On the challenges of reversing multi-articulated vehicles

Licentiate thesis, 2026

Reversing articulated heavy vehicles, particularly multi-articulated ones, is a difficult task for drivers. The task is challenging in two aspects. First, the articulation angles are unstable while driving backwards, requiring active steering control from the leading unit to avoid inter-unit clashes. Second, the travel direction of the last unit is indirectly controlled, requiring a thorough understanding of the articulated vehicle's complex dynamics. Many driver-assistance systems and autonomous functions have already been developed for this or similar tasks. As many of them are based on vehicle models, two typical models of articulated vehicles are presented and discussed in detail, namely, a kinematic model and a dynamic model. The existing functions and research in reversing control are divided into three categories: articulation stabilization, path planning, and path following, which are thoroughly reviewed here.

The present thesis covers three research directions within reversing control. The first direction is a study aimed at explaining why the rearview dynamic guideline is insufficient for assisting articulated vehicles and why driver aids related to control and planning, such as those in the three categories above, are needed. The present thesis also aims to identify directions for potential improvements based on existing research and promote future developments in those areas, so that a driver-assistance system can be attractive enough for industrialization. The second direction is based on a study that develops a new geometrical method to determine the fundamental limitation of articulation stabilization for vehicles with up to three articulations. The determined limitation provides a feasible range for articulation stabilization control to enhance performance and coverage. Path-following solutions can benefit from improved articulation stabilization. Vehicle poses that fall outside the fundamental limitation can be used for the last few meters to reach final poses, which opens more possibilities for path planning. The third and final direction presents a preliminary study conducted to quantify the mismatch between the widely used kinematic model and the actual vehicle dynamics at low speed using a single-articulated vehicle. The preliminary study shows the potential to improve the performance of model-based assistant solutions from a vehicle modeling perspective, and it also promotes ongoing research on multi-articulated vehicles.

The driver aid that can fully eliminate the challenge of reversing posed by the complex motion of articulated vehicles must involve controllers that assist with path planning and following. Presenting predicted vehicle motion based on current vehicle states and driver input as dynamic guidelines, along with a top-view, can assist the driver with collision avoidance, but not with maneuvers that change the steering angle. Most existing research has developed path-following and planning controllers based on the kinematic model. Before those controllers can be industrialized, it is critical to determine the limit in real operations at which the vehicle deviates from the model to the extent that the control approach becomes insufficiently robust. Scenarios outside the limit need to be handled with either more accurate models or more robust control approaches. A driver aid must be able to cover sufficient scenarios that can occur in real operations to be considered well-developed and attractive for industrialization, with additional costs to the vehicle.