Combined Tire Friction Modeling for Path Tracking Control at the Limits of Handling

Institut

Lehrstuhl für Regelungstechnik

Typ

Masterarbeit /

Inhalt

theoretisch /  

Beschreibung

Problem description: Autonomous driving is a well established field promising increased
safety, comfort and economics of transportation. It is however a complex task requiring the
successful interaction of different software modules. Usually, the task is split into the sub-
tasks of perception, planning and control. This Master’s Thesis focuses on the control part,
where a pre-defined collision-free, drivable path is provided and the task is actuation of throttle,
brake and steering to track this reference while keeping the vehicle stable and within its limits [1].
Model Predictive Control (MPC) is a modern optimization-based control method which works
by repeatedly solving discrete-time open-loop optimal control problems over a finite horizon and
each applying only the first step and re-optimizing at the next step [4]. MPC allows feasible
approximation of optimal controllers for nonlinear systems under constraints. This makes it
suitable to apply for autonomous vehicle path control near the limits of handling where nonlin-
earities and coupled constraints inherent in the vehicle dynamics are relevant to consider in the
controller design.
MPC relies on a model to predict system behavior, which includes a model of the tire-road
interaction. There exist multiple approaches to model tires including neglecting them and using
friction limits instead [6]. The competing approach is to include a lateral tire model connect-
ing lateral slip angles and lateral forces at the tire, which allows the MPC to correctly predict
behavior beyond the limits such as drifting [3]. This Nonlinear MPC (NMPC) is therefore able
to prevent or recover from these undesirable states. The problem arises that this more sophisti-
cated approach of using nonlinear lateral tire curves however still requires an additional coupled
friction constraint to properly incorporate the limit of combined longitudinal/lateral forces that
tires are naturally subject to.
This thesis aims to take a step away from constraints towards a model more closely modeling
the physical vehicle. The goal is to utilize a combined friction model for the tires linking lateral
and longitudinal slips with respective forces at the tire or axis [2]. The more accurate com-
bined tire model would allow the MPC formulation to omit any artificial friction constraints and
incorporate the tire limitations in a natural way within the prediction model. This improved
prediction model promises more optimal control strategies at the combined limits as well as the
natural integration of a function resembling an electronic stability program (ESP) directly into
the tracking controller.
For implementation, the software package acados [5] offers fast compilable code to solve
optimal control problems arising from MPC formulations and offers interfaces to the languages
Python and C. Simulations in an existing Python-environment are to be performed showing the behavior of the resulting MPC in safety-critical situations.

 

Voraussetzungen

• Background in systems & control theory (including MPC), as well as optimization,
• Knowledge about vehicle dynamics is recommended,
• Good programming skills in Python,
• Analytical, problem solving mindset and a high degree of autonomy.

Möglicher Beginn

sofort

Kontakt

Joscha Bongard
Raum: MW0205
Tel.: 089/28915679
joscha.bongardtum.de

Ausschreibung