Autonomously docking, using object detection, and path planning

Sammanfattning: With today’s climate change Volvo Construction Equipment (CE) must develop machines that emit less carbon dioxide or no carbon dioxide at all. One way to go is to electrify their machines. Volvo CE has developed an autonomous electric machine called HX02. Today the HX02 is charged with an inverted pantograph. This solution is very bulky and expansive to install in worksites. In this master thesis a new charging solution is developed. To use the machine as part of its own charging procedure, meaning that the machine can drive itself into the charging port, with no extra help from a charging station. In this master thesis, the research questions is answered with a case study methodology, with the case study methodology there is a few cases will clear boundaries. The first part of the master thesis is to develop a path planner. Since the machine has the ability to steer all four wheels, one of the research questions was to examine which steering method is the best. When the path planners was made, it could be concluded that the best steering method was the four-wheel together with the crabsteering method. The second part of the master thesis is about how to know where the charging station is. There is a camera on the machine that uses an object detection software. Four different object detection software were tested, where of three of them is based on a Faster-Region Based Convolutional Neural Networks (RCNN) model, based on different data-sets. The fourth was a fiducial marker system, called ArTag. In simulation it was concluded that one of the Faster-RCNN models was the best for accurate detection of the object. In this master thesis it was concluded that to make the machine a part of the charging procedure, a new camera has to be tested, because the current camera had a fish-eye lens and this gave the camera a big error. Due to the large error of the camera, the machine could not dock with the charging station. Therefore, a different type of camera is required to redo the tests.