• Lattice Based Motion Planning
• Model Predictive Control
• Wheeled Robots

Over the past decade, the increased safety standards have accelerated the need for fully automated vehicles operating in harsh environments, both to reduce the labor force and create a safer work environment.
Large, heavy machines are required in these types of environments, but the heavier the vehicle, the lower its maneuverability, and thus the lower the maximum safe speed and possibly productivity. Four-wheel steering vehicles, when compared to the two-wheel ones, have higher maneuverability and high-speed stability. This motivates the popularity of this architecture, especially in these environments. However, most of the current state-of-art motion planning and control algorithms for wheeled vehicles are designed for single-axle steering vehicles.
In this thesis, we propose an autonomous navigation stack for a four-wheel steering vehicle. The work consists of two parts: developing a suitable path planning architecture for these machines to plan a kinematic feasible path, and developing a controller to track it.
The resulting architecture: 1) extends the lattice-based motion planning method for four-wheel steering vehicles by proposing and validating several steering functions to generate the vehicle’s motion primitive, and 2) leverages a nonlinear model predictive controller capable of avoiding collisions with obstacles in a dynamic environment. We validate the approach via simulations on a digital twin of the Volvo HX01 hauler moving in both benchmark and realistic environments.
The work was done in collaboration with the University of ̈Orebro, which provided the requirements and a starting framework, i.e., the navigation-oru stack, designed for single-axle vehicles.