Abstract:
Over the past few decades, autonomous vehicles have been widely considered as the next
generation of road transportation. As a result, relevant technology has been rapidly developed, and one specific topic is enabling autonomous vehicles to operate under demanding
conditions. This requires the autonomous driving controller to have a good understanding of the vehicle dynamics at the limits of handling, and is expected to improve the
performance as well as safety of autonomous vehicles especially in extreme situations.
Furthermore, there has been application of techniques such as torque vectoring and four-
wheel steering on modern vehicles as part of the driver assistance system, while such
multi-actuation can be deployed on an autonomous vehicle in order to further enhance its
performance in response to challenging manoeuvres and scenarios.
This thesis aims to develop a real-time path tracking control strategy for an autonomous
electric vehicle at the limits of handling, taking advantage of torque vectoring and four-
wheel steering techniques for the enhanced control of vehicle dynamics. A nonlinear
model predictive control formulation based on a three degree-of-freedom vehicle model
is proposed for control design, which takes into account the nonlinearities in vehicle dynamics at the limits of handling as well as the crucial actuator constraints. In addition,
steady-state references of steering inputs as well as vehicle states are generated based
on a bicycle model and included in the control formulation to improve the performance.
Two path tracking models with different coordinate systems are introduced to the control
formulation, and compared to understand the more suitable one for the proposed path
tracking purpose. Then the path tracking performance with different levels of actuation
is investigated. According to the high-fidelity simulation results, the vehicle achieves the
minimum lateral deviation with the over-actuation topology including both torque vectoring and four-wheel steering, which illustrates that the over-actuation formulation can
enhance the path tracking performance by enduing the vehicle with the best flexibility as
well as stability during operation at the limits of handling.
Before being implemented on the vehicle, the performance of the proposed control strategy is further assessed with regards to real-time operation. After evaluating the control performance with different prediction horizons and sampling time, the most suitable
setup is identified which compromises between the control performance and the capability of real-time execution. Finally, the control algorithm is implemented on a real vehicle
for practical testing. The controller is tested in four different scenarios, and the results
demonstrate that the proposed controller is capable of path tracking control and vehicle
stabilisation for multi-actuated autonomous vehicles at the limits of handling.
In general, this thesis has proposed a path tracking controller for autonomous vehicles
which takes into account nonlinear vehicle dynamics at the limits of handling. Following
some necessary simplification, the developed controller has been successfully deployed
on a real vehicle in real time, and the control performance has been validated in several
challenging scenarios. The controller proves itself to be able to improve the vehicle’s
flexibility as well as to stabilise the vehicle at the limits of handling, and furthermore, it is
able to accommodate relatively large side slip angles during the demanding manoeuvres
as well.