • Mechanics of Materials
  • Mechanical Engineering
  • Computer Science Applications


Dive into the research topics of 'Load torque estimation for an automotive electric rear axle drive by means of virtual sensing using Kalman filtering'. Together they form a unique fingerprint.

Cite this