Bayesian Parameter Estimation for Heavy-Duty Vehicles: SAE Technical Paper No. 2017-01-0528

Eric Miller, Adam Duran, Arnaud Konan

Research output: Contribution to conferencePaper

2 Scopus Citations


Accurate vehicle parameters are valuable for design, modeling, and reporting. Estimating vehicle parameters can be a very time-consuming process requiring tightly-controlled experimentation. This work describes a method to estimate vehicle parameters such as mass, coefficient of drag/frontal area, and rolling resistance using data logged during standard vehicle operation. The method uses Monte Carlo to generate parameter sets which is fed to a variant of the road load equation. Modeled road load is then compared to measured load to evaluate the probability of the parameter set. Acceptance of a proposed parameter set is determined using the probability ratio to the current state, so that the chain history will give a distribution of parameter sets. Compared to a single value, a distribution of possible values provides information on the quality of estimates and the range of possible parameter values. The method is demonstrated by estimating dynamometer parameters. Results confirm the method's ability to estimate reasonable parameter sets, and indicates an opportunity to increase the certainty of estimates through careful selection or generation of the test drive cycle.
Original languageAmerican English
Number of pages8
StatePublished - 2017
EventWCX17: SAE World Congress Experience - Detroit, Michigan
Duration: 4 Apr 20176 Apr 2017


ConferenceWCX17: SAE World Congress Experience
CityDetroit, Michigan

NREL Publication Number

  • NREL/CP-5400-67699


  • Bayesian
  • dynamometer
  • Markov-chain Monte Carlo
  • Metropolis-Hastings
  • parameter estimation
  • vehicle system


Dive into the research topics of 'Bayesian Parameter Estimation for Heavy-Duty Vehicles: SAE Technical Paper No. 2017-01-0528'. Together they form a unique fingerprint.

Cite this