A Backlash Compensator for Drivability Improvement Via Real-Time Model Predictive Control

Cristian Rostiti, Yuxing Liu, Marcello Canova, Stephanie Stockar, Gang Chen, Hussein Dourra, Michael Prucka

Research output: Contribution to journalArticlepeer-review

18 Scopus citations


Nonlinear dynamics in the transmission and drive shafts of automotive powertrains, such as backlash, induce significant torque fluctuations at the wheels during tip-in and tip-out transients, deteriorating drivability. Several strategies are currently present in production vehicles to mitigate those effects. However, most of them are based on open-loop filtering of the driver torque demand, leading to sluggish acceleration performance. To improve the torque management algorithms for drivability and customer acceptability, the powertrain controller must be able to compensate for the wheel torque fluctuations without penalizing the vehicle response. This paper presents a novel backlash compensator for automotive drivetrain, realized via real-time model predictive control (MPC). Starting from a high-fidelity driveline model, the MPC-based compensator is designed to mitigate the drive shaft torque fluctuations by modifying the nominal spark timing during a backlash traverse event. Experimental tests were conducted with the compensator integrated into the engine electronic control unit (ECU) of a production passenger vehicle. Tipin transients at low-gear conditions were considered to verify the ability of the compensator to reduce the torque overshoot when backlash crossing occurs.

Original languageEnglish (US)
Article number104501
JournalJournal of Dynamic Systems, Measurement and Control, Transactions of the ASME
Issue number10
StatePublished - Oct 1 2018

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Information Systems
  • Instrumentation
  • Mechanical Engineering
  • Computer Science Applications


Dive into the research topics of 'A Backlash Compensator for Drivability Improvement Via Real-Time Model Predictive Control'. Together they form a unique fingerprint.

Cite this