Autonomous Motion Control

QC 20260907

Abstract

Autonomous trucks are expected to play an important role in future freight transport, but their deployment beyond favorable weather conditions remains an open challenge. Winter roads are especially demanding for semitrailer trucks. Snow and ice reduce tire-road friction, increase stopping distances, limit steering capability, and may lead to vehicle instabilities such as jackknifing or trailer swing. For autonomous driving systems, this means that the vehicle must not only follow a planned trajectory, but also understand when the road surface makes the planned motion unsafe or physically infeasible. 

This thesis studies autonomous motion control of semitrailer trucks in lowfriction conditions. The work focuses on model predictive control, a control framework that uses a vehicle model to predict future motion and optimize steering and acceleration while respecting constraints. A central theme of the thesis is the trade-off between model fidelity and practical applicability. Detailed dynamic models can describe important physical effects, but they may be difficult to use in practice because they require many parameters, additional state estimates, and careful numerical treatment. Simpler kinematic models are easier to use, but may neglect slip-angle and friction effects that become important at higher speeds and on slippery roads. 

The thesis addresses this challenge through four connected studies. First, kinematic and dynamic tractor-semitrailer models are compared for model predictive control. Second, time-optimal lane changes are computed for different friction levels to understand the physical limits of emergency maneuvers. Third, a friction-adaptive nonlinear model predictive controller is developed using an extended kinematic vehicle model and friction-dependent objectives and constraints. Fourth, the controller is evaluated both in simulation and in winter experiments with a full-scale autonomous tractor-semitrailer. 

The results show that friction information is important both for understanding emergency maneuver limits and for designing motion controllers that avoid excessive use of tire forces. In simulations, the proposed controller reduces tireforce demand compared with a more aggressive baseline controller. In realworld double lane-change experiments on snow, it completes the maneuver while a classical tracking-control baseline fails. The thesis therefore demonstrates that friction-adaptive predictive control can improve the safety-relevant behavior of autonomous semitrailer trucks in low-friction conditions. 

Link to DiVA