Skip to main content
To KTH's start page To KTH's start page

Exploiting over-actuation for improved active safety of autonomous electric vehicles

Time: Mon 2022-06-13 13.00

Location: E3, Osquars backe 14, Stockholm

Language: English

Subject area: Vehicle and Maritime Engineering

Doctoral student: Wenliang Zhang , Fordonsdynamik

Opponent: Professor Dr.-Ing. Steffen Müller, Berlin Institute of Technology

Supervisor: Lars Drugge, Fordonsdynamik; Associate Professor Mikael Nybacka, Fordonsdynamik

Export to calendar

QC 20220513


The increasing demand for road vehicles has led to challenging road safety and environmental issues. The deployment of active safety systems and autonomous vehicles can contribute to safer roads by assisting or replacing human drivers in the task of maintaining vehicle control in critical conditions: e.g., an obstacle-avoidance manoeuvre. Road vehicle electrification can bring about environmental benefits and at the same time enable the development of over-actuated vehicle platforms. Over-actuation can be explored together with active safety and automated driving systems in order to enhance vehicle safety. On the other hand, to achieve their best possible performance, such safety and automated systems require the knowledge of vehicle states such as sideslip angle as well as reliable trajectories. However, measuring such crucial states can be overly expensive on production vehicles.

The studies presented in this thesis aim to explore how over-actuation can improve path-following and yaw-stability performance of autonomous electric vehicles in critical manoeuvres and investigate the associated state estimation and trajectory planning problems.

To achieve these goals, this thesis focuses on five aspects. First, it explores vehicle dynamics modelling by introducing vehicle and tyre models of various levels of complexity. In particular, the camber effect on lateral tyre forces was modelled using a simple, yet effective, component, which allows for individual camber control of each wheel. Second, it addresses the state estimation problem by designing and evaluating three moving horizon estimation (MHE) based estimators and an unscented Kalman filter. The evaluation in three critical manoeuvres showed that the estimator MHE outperformed the other algorithms, with improved convergence rate, accuracy and response to external disturbances and modelling errors, due to its consideration of a sequence of most recent measurements and process noises. Third, trajectory planning is studied through optimal control formulations and by examining the effect of model complexity in critical driving scenarios. It was shown that the advanced double-track planner with load transfer and the Magic Formula tyre model was desired to achieve more consistent trajectory planning and tracking performance as well as smaller peak yaw rate and sideslip angle. Fourth, the path following and yaw stability problem is tackled in the model predictive control framework and by exploring various over-actuation configurations -- active front steering (AFS), torque vectoring (TV) and active camber (AC). The results in safety-critical conditions showed that AFS + TV improved yaw stability, path following and passing velocity compared to AFS, and AFS + AC performed better than AFS + TV. The integrated control of AFS + TV + AC further enhanced vehicle safety and was more robust to reference trajectory variations, as a result of its more effective actuator and tyre utilisation. Finally, this work details the framework for optimal control implementation, which facilitates efficient computing, smooth parameter tuning and results analysis, as well as sustainable code development.

The research presented in this thesis has contributed to the modelling, formulation and control of autonomous electric vehicles by exploiting over-actuation for enhanced vehicle safety. It has been shown that over-actuation control strategies can be a promising solution for improving active safety, and thus they contribute to a safer and more sustainable future transport.