Parameter and state estimation in Vehicular Systems using nonlinear adaptive observer and data fusion techniques (Project funds: £114,037.36)

  • Herrmann, Guido (Principal Investigator)
  • Phil, Barber (Co-Principal Investigator)
  • Burgess, Stuart C (Co-Investigator)
  • Wragge-Morley, Robert T (Researcher)

Project Details


The vehicle CAN signals (principally engine torque and wheel speed) are to be used to identify the gradient the vehicle is traversing, and the current mass of the vehicle. Adaptive control engineering based machine learning techniques are employed along with data fusion methods to produce a robust real-time estimate (using also accelerometer data). In addition the effect of extreme gradient change dynamics on powertrain based estimation methods is examined.

Key findings

The conclusion is that the uncertainty in the calibration of signals such as torque will ultimately define the calibration of the derived parameters. The impact of certain types of input signal disturbance can be reduced by understanding their relationship with coincident driving events and making use of data fusion or other learning algorithms. Heuristics such as the fact that mass does not significantly change during driving, or that longitudinal acceleration is ostensibly zero whilst the car is stationary, can also be used as premises in learning algorithms.
Effective start/end date1/06/1230/11/15