Autonomous Path Planning &
Parking Assistance System
Autonomous vehicles have, during the past years, become a more interesting topic for vehicle manufacturers due to the increased safety, efficiency and reduced environmental impact that this type of technology can provide. Autonomous driving is especially suitable for driving trucks forward and backwards in a narrow area when high precision is required.
The system is a LEGO truck constisting of a tractor and a semitrailer connected via a dolly. Mounted on it are motors, an RPi an EV3-unit and a battery pack.
A GUI has been developed for easy simulations. It visualizes the world that the truck operates in and displays the paths which the controller is following. It is possible to select a start position along with a goal position. It is also capable of simulating the calibration routine.
The visualization system projects a map on the floor in Visionen where all the different obstacles and a goal are shown. As the truck performs a task, executed on an RPi 4, its position and paths are continuously updated and displayed.
A path following MPC controller has been developed. The MPC controller minimizes the path following errors, i.e. the distance to the path, the orientation error and the joint angle errors of the truck. With the introduced MPC controller we see significant improvement in both performance and robustness compared to previous LQ controller. The truck can now complete missions where the LQ controller fails.
While the truck is moving it is possible that it deviates from the planned path shown in green bellow. The MPC controller avoids obstacles by adding constraints to the trailer’s lateral deviation from the path using ray tracing. The result can be seen as a tunnel that the truck is constrained to move within.
A calibration phase for the steering angle was implemented using a PI controller. The truck’s steering angle sensor gives only relative measurements, which means that steering angle may be biased at start-up. The calibration phase rapidly estimates this bias.